There is a growing view that intelligence evolved as a product of social

interdependence. The unique development of human intelligence was

probably linked to the use of spoken language, but language itself
evolved in the context of social interaction, and in its development it has
shaped - and been shaped by - social institutions. Taking as their
starting-point the social production of intelligence and of language,
scholars across a range of disciplines are beginning to rethink fundamen-
tal questions about human evolution, language and social institutions.
This volume brings together anthropologists, linguists, primatologists
and psychologists, all working on this new frontier of research.
Social intelligence and interaction
Social intelligence and interaction

Expressions and implications of the social

bias in human intelligence

EDITED BY
ESTHER N. GOODY
University of Cambridge

CAMBRIDGE
UNIVERSITY PRESS
CAMBRIDGE UNIVERSITY PRESS
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, Sao Paulo

Cambridge University Press

The Edinburgh Building, Cambridge CB2 8RU, UK

Published in the United States of America by Cambridge University Press, New York

www.cambridge.org
Information on this title: www.cambridge.org/9780521453295
© Cambridge University Press 1995

This publication is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.

First published 1995

A catalogue record for this publication is available from the British Library

Library of Congress Cataloguing in Publication data

Social intelligence and interaction: expressions and implications of
the social bias in human intelligence / edited by Esther N. Goody.
p. cm.
Based on papers presented at a workshop sponsored by the
Wissenschaftskolleg zu Berlin, held in the spring of 1990, with the
theme Some implications of a social origin of human intelligence.
ISBN 0 521 45329 1. - ISBN 0 521 45949 4 (pbk.)
1. Intellect — Social aspects. 2. Social intelligence.
3. Sociolinguistics. I. Goody, Esther N.
BF431.S617 1995
1 5 3 . 9 - d c 2 0 94-8853 CIP

ISBN 978-0-521-45329-5 hardback

ISBN 978-0-521-45949-5 paperback

Transferred to digital printing 2008

 Contents

List of contributors page ix

Conventions used in transcripts x
Preface xi

Introduction: some implications of a social origin of 1

intelligence
ESTHER GOODY

Part I Primary processes

1. The ape legacy: the evolution of Machiavellian intelligence 37

and anticipatory interactive planning
RICHARD W. BYRNE
2. How to invent a shared lexicon: the emergence of shared 53
form-meaning mappings in interaction
EDWIN HUTCHINS AND BRIAN HAZLEHURST
3. Hunter-gatherers' kinship organization: implicit roles and 68
rules
NURIT BIRD-DAVID

Part II The interactive negotiation of meaning in

conversation

4. On projection 87
JURGEN STREECK
5. Interaction sequences and anticipatory interactive planning 111
PAUL DREW
6. Where does foresight end and hindsight begin? 139
DAVID GOOD
viii Contents

Part III Genres as tools that shape interaction

7. Politeness strategies and the attribution of intentions: the 153

case of Tzeltal irony
PENELOPE BROWN
8. Interaction planning and intersubjective adjustment of 175
perspectives by communicative genres
THOMAS LUCKMANN

Part IV Expressions of a social bias in intelligence

9. Divination as dialogue: negotiation of meaning with random 189

responses
DAVID ZEITLYN
10. Social intelligence and prayer as dialogue 206
ESTHER GOODY
11. Interactional biases in human thinking 221
STEPHEN C. LEVINSON
12. Stories in the social and mental life of people 261
MICHAEL CARRITHERS

Consolidated bibliography 277

Index 298
 Contributors

Nurit Bird-David Department of Anthropology, University of Tel

Aviv, Israel
Penelope Brown Max Planck Institute for Psycholinguistics, Nijme-
gen, The Netherlands
Richard W. Byrne Scottish Primate Research Group, Department of
Psychology, University of St Andrews, Scotland
Michael Carrithers Department of Anthropology, Durham University
Paul Drew Department of Sociology, University of York
David Good Faculty of Social and Political Sciences, University of
Cambridge
Esther Goody Department of Social Anthropology, University of
Cambridge
Brian Hazlehurst Departments of Anthropology and Cognitive
Science, University of California, San Diego
Edwin Hutchins Department of Cognitive Science, University of
California, San Diego
Stephen C. Levinson Max Planck Institute for Psycholinguistics,
Nijmegen, The Netherlands
Thomas Luckmann Fachgruppe Soziologie, Universitat Konstanz,
Germany
Jiirgen Streeck Department of Speech Communication, The Univer-
sity of Texas, Austin, Texas
David Zeitlyn Wolfson College, Oxford
 Conventions used in transcripts

CAPS spoken loudly

underlined spoken with emphasis
extended sound
(h) soft inbreath
.hhh audible inbreath
.kh inbreath with other audible sounds
[ two speakers speaking simultaneously: position of bracket
indicates point of overlap onset
= short transition time (no pause between lines of speech)
- word broken off
> speeding up
unintelligible
( ) bounds uncertain transcription
(( )) bounds transcriber's comments
- • indicates target turns referred to in text
(0.1) pause (in seconds)
(0 micropause
 Preface

It is both a hazard and a delight of anthropologicalfieldworkthat the more

completely one becomes immersed in a society and culture totally
different from one's own, the more similar people seem to kin and friends
at home. Despite the manifest, subtle and profound differences there is a
level on which people seem to feel and act in basically similar ways.1 The
dynamics of this dialectic between socio-cultural uniqueness and com-
mon humanity lie in part at the intersection between cultural forms and
inter-personal interaction. This is an area I first explored in papers on the
links between greeting, giving and constraining (1972) and on question-
ing (1978a). However these two problems raised more general issues
concerning the inferring of intentions in interaction, and thus the
significance of social roles for making interaction more predictable
(1978b). Ethologists are now suggesting that primate intelligence was
directly linked to the challenges of social interdependence. This insight
places the problems of greeting, questioning and inferring intentions in
an even wider context. What can we learn about the nature of human
society by taking seriously the possibility that human intelligence is in
this fundamental sense social intelligence?
It is difficult to know where to begin with such a general problem,
particularly if there is a commitment to a firm empirical base. The
Working Papers were essays directed at particular aspects: the impli-
cations for primate social intelligence of an emerging spoken language;
the new potentiality of language for meeting the challenges of social
interdependence; language and the emergence of institutionalized gender
roles; language and the emergence of rules. During my attempts to pull
together these several themes it became clear that these very preliminary
ideas needed the challenge of scholarly debate, and exploration in a range
of empirical contexts. The Wissenschaftskolleg zu Berlin made this
possible by sponsoring a workshop on Some implications of a social orgin of
human intelligence in the spring of 1990 to which the contributors to this
volume were invited. Each had already begun to explore some aspect in
their own work, though not necessarily under the rubric of social
xii Preface

intelligence. A set of the draft Working Papers in Social Intelligence was

circulated to participants with the invitation to respond with a paper
related to their own research. Since the present volume grew out of papers
written for this workshop it seems appropriate to begin the Introduction
with themes from the Working Papers. Discussion during the Workshop
and during revision of the individual papers has led to several issues being
more fully worked out, for which grateful thanks to stimulating
colleagues.
In several ways this volume owes its existence to the Wissenschaftskol-
leg zu Berlin. First, because it gave the editor a year of peace, marvellous
facilities and stimulating colleagues which made it possible to pursue
background reading and to draft further Working Papers for the social
intelligence project. Second, because their funding of the workshop on
social intelligence made it possible to bring together scholars from several
fields who were working already on issues related to social intelligence.
And finally because, in the setting of Berlin in May, our discussions of the
exploratory papers led to unanimous agreement that we should publish
revised papers and so extend the debate.
Particular thanks are due to Wolf Lepenies and Joachim Nettelbeck
who were willing to believe that it would be worthwhile to find funds and
facilities to bring participants from far and wide for the Social Intelli-
gence Workshop to talk about such an amorphous topic. Stephen
Levinson and Penny Brown, who were in Berlin launching the Max
Planck Projectgruppe fur Kognitive Anthropologie, gave generously of
time, ideas and enthusiasm during the planning of the workshop. Dr
Stefan Strohschneider of the Projectgruppe fur Kognitive Anthropologie
acted as scribe.
Particular thanks are also due to a number of individuals who let me talk
to them about social intelligence and said it was all worthwhile: First, to
Jack Goody, whose patience is matched only by his willingness to lend a
hand to innumerable chores; and to John Barnes, Michael Cole, Elizabeth
Duvan, Maurice Godelier, Ward Goodenough, John and Jenny Gum-
perz, Robert Hinde, Stephen Hugh-Jones, Phyllis Lee, Skip Rappaport,
and S.J. Tambiah.
The logistics of putting together a manuscript from individual papers
can be daunting; they were made bearable by Sarah Green's skill with a
word processer and preparation of the index, and by Sue Kemsley's
diligence with the bibliography. Sandy Anthony's sub-editorial sharp
sight has made the final stages surprisingly untraumatic.
Several participants admitted later that they had wondered whether it
would be possible to find common ground across such disparate disci-
plines. In fact we were repeatedly surprised at the ease with which topics
 Preface xiii

and issues crossed subject boundaries. The problematic of the impli-

cations of intelligence as social provided a common perspective - though
by no means unanimity!

Note
1 There are certain key ethnographies which explore this question in relation to
emotions, a direction not pursued here but which is part of the wider puzzle:
Briggs's Never in anger (1970); Rosaldo's Knowledge and passion (1980);
Reisman's two books on the Fulani, Freedom in Fulani social life (1977), and
First find your child a good mother (1992).
 ESTHER GOODY

Introduction: Some implications of a social

origin of intelligence

Recent work in ethology persuasively argues that the striking advance in

primate intelligence over that of lower mammals is a product of social
interdependence. This finding raises two kinds of questions for students
of human society. In the first place, if the constraints of social interaction
generated primate intelligence, what was the ratchet that led to the
emergence of incomparably greater hominid intelligence, that is to Homo
sapiens sapiens} For reasons outlined below, this stimulus seems likely to
have been closely related to the gradual emergence of spoken language.
Although discussion of the role of language in human evolution can, for
the present at least, be only speculative, the problem has a fascination
which justifies such exploratory thinking. We need to ask how language
might have altered primate social life in ways that demanded, and
rewarded, more complex intelligence.
A second question looks forward, not back; it concerns the contempor-
ary nature of human intelligence. If human intelligence evolved in
response to the challenges of social living, what are the implications for
understanding thought, interaction and social forms? This introduction
outlines some of the dynamics likely to be related to both questions in a
way that is intended to raise problems for further analysis and research.

Themes from Working Papers

1. Primate intelligence as a response to social interdependence
In a seminal paper, 'The social function of intellect', Nicholas Humph-
rey (1976) has argued for a social origin of primate intelligence. He begins
by noting that the higher primates, and particularly chimpanzees, are on
many measures highly intelligent. Yet there is nothing in their life as
foraging animals which demands such a level of intelligence. In this, he
says, they are comparable to what we know of early man from studies of
hunter-gatherer society. For both species, foraging follows customary
patterns and the use of tools is very limited, both being based on imitation
of others, or trial and error. Innovation requiring 'higher-order' or
2 Esther Goody

'creative' intelligence - that is, the ability to make inferences from novel
conjunctions of events, is rare. The stone tools of Homo erectus appear to
have hardly changed at all over one million years. This is a problem for
ethologists because it is inconceivable that a creature would develop skills
that are seldom or never used. Indeed, their assumption is the opposite:
that skills only develop in response to pressures from the environment,
such that the emergent skill makes a significant contribution to improved
chances of reproductive survival. So what use is higher-order intelligence
to anthropoid apes and stone-age man, if it doesn't provide an advantage
in dealing with the natural environment?
Humphrey suggests that the most difficult problems facing chimpan-
zees are other chimpanzees; that it was in dealing with the social
environment that creative intelligence evolved. This suggestion makes
sense once one recognizes the peculiar situation of interdependence that
characterizes both ape and human society. Both live in groups. And living
in groups requires being able to pursue individual goals effectively
without alienating one's fellows, breaking up the group, or creating a
situation of conflict within the group so that it becomes vulnerable to
outside attack.
Thus social primates are required by the very nature of the system they
create and maintain to be calculating beings; they must be able to calculate
the consequences of their own behaviour, to calculate the likely behaviour
of others, to calculate the balance of advantage and loss - and all this in a
context where the evidence on which their calculations are based is
ephemeral, ambiguous and liable to change, not least as a consequence of
their own actions (Humphrey 1988:19).
Acting on such models of the behaviour of others involves social
transactions; there is a constant trading off between partners. If one
animal or person wishes to change the behaviour of another he must take
into account the other's goals and tactics. So in addition to the cognitive
skills required to perceive the current state of play (low-level intelli-
gence), the social gamesman, like the chess player, must be capable of a
special kind of forward planning. As each move may call forth several
alternative responses, and ego's own response choice must vary accord-
ingly, this situation generates a decision-tree model.
In short, effective social living requires anticipation of the actions of
others, calculation of short- and long-term costs and gains, and close
attention to signals about the consequences of one's own behaviour. The
higher primates, and man, have the ability to model this interdependence
of one's own and others' behaviour at the cognitive level.1 In order to
facilitate thinking about this kind of thought it is useful to give it some
kind of representation; for convenience it might be termed anticipatory
interactive planning, or AIP.
 Introduction 3

Anticipatory interactive planning was a response to social living among

primates. In turn, AIP set challenges which generated progressively
increasing intelligence in the hominid line (Jolly 1966a; Humphrey 1976;
papers in Byrne and Whiten 1988).

2. Hominids, or at least Homo sapiens,, were able to utilize spoken

language in the representation of their own and others3 contingent responses.
Language clearly facilitates AIP in several ways, and must have made it very
much more powerful. The emergence of language, then, may have been critical
in the generation of hominid intelligence
Social intelligence itself cannot account for the emergence of Homo
sapiens, since it is shared with many higher primates. If spoken language is
found only in humans, how is this related to the emergence of our species?
How might social intelligence have led to, and been enhanced by, the
development of language?
Progressive increase in cranial capacity through the Homo species to
Homo sapiens suggests a continuing increase in intelligence during
hominid evolution. This accelerates markedly with H. erectus and early
sapiens.2 Until recently it had been thought that language occurred
suddenly through a mutation ('a unique genetic event') in which grammar
and syntax were 'wired in' (Chomsky 1968, 1980; Premack 1986).
However, several authorities have recently argued that language must
have appeared in two stages: Lyons suggests that a gestural language
preceded the emergence of human spoken language (1988). Recently
Donald has proposed a protolanguage based on gesture and mime related
to a mimetic form of cognition (1991). He sees both this protolanguage
and cognition as distinct from spoken language and its related cognitive
structures. Bickerton (1990) proposes a spoken protolanguage consisting
essentially of a simple lexicon, perhaps later augmented by grammatical
elements which were, however, quite independent of the grammar and
syntax of 'true language' that appeared only with a genetic mutation.
Thus these authorities retain the Chomskian premise that 'true language'
was the result of a unique genetic event rather than developing gradually
on the classic Darwinian model. This position makes it difficult to view
early protolanguage as preadaptive for more complex forms since it is the
disjunction between them that is stressed. In Bickerton's case this
difficulty is ironic, since his insightful discussion of the emergence of
grammatical elements in response to challenges for more clarity and
specificity of reference is a model for an adaptionist account of the
emergence of grammar and syntax.
Other scholars are beginning to pursue an adaptionist model that seeks
to relate the initial emergence of spoken language to such primate features
as the evolution of hemispheric specialization for both manual function
4 Esther Goody

and language, bipedal posture and tool use.3 In such a view early
protolanguage could have been extremely crude permitting only simple
reference, with phonemes, grammar and syntax very gradually emerging
as particular responses to specific problems of using spoken language (e.g.
tense, negation, thematic roles). In their several papers and joint book
(forthcoming) Lindblom, MacNeilage and Studdert-Kennedy persuas-
ively argue the general case for a Darwinian view of gradual emergence of
spoken language, laying out in detail how phonemesfitsuch a pattern, and
addressing the critical issue of the structural and behavioural acquisition
of the ability to produce speech. Based on his studies of primate and
human structures Lieberman has long argued that spoken language must
have evolved through progressive modification of the vocal tract and
associated cognitive specialization (1968, 1991). The current picture
suggests that early Homo erectus already had a vocal tract differing
significantly from the apes and Australopithecus species. As there would
be no reason for such modification without the advantage of spoken
language,4 the clear implication is that hominids have been using some
sort of spoken language for over one million years.5
Robin Dunbar (1993) proposes another, very persuasive link between
hominid intelligence and spoken language. He argues that primate group
size and intelligence can be shown to increase in parallel, supporting the
Humphrey, Jolly, Byrne and Whiten view that social living was the
critical challenge; as primate groups got larger, demands of cognitive
representation multiplied. But Dunbar points out that many sub-human
primates rely on mutual grooming for the servicing of social relationships,
and that for large groups this mechanism ceases to be effective, since it is
time-consuming and a one-to-one interaction. Even a simple language, on
the other hand, would have permitted the 'servicing' of many social
relationships - simultaneously, and at a distance. Indeed, he suggests that
initially instrumentally focused information may have been of secondary
importance.
In accounting for the emergence of early language, Dunbar retains the
premise that ecological constraints were primary in hominid evolution.
He suggests that conditions for securing subsistence required early
hominids to live in groups too large to be socially maintained through
grooming. Thus, being forced to live in large groups, early hominids
evolved 'gossip' to service social relationships. The argument for a social
facilitation function of early language is very strong. Anthropologists
since Malinowski (1927) have recognized its continuing significance.6
However there is no real argument made for ecological constraints as the
initiating factor, and it is in any case superfluous. We know that some
primates use complex vocal signals (e.g. Cheney and Seyforth 1990). It
seems quite probable that human protolanguage developed from such a
base. An alternative hypothesis would be that some early hominid
 Introduction 5

group(s) began to elaborate vocal signals in a way that facilitated the

servicing of social relations in groups larger than those possible using
grooming alone. The challenge of using protolanguage in larger groups
may well have driven the first simple regularities of usage, precursors of
grammar and syntax.
If, as it seems probable, the long-term intimacy of mother-infant-
child-sibling communication is central to the transfer of language skills
between generations, then this posed the problem for early hominids of
how children of different mothers (and different matrilines) understood
each other. Mothers who were themselves matri-siblings might transmit
the same language usages to their children, but there must be a point at
which the group becomes too large for this to balance changes of usage.
Perhaps initially protolanguages were shared only within small groups.
Larger groups of the sort Dunbar identifies as truly depending on spoken
language for maintaining social relationships might then have been
possible only after rudimentary grammars emerged which could cross
boundaries of domestic bands.
The invisibility of social behaviour, especially spoken language, in the
archaeological record has inevitably led to a focus on 'bones and stones',
skeletal fossils and tools. Clearly both indirectly reflect social forms, but
not without conjecture. If sociality does prove to be central to hominid
cognition, then an evolutionary account must also give it a key role. A
parallel development of hominid intelligence and of language invites us to
ask in what ways early language might have facilitated social intelligence.
Among the most obvious effects would be:
(a) Reference by name to things and actions permits joint attention, and
thus coordination of complex activities.
(i) This reference would then have to be mentally represented.
(ii) The processes of coordinating joint activity would also have to be
mentally represented.
(b) The emerging structures of grammar and syntax permit much more
complex and more rapid conversation.7
(i) These grammar and syntax structures have to be cognitively
represented.
(ii) Grammar and syntax processes have presumably become stan-
dardized, then routinized and finally automatized,8 as has the
motor control of speaking. We are not aware of how we articulate
and produce the sounds in our speech (How do you say 'feather'?)
Nor are we conscious of whether we put subject or verb first in a
sentence. Automatization appears to free many complex pro-
cesses from awareness, as car drivers are often startled to realize
(see Velmans 1991 and comments).
6 Esther Goody

(c) Bickerton argues that a major vector of animal evolution is the

complexity of representational system. He distinguishes between Prim-
ary Representational Systems (PRS), based on processing and represen-
tation of sensory input, and Secondary Representational Systems (SRS)
in which the output of the PRS is mapped onto an externalized language.
Species having only a PRS still respond to their environment in systema-
tic ways which show they have 'tacit concepts' - like a frog's response to a
fly. But a SRS permits the labelling of tacit concepts so that they become
explicit (1990).
What difference does it make that concepts are labelled? There are
important processes nested within the use of reference in spoken lan-
guage. Reference (by name or words) permits classification of things,
actions, feelings, events, etc. by making it possible to be explicit about the
categories to which they belong. Such classification is built into languages
so that we use it automatically (animal > dog > terrier > my terrier Spot).
Cognitive psychology suggests that this must have had important conse-
quences for the organization of perception and memory. If we can
process, organize and recall categories then we can handle hugely more
information than if we must deal with particular instances. This must
have represented a major advance in human cognition.9
The very process of establishing common meanings for lexical items
places them in the domain of shared knowledge, as a part of what
Hutchins and others have called distributed intelligence (see Hutchins
and Hazlehurst, Chapter 2, this volume). A shared lexicon represents the
coordination of meanings which makes the coordination of actions poss-
ible on a quite new level compared with what is possible on the basis of
private inference.10 Another way of putting this social reality of language
is to note that a Secondary Representational System does far more than
multiply the power of the individual's primary representational system; it
gives every individual the power to enter into the linguistic represen-
tations of others, and to use these shared secondary representations to
model cognitively the understandings as well as the intentions of others.
And of course the fact that we use language to influence others means that
we can move from cognitive AIP models incorporating reference to using
these categories in speech to seek to implement AIP strategies. Language
as a SRS bridges individual cognitions through cognitive modelling using
shared meanings. At the same time it makes possible the coordination of
joint action between individuals by speaking about common goals,
instrumental means etc. using these shared meanings.
(d) Classification in turn is a prerequisite for the emergence of social
roles (see below) and social rules (see below) which were necessary tools
for the construction of even the simplest institutions of human social
living. Both roles and rules now become elements in AIP representations,
 Introduction 7

simplifying them in some ways, but also making them more powerful. At
the same time roles and rules constrain the behaviour of others, making it
more predictable.
(e) Language permits the individual to act much more effectively on his
social world in two modes: information and control.
(i) The cognitive modelling of the contingent actions of others,
anticipatory interactive planning, depends on information about how
others will act. While past experience may give clues for inferring
responses, this process is obviously limited. Language permits the
explicit exchange of information. (If I want another to help me get food it
is useful to know whether he is hungry. He may have mentioned this, or I
can ask him. Or I can ask someone else if he is hungry, or . . .)
This raises the issue of distortion of information as one AIP strategy.
Ethologists have seen deception as the key to primate social intelligence,
as suggested in the title of a recent important book Machiavellian
Intelligence (Byrne and Whiten 1988) and further discussed in Byrne's
contribution to the present volume. Barnes's study of lying (1994)
pursues this theme in human social life. The linguist Grice proposes that
effective use of language depends on our being able to assume that others
speak truthfully. Indeed deception is powerful precisely because we are so
dependent on correct information for modelling our own and others'
actions. AIP models of others' intentions can include the intention to
deceive, their perception of our awareness of this deception, and our own
counter deception, and so on. In Chapter 6, Good suggests that the
ambiguity of conversational exchange may be one way of preserving the
freedom to respond appropriately to deception.
(ii) However, AIP is not an end in itself but a means towards reaching
our goals. AIP strategies must be implemented. The other mode of acting
upon the world which language profoundly enhances is that of control.
The use of language to manage relationships with others is extremely
powerful, and dauntingly complex. On the simplest level commands both
organize action and express dominance. Successful commands are prob-
ably the most effective AIP strategy of all, since they secure direct
compliance with one's own goals without the need for calculating
alternative strategies or engaging in negotiation. But of course language
also permits social cooperation and negotiation of joint strategies. Con-
versational analysis reveals, on a less explicit level, the subtle nature of the
cooperative negotiation of meanings in ordinary daily life. And the
identification and analysis of speech acts has led to the recognition that
speech often conveys several kinds of message simultaneously. The
contributions of Streeck, Drew and Good in Part II of this volume
present current thinking on key aspects of interactive negotiation in
conversation.11
8 Esther Goody

It is clear that information and control tend to be merged in most use of

spoken language; language is used to manage social relations at the same
time that it conveys information. However, management of relationships
is probably the single most central use of language. Malinowski noted its
importance long ago when he pointed out that communication apparently
carrying no content whatever (which he termed 'phatic'12) was still
important for maintenance of social life. Greeting forms are the most
familiar example (see E. Goody 1972). Indeed Dunbar suggests that
spoken language may have begun as companionable chatter with minimal
referential content ('gossip') which served to maintain social integration
(1992a). The formal constraints on use of language represented by
institutionalized avoidance and joking relations (E. Goody 1978b), and by
the universal politeness forms (Brown and Levinson 1978, 1987) reflect
points where social relationships can be threatened by casual use of
language. Garfinkel's experiments (1967) with altering expected res-
ponses in interactions between associates show again how fragile are the
routines we construct to give meaning to our normal use of language, and
how threatening is any deviation from what we expect. We depend on
customary use of language to give constant evidence of the social validity
of our relationships.
(f) Finally, language makes possible the objectification of belief (Rappa-
port 1988; Goodenough 1990). With language we construct social worlds
that have ultimate reality. The totemic world of the Australian Abori-
gines, the layered world of Hindu reincarnations, the medieval realms of
heaven and hell - these exist only through the complex representations
made possible by language. It is the sharing of these beliefs, impossible
without language, which makes them real.13 And it is this reality which in
turn sets premises for shared goals which make possible joint social action
at least part of the time.
Language and AIP together generated the complexity of mental
representations which characterize the cognition of Homo sapiens.

3. Language was essential to the awareness of the embedding of intentional-

ity which distinguishes Homo sapiens (Dennett 1983,1987)
The awareness of the self as both actor and object is probably
dependent on language. Mead has given a brilliant account of how
hearing ourselves speak brings about this recognition. Among the higher
primates only chimpanzees have been shown to recognize themselves in a
mirror, but mirrors are not a part of the natural chimpanzee world. Even if
they occasionally catch sight of themselves in a pool of water, chimpan-
zees cannot represent this 'self to others in speech; it cannot be
objectified. If primate creative intelligence is based on the capacity to
represent mentally the reciprocal contingency of one's own and others'
 Introduction 9

actions, perhaps language has made it possible for Homo sapiens to

become at least partially aware of this process of representation. In AIP
terms, awareness of the 'self could be seen as the recognition by the T of
the planning of interactive strategies in which the self is also one of the
actors ('me'). Linguistic representation of intentions may also have been
critical in the development of the human skill in pretending. Leslie (1987)
has suggested that pretending requires the capacity simultaneously to
maintain two levels of belief; one knows the way the world 'really is', and
at the same time one posits a different state of the world and manipulates
this, for play or fantasy. Children with the mental deficit of autism appear
to be unable to sustain such dual levels of representation. Again there is
some evidence that chimpanzees engage in pretence in play. But they do
not represent this pretence to themselves, or to other chimpanzees, as a
basis for alternative hypothetical worlds. Only with language can pre-
tence be socially objectified.
With language, Homo sapiens is able to objectify the self as well as others
in representation of AIP strategies.

4. The capacity to learn from others is enhanced by language in several

ways. Homo sapiens depends on learning rather than instinct for the
transmission of adaptations between generations; effective learning on the
scale necessary for the transmission of culturally developed adaptations
depends on language
The Vygotskian view sees learning as based on activity carried out
jointly by novice and expert.14 The novice begins by watching, then joins
in with the simplest tasks, then gradually takes responsibility for the
whole activity. In carrying out the activity together with the 'expert', the
novice establishes routines which form the basis for mental schemata of
the activity. These schemata model both the actions of the novice and of
the expert. At first the novice need not 'understand' what s/he is doing,
since the expert organizes and guides the process. As the novice comes to
understand, and to master component skills, s/he is able to take responsi-
bility for parts of the activity, and eventually for the whole. Vygotsky
argued that cognitive processes take place first on the social plane, and
that these joint processes are internalized to become the individual plane.
During the learning process there is a 'space' where the novice is able to
cooperate, but not to take responsibility - this has been termed the zone of
proximal development (Vygotsky 1978:84-91; Wood, Bruner and Ross
1976; Cole 1985:154-9; RogorT 1990). This view implies that the novice
can jointly perform an activity before s/he is able to do it alone. Some
chimpanzee learning is clearly of this sort, as with the use of straws to
extract termites. What does language add to this process of social
learning?
10 Esther Goody

(a) When learning is accompanied by speech, the schemata are objecti-

fied. The task may be discussed, corrected, elaborated. The activity being
learned can become an explicit object for learning. The Utku Inuit have
partly verbal routines between parents and infants which act out fearful-
ness, and the appropriate response of conciliatory dependence. These
routines 'teach' the child to control anger at all times, and control of anger
is a prerequisite for membership in an Utku community (Briggs 1970).
(b) With language, skills can be explicitly identified with roles, and
indeed tend to form a central component of the social definition of roles.
The ubiquity of a sexual division of labour as a core premise of every
society is a paradigmatic instance of this process. Many authorities have
noted that tasks which are culturally specified as gender-specific often
lack any features objectively restricting them to that sex (e.g. LaFontaine
1978). It seems likely that once terms for gender roles based on physiologi-
cal differences are in use, activities typical for that gender come to be
referred to by these terms and to be regarded as 'naturally' gender-
specific in the same way as physiological differences themselves. In this
way language enters into the social construction of reality (E. Goody
n.d.).
(c) Many simple societies have not institutionalized formal teaching/
learning roles beyond the expert/novice distinction integral to coopera-
tive activity between adults and children, and the modelling and control
implicit in the relationship of parents to children (E. Goody 1982, 1989,
1993; RogorT 1990). However with language, teaching roles become
possible because learning can be objectified. Many simple societies link
teaching with role transitions, usually heavily embedded in ritual, as in
'initiations' at puberty. What is taught, and learned, tends to be a blend of
deference to the authority of seniors and the rights and obligations of
adult roles (cf. Richards 1956).
(d) With language, culturally developed skills become cultural capital,
to be transmitted, shared, or restricted. Skills become objectified.
With language, the schemata on which AIP depends can be more
effectively taught through joint activity; culturally developed skills can be
transmitted between generations, both through joint activity, and
through explicit teaching.

5. The social nature of learning replicates, and objectifies, the social

character of AIP in which one's own actions are represented as contingently
interdependent on others3 responses. With language, AIP representations
come to be expressed in internal dialogue.
Vygotsky (1962) argues that thought is internalized speech. Initially
 Introduction 11

the infant participates jointly, apparently passively, with adults and older
children in speech acts; mothers of newborn babies have been found to
hold conversations with the infant in which they also supply responses.15
Mothers of older babies engage in play routines which scaffold the
learning of vocabulary and apparently of syntactic frames (Bruner and
Sherwood 1976). The child's increasing competence in speaking is
expressed in conversation with those around her, but also in conver-
sations with herself (egocentric speech) to which every new parent listens
with awe. Older children continue to use egocentric speech, for instance
when confronting a difficult problem where the steps of a solution are
uttered verbally as self-instruction. Egocentric speech is 'a form found in
the transition from external to inner speech' (Vygotsky, quoted in
Wertsch 1985a: 108-28). Gradually, at about the age of seven, egocentric
speech disappears as the child comes to internalize it as inner speech,
which nevertheless retains its regulative force (see also Luria 1979:chs. 5
and 6; 1981).
For Vygotsky the significance of inner speech is that it 'enables humans
to plan and regulate their activity and derives from previous participation
in verbal social interaction' (quoted in Wertsch 1985a: 111). This model of
human planning and regulation maps neatly onto our model for primate
AIP, with the addition, of course, of spoken language. A basic axiom of
Vygotsky's theory is 'that development... occurs through the decontex-
tualization of mediational means'. For ape AIP the mediational means
must be some form of cognitive representation, but presumably highly
contextualized, anchored in the immediate here and now. Language is the
prime mediational means for action between humans, and is also central
to cognitive representations. However as a Secondary Representational
System, language is decontextualized; shared lexical meanings must hold
across contexts. Unfortunately we cannot know what form is taken by
apes' inner thoughts. But it cannot include a shared verbal lexicon, which
means that ape plans cannot include labelled concepts, or the possibility
of basing AIP strategies on explicitly shared meanings. Nor can ape inner
thoughts be self-regulatory in the sense of 'talking to oneself - internal-
ized egocentric speech - which explicitly represents intentions to oneself.
The use of inner speech in planning and regulating human actions gives
AIP a concreteness not possible without language. However the fact that
inner speech at times gives us awareness and control over AIP is probably
misleading. For introspection forces one to admit that intentional model-
ling of contingencies and planning of strategies is intermittent at best. Or
better, we move between levels of awareness. At times attention is sharply
focused on a problem arising from social interdependence and we
intentionally work out alternative solutions; at other times we muse on
12 Esther Goody

such problems, but attention is directed elsewhere; yet again we are

engrossed, perhaps in music or a technical problem, and totally ignore any
social dimension.16 Just as apes were using AIP without language, so
humans seem to continue to monitor social interdependence on non-
verbal levels at the same time as we make use of linguistic represen-
tation.17 Although we are seldom aware of doing this, conversation itself
appears to involve constant monitoring and moment-by-moment adjust-
ments of each speaker to what the other says and the way it is said. This
reciprocal monitoring is so integral to conversation that often the mean-
ing which emerges from this mutual contingency of response was neither
anticipated nor intended by either speaker (Chapters 4, 5 and 6 in this
volume).
Both through inner speech, which is a sort of dialogue with ourselves
(between me and I), and through our close attention to conversational
partners, spoken language seems to have constructed a dialogue template
for social cognition. In inner speech and in conversation, dialogue and the
dyad are built into human cognition. This dyadic premise and the
dialogue template can be seen to underlie the human reliance on prayer as
'supplicatory speech addressed to some external force or being'. It is also
expressed in the social construction of unseen mystical powers and the
personification of good and evil which form the basis of religions.
When language transforms thought into internal dialogue, internal
dialogue provides a mode for responding to cognitive problems.

6. The A IP structure of thought which incorporates mental representations

of others' responses to one's own actions is itself a model for reciprocity in
social relations
If the way we think already includes the linking of our actions to the
responses of others, then the expression of this in giving and receiving
'things' is an externalization of thought processes. This highlights the
complexity of AIP, which must both model contingent responses and
model strategies for securing actions from others which are favourable to
ego's goals. AIP moves constantly back and forth between modelling and
strategic action. Thus 'giving' in its most basic form is the bestowal on
another of something expected to favourably influence behaviour. This
would fit with the fact that as well as giving 'things', we 'give' actions:
respect, affection, offence . . . (E. Goody 1972).
It is also consistent with the 'mystical' quality of the gift to which
Mauss drew attention; the giver of a gift is somehow still linked to it, the
act is incomplete until a return gift is made.18
The power and ubiquity of the gift in human societies can be seen as
related to its capacity both to represent the contingency of social relations,
and at the same time to permit manipulation of this contingency.
 Introduction 13

7. One implication of the view that thought models reciprocal socially

contingent actions is that logical3 thought is 'artificial'
Formal logical thought requires finding a way to represent events that
does not involve the A IP format of action and response, interpersonal
cause and effect. The role of writing seems to be critical in the emergence
of such logic, as Jack Goody has argued (1977, 1987). It permits
externalization of thought apart from the individual's goal strategies. One
can see Plato's use of the dialogue as a vehicle for demonstrating the truth
of his propositions as transitional between conversation and writing.
Writing makes possible, though obviously it does not necessitate, the
representation of information separately from the conditions of goal-
realization (E. Goody 1978a). Levinson's chapter in this volume (Chapter
11) explores evidence for and implications of a separation between
interactive (social) intelligence and logical reasoning.
Formal logical thought employs conceptual tools developed through
literacy, and is not inherent in human cognitive processes.

8. The constraints of social interdependence on which language built to

produce hominid intelligence may also be central for the emergence of human
social institutions. When language enhanced predictability, information and
control, the result may have been the emergence of social roles and rules
What are the implications of the view of intelligence as fundamentally
social for the construction of social institutions on which human society
depends? One way to look at this is to ask what factors AIP depends on for
efficacy. On the most general level these seem to include predictability of
others' actions, information concerning the meaning of others' actions,
and control over our own and others' actions.19 If predictability, infor-
mation and control were critical constraints in the emergence of spoken
language, they probably also shaped social institutions.
The forms of social institutions vary greatly across human societies. So
great is this variation that some anthropologists have argued that it is
futile to seek regularities in their form and function; that rich description
can be the only legitimate goal. While this is certainly true at the level of an
attempt to compare total institutions, successful comparison has always
depended on using analytical concepts that represent common features.
Thus very useful studies have been made of joking relations, cross-cousin
marriage, descent-group dynamics, gift giving, legal institutions, the
feud, ritual, etc. It is a question of matching the conceptual tools to the
level of process being examined. In working at the level of the emergence
of social institutions, there are good reasons to consider as basic the
constructs of social role and rule. Our question then becomes, how do
social roles and rules mediate between individual AIP cognition and
social institutions?
14 Esther Goody

Social roles
One of the possibilities created by the use of reference in speech is the
naming of categories. Categorization has basic effects on both perception
and memory (e.g. Miller and Johnson-Laird 1976). Some insight into the
implications for social institutions is provided by recent attempts to
model the significance of the emergence of the role of 'father' in human
societies (Wilson 1980, Fortes 1983). Both authors argue that the social
recognition of the father marked the beginning of human society. Yet
neither discusses the role of language in this transition. Their arguments
in fact presume spoken language. Wilson writes of the power of the
promise as a commitment to future action; but promises presume
complex spoken language which permits representation of obligation,
and of the future. Fortes links the emergence of the role of the father with
the internalization of rules, without which people cannot live together in
society. But it is hard to see by what process rules could be formulated or
invoked, let alone internalized without spoken language. Further, the
kind of rules Fortes is discussing refers to categories of actors and actions;
not to one particular father only, but to the role of father; not to the sin of
sleeping with a particular female, but to the sin of sleeping with a 'mother'
or a 'sister'. Indeed it seems probable that it is the very extension of norms
generated in domestic relationships to other kin which is paradigmatic of
both social roles and social rules. It is surely worth reflecting on the fact
that one of the few true universals in human societies is some form of
kinship terminology linked to kin roles.
Another of these rare universals is a division of labour based on sex.
The emergence of a complex sexual division of labour must also have
depended on language and the capacity to categorize behaviour. While
the actual tasks assigned to each sex vary widely across societies, the
grouping of tasks into those appropriate to one sex or the other appears to
be universal (E. Goody n.d.).
The naming of regularities of behaviour associated with positions in a
social structure as roles entailing recognized rights and obligations is thus
a feature of all human societies. In this way roles provided a powerful new
kind of information for AIP. Roles of this kind create frames for
predictable behaviour, associated with sanctions. Further, they define
behaviour in dyadic terms - a maternal role implies a child role, and it is
the interdependence of the two roles which is delineated by the rights-
duties-sanctions complex. Thus a definition of appropriate role behav-
iour includes both the proper behaviour of ego, and the proper reciprocal
responses of the role partner. In this way it makes AIP more powerful,
since acting in a clearly defined role makes the behaviour of both members
of the dyad more predictable. Indeed, their behaviour is predictable not
only to the participants, but to others (in the same society).
 Introduction 15

Roles also facilitate AIP through increasing predictability on another

level by implying sanctions if one or other member of the dyad does not
behave in the correct, predictable way. Thus roles include a control
element that facilitates the AIP cognition of all members of the society.
The classifying and labelling of role behaviour has an added signifi-
cance at the level of the social system because sanctions are part of the
mechanism for learning and maintaining 'social structures' and 'social
capital'. Roles lead out from individual behaviour to the reproduction of
the social structure.20

Rules
'Rule' is one of a fuzzy set of terms for regularities of behaviour. Other
terms include 'routines', 'regularities', 'norms' and 'laws'. These terms
have in common the fact that they refer to an apparent regularity in
behaviour in a specific domain. But they differ profoundly in the extent to
which the regularity is (a) intentional (routines need not be intentional,
and may not be subject to conscious awareness); (b) socially sanctioned (as
in norms); (c) explicitly formulated (as in rules); or (d) embedded in legal
institutions (as with laws). These regularities of behaviour vary on two
dimensions at least: they vary in their degree of explictness, and the
formality with which they are formulated; and they vary in how devi-
ations are sanctioned. At one extreme, routines are sanctioned only by
those who participate regularly in them, and the sanction may be
unconsciously produced and subliminally perceived (as when another's
frown expresses unacknowledged discomfort at an unfulfilled expec-
tation, and warns the actor of his departure from the normal routine,
perhaps again without his being aware of it). Other forms of sanction
range from complaints and protests, to punishments, divine retribution,
or rebellions. Sanctions can, of course, be positive as well as negative, and
either individual or social.
As Garfinkel has demonstrated, once a regularity of interactive behav-
iour becomes established, failure to follow the expected pattern makes
participants feel uncomfortable; a failure is also understood to be a
statement about the relationship. Such behaviour has meaning attributed
to it, whether or not that meaning is 'intended', whether or not it is
'correct'. As AlP-users, we depend on predictability in order to construct
our mental representations of plans based on anticipated responses.
Whatever the origin of predictable patterns of behaviour (whether roles,
rules, or personal relationships), when our expectations are violated we
feel uncomfortable and, unless under some form of constraint, we do
something to cause the wayward behaviour to return to 'normal'. Norms,
rules and laws are terms used to refer to the kind of regularity that is
publicly defensible. In this way they are indicators of expected regularity,
16 Esther Goody

and their defensibility asserts their social legitimacy. Norms, rules and
laws have to do with the synchronizing of actions of individual actors in
socially defined forms. They are about the transformation of individual
attempts to reach goals through cooperation with or the coercion of
another individual into socially constrained forms of action. The mechan-
isms of this synchronization, this transformation, are not always clear,
and indeed should be far more central to our study of society than they
are.21
The general mechanisms by which routines shape individual inter-
action as reciprocally interdependent must draw on and be represented in
AIP modelling. The processes by which such routinized expectations
come to be more widely shared as norms, rules or laws, are social rather
than cognitive, but they must also be cognitively modelled. Such social
processes clearly include the use of language to categorize and label, the
linking of labels to roles, the sharing through language of expectations as
norms, the explicit framing of rules and laws, the creation of cultural
representations (such as mystical beings, totemic myths, and kinship
systems), and the framing of cultural premises of causation, obligation
and morality. The single feature that all these processes would seem to
share in addition to the dependence on language is that their products are
the result of social interdependence. These products - norms, beliefs,
institutions and cultural premises - are emergent from, constructed
through, the mutual negotiation of predictable patterns of behaviour. In
Heritage's terms (1984), people are made accountable for their actions in
social settings. This accountability produces, literally, accounts - to
ourselves and to others - of our actions. It is these accounts which
constitute the shared stuff of beliefs and institutions.

The cognitive modelling of others' contingent responses (AIP) seems

likely to have driven the dramatic growth in primate intelligence.
Ultimately AIP concerns the modelling of intentions, both my own
intentions and what I deem to be the intentions of others.22 Spoken
language constitutes a uniquely powerful tool, both for modelling inten-
tions and for seeking to influence them. Language, increasingly as
grammar and syntax began to evolve, also proved to be a tool for
constructing reality. More complex language made it possible to tell
stories about the past, and about distant people and places; it made it
possible to ask, and try to answer, questions about how and why things
happen; language made it possible to map understanding of intentionality
onto imagined worlds.
 Introduction 17

Themes from individual chapters

The original invitation urged participants to bring even rough drafts to
the Workshop on Implications of a Social Origin of Intelligence, since
this was intended as an opportunity to explore ideas. At the conclusion of
the Workshop all agreed that we wished to pursue these questions by
publishing papers; the present volume is the result. The contributions fall
naturally into four sets focusing on different aspects, and different levels
of social intelligence. The chapters in Part I concern what might be
termed primary structures or processes. Part II concentrates on the
implications for an hypothesis of social intelligence of the interactive
negotiation of meaning in conversation. The two chapters in Part III
describe how communicative genres are both shaped by wider institu-
tional and historical processes, and yet act as tools with which individuals
can manage interaction. Finally, Part IV includes papers which discuss
ways in which the social bias in human intelligence is expressed in
reasoning, in interaction, and in social institutions. Some chapters of
course speak to more than one of these domains, and all address other
issues as well. Here I pick up themes which are central for the general
problem of the implications of an intelligence which is social in origin and
in dynamics.

Part I Primary processes

The stimulus to reformulate our understanding of human intelligence as
profoundly social came from ethologists' reassessment of the develop-
ment of primate intelligence. Richard Byrne's contribution to this
volume (Chapter 1) provides an account of the current state of this
exciting work, and argues for the central importance of the capacity for
deception in the rapid increase in primate intelligence. Deception in
social relation has the special quality of challenging the other to outwit the
deceiver. Transactions built on deception thus are not closed, modelling
and planning must always allow for revision to take account of further
deception. The ape who was most skilful at deception literally outwitted
the others in competition for scarce resources, resulting in selection
pressure towards increasing deceptive skill. Byrne sees hominid intelli-
gence as the result of yet more deception, i.e. of the continued operation of
this same mechanism.
It is in the nature of AIP to model alternatives contingent on varied
responses to ego's actions; this is evident in Humphrey's use of the chess
game as a model for social intelligence. We would expect strategies for
action, and for conversation, to reflect this need for flexibility. Perhaps
during the evolution towards the hominidae there was a progression from
18 Esther Goody

modelling others' current actions, to modelling likely future actions, to

modelling alternative possible future actions. The fact that others were
also modelling with increasing complexity would drive the sort of
escalation posited for Machiavellian intelligence. Here the element of
deceit would be an aspect of goal-directed behaviour rather than the goal
itself.
Byrne argues that in contrast to monkeys, there is evidence that apes
model intentions of others as well as their immediate behaviour. Hence
AIP has to include representation of possible intentional states as well as
of instrumental actions; the possibility of deception would drive the
modelling of intentional states of others. The need to model intentional
states in others must have been directly linked to the recognition of
intentional states in oneself, and chimpanzees' ability to recognize their
own reflections suggests that this is so. Vygotsky would see recognition of
our own intentional states as growing out of the practice of modelling
others' intentions.
However, the capacity for anticipatory interactive planning provided a
basis for cooperation as well as deception. Quite complex primate
cooperation has been documented, both at the level of alliance against a
more powerful individual (Harcourt and deWaal 1992) and of the males of
a band against a hunted animal (Boesch and Boesch 1989). Stephen
Levinson (Chapter 11) argues that cooperation is more difficult than
deception since it requires mutual coordination of several individuals. It
is clear that spoken language provided a many-faceted tool for forging and
utilizing cooperation. Language facilitates cooperation at the level of the
communication of shared meanings, and by controlling coordination of
actions across considerable space, in the dark, and for the future.
Critically, language permits joint planning, transforming individual AIP
into a social process, and individual AIP strategies into social - coopera-
tive - action. If there is an answer to the puzzle of the hominid transition,
it may well lie in the power of the tool of language to transform individual
AIP strategies into cooperatively planned and executed social action.
However the establishment of shared meanings is a problem. Bickerton
(1990) considers reference to be the first of two major achievements which
led to 'true' language. Levinson's chapter outlines in detail the difficulties
in knowing what others mean by our words (and of course vice versa).
This is the background to Edwin Hutchins's and Brian Hazlehurst's
model for establishing shared meanings for a lexicon (Chapter 2).
Critically, this model only works if there are two (or more) actors to
exchange words and establish meanings. Each confirms the meaning for
the other by responding appropriately and 'sending it back to him' in later
exchanges. This is indeed an appropriate model for social intelligence
itself. It only 'works' because there are two or more actors; and it 'works'
 Introduction 19

through each actor taking account of information provided by the other to

model contingent AIP strategies. The shared meanings which are the
product of this process are social in the special sense that they cannot be
constructed by an individual alone; they cannot be the product of any
single cognitive process by itself, however brilliant the mind. If there is
anything 'wired in' about human cognition, it may be the 'other' slot
required for establishing common meanings and for the cognitive model-
ling of social interdependence. Recent work on infants' attunement to
maternal signals suggests that this faculty develops virtually from birth
(e.g. Trevarthen 1988).
This model for the construction of a lexicon could also be used to
represent Vygotsky's view of how a child learns through joint participa-
tion with an 'expert' (Vygotsky 1962; Goody, Chapter 10, this volume).
Since differential skills/knowledge among participants would be a feature
of even the simplest hominid society, it would be interesting to incorpor-
ate this into the Hut chins and Hazlehurst model.
Mead's account of the recognition of self as both object and agent
through hearing one's own speech, and observing others' response to it,
takes on new meaning here. Indeed there is a sort of parallel between the
duality of 'speaker' and 'hearer' and that between self as subject and as
object. When we speak, we hear what others hear, and judge its meaning
partly by how they respond. So the 'circuits' between speakers are,
literally, replicated internally within each speaker. This argues strongly
for spoken language as having played a central part in the emergence of
self-consciousness as objectifiable both for each individual, and in human
social formations.
The sensitive and original work of Nurit Bird-David (Chapter 3) with
the Nayaka gatherer-hunters of south India has given new depth to our
understanding of social life in tiny, face-to-face communities living at this
lowest level of economic subsistence. Although of course they have
spoken language, in fact she found that these people speak little among
themselves. Relations between people who are in intimate proximity
without pause, year after year, are patterned by spheres of avoidance and
silence as well as by talk. With this strong sensitivity to individual
privacy, ways of calling to and of addressing people take on special
importance. Kin terms may be used like names, for address, as well as to
attract a relative's attention. This, she points out, is a basis on which
members of the community come to take alternative perspectives of one
another. The individual one person calls to as 'father' is another's
'mother's brother'. We see here the emergence of social roles.
Instead of formally explicit roles and rules, Bird-David describes (with
difficulty, as they are implicit in nature) de facto patterns which are a focus
of feelings of obligation or uneasiness. She calls these 'quasi-norms', a
20 Esther Goody

term which nicely reflects their partially binding, unframed character.

What she describes is an (open) set of individuals who have identicial life-
styles, living in close proximity, and who somehow recognize that they
feel the same way - approving or disapproving - about certain patterns of
behaviour. Clearly here quasi-norms emerge from the constraints of daily
practice. And in this tiny society individuals interact with sufficient
frequency and stability that such norms 'need not' become explicit.23 The
picture which results is not of a Machiavellian society, but rather of
simple but basic forms of cooperation, both in daily activities and in the
management of social relationships themselves.
Although they do not speak a great deal about feelings, plans or the
meaning of what they do, Nayaka use speech to manage many sorts of
social action. Perhaps the Hutchins and Hazlehurst model for the
establishing of a lexicon also fits the way in which the Nayaka establish
common meanings for patterns of avoidance, de facto roles and quasi-
norms. There are clearly important continuities between the emergence
of lexical terms and implicit patterning of action.
The chapters in Part I highlight a number of primary processes: the
challenge posed for cognitive modelling by deception, but also the power
of cooperation; a model for establishing shared meanings for a lexicon
which depends on systematic reciprocal exchange and confirmation of
information between individuals; the emergent nature of de facto roles
and implicit rules. These are not the only primary processes involved in
social intelligence, but they do appear to be basic to many of the particular
forms discussed in later chapters.

Part II The negotiation of conversational meaning

These three chapters deal, at different levels, with the fine-grained
analysis of informal conversation. They follow in a tradition based on the
work of Garfinkel, Schegloff, Sacks and their colleagues in ethnometh-
odology and conversational analysis, which seeks to identify the routines
and conventions which structure everyday interactions. These studies
have shown that ordinary everyday life is not the casual, haphazard affair
we often assume it to be. Instead, even very ordinary behaviour like taking
turns in conversation follows surprisingly precise patterns, and people
become highly agitated at any deviation. Increasingly researchers are
asking fundamental questions about the way these implicit patterns
emerge, how they influence the way in which people manage interaction,
and how they influence the construction of meanings within this
interaction.
A central point in Paul Drew's contribution is the importance of
distinguishing between cognitive strategies and strategies for interaction.
Close analysis of ordinary conversations reveals complex links between
 Introduction 21

the two. These links tend to be mediated by conversational micro-

routines such as the 'prefatory components' analysed by Streeck (Chapter
4) and the 'repairs' described by Drew (Chapter 5).
On the basis of his very detailed analysis of the sequencing of gesture
and speech in conversations, Jiirgen Streeck has isolated what appears to
be a very general 'design feature of human action' - a prefatory compo-
nent, expressed in speech and in gesture, which precedes overt action.
Prefatory components ('pre's') prefigure intended actions by projecting
indications for others to recognize. Such 'pre's' enable others to antici-
pate intended actions and to respond accordingly, thus synchronizing
understandings of the participants. In their very nature, 'pre's' are
tentative initial formulations of more important issues. They can be
picked up or ignored by the partner in interaction 'orT-stage', i.e. without
any formal rejection of the overture. Because they are tentative, 'pre's' can
easily be subsequently redefined without challenging the validity of the
original form. As an example of their function in avoiding possible
conflict in interaction, Streeck discusses an invitation made (tentatively)
in 'pre' form which was ignored and never followed through.
Paul Drew uses formal conversational analysis to look at the relation-
ship between conversational forms, cognitive strategies, and strategies for
interaction. He identifies several types of interaction sequence: teases, the
conversion of disagreements into agreements, and 'pre-requests'. In each
case he distinguishes between the structural organization of the sequence,
and the procedures through which it is utilized. An interaction sequence
involves a patterned exchange of a certain form between speakers. A tease,
for instance, challenges the validity of a prior statement in a gentle way
which requires what linguists call 'repair' - that is, a revision of the
original comment so that it avoids the initial 'error'. All these sequences
are ways of carrying through a given conversational goal (of the indivi-
dual) while maintaining the joint enterprise of the conversation itself.
Thus the actual outcome is a product of joint interaction.
The several specific analyses of conversational sequences raise a
number of critical issues for understanding the fit between individual
cognitive modelling and the interaction between two (or more) different
people. One such issue is the location of the synchronization of individual
cognitive modelling within the interdependence of joint conversation
routines and procedures. It would seem that procedures for conver-
sational sequences are culturally framed and also linguistically patterned.
Routines like a tease or a pre-request can be seen as tools available both 'to
think with' and 'to interact with'.
Another issue concerns the level of awareness in the use of conver-
sational sequences. In one sense sequences are emergent in conversation,
as when one person's comment is interpreted as meriting a challenge
which takes the form of a tease, leading to a repair of the original
22 Esther Goody

comment. AIP modelling here may be at the general level of an intention

to maintain a given style of conversational interaction; hence the choice of
a tease rather than a more direct challenge. But because the tease response
is available as a routine, detailed planning is not required for its execution.
It seems likely that routinization tends to reduce the level of awareness
with which interactions are planned and executed. Levinson explicitly
suggests this for what he calls 'heuristics' of spoken language (Chapter
11). Routinization of procedures requires recurrent practice and cultural
framing. Both conditions seem to occur where a given conversational task
is frequently encountered, as with the need for repair, avoidance of
disagreements, and securing cooperation with individual goals.
The distinction between cognitive strategies and strategies for action is
useful for thinking about both these issues. The synchronizing of AIP
modelling of plans and goals of two (or more) individuals can be seen to
occur both with cognitive strategies, and with strategies for action.
Routines must be modelled at some level, whether with awareness or not.
As routines are both linguistically and culturally framed, different actors
are using more-or-less the same model for their cognitive strategies. Thus
when these are brought into action strategies, the turns specified in the
procedures are familiar to both participants, and the sequence can unfold
smoothly.
The work of both Streeck and Drew provides striking evidence for the
patterning of interaction on a level of which we are not aware. We see the
cognitive modelling of AIP to be accompanied by communicative micro-
strategies which permit delicate adaptation to each other's responses. To
follow a formulation used by Thomas Luckmann (Chapter 8), the micro-
communicative strategies of prefatory components of conversation
permit the 'mutual adjustment of perspectives' necessary for effective
solution of any communicative problem. Intuitively one feels that the
modelling of cognitive strategies moves between levels of awareness
depending on many factors, ranging from attention, to routinization of
sequences, to degree of complexity, to whether responses of the other are
as expected, to whether a given cognitive strategy is to be translated into
action. This is clearly an area which needs further working out.
The inclusion of gesture in Streeck's work has a special relevance to the
role of language in Homo sapiens' social intelligence. John Lyons (1988)
and Merlin Donald (1991) have suggested that a language of gestures may
well have preceded spoken language. A central aspect of this argument is
the continuing importance of gesture in contemporary human conver-
sation, and indeed the grammatical and lexical complexity and richness of
sign languages used by the congenitally deaf. Streeck's evidence that
gesture is used in the subtle negotiation of topic and intention is clearly
consistent with the view of the importance of gestures in early hominid
language. The contention of both these authorities that gestural language
 Introduction 23

was prior to and separate from the emergence of the earliest spoken
language seems extreme. If indeed spoken language emerged very
gradually it may have been intimately linked with gesture from the
beginning, at the time when the vocal tract was not yet specialized for
complex speech.
If the power of language to structure the close coordination between
interacting individuals is revealed in the conversational analyses of
Chapters 4 and 5, in Chapter 6 David Good shows that there is still
pressure to maintain individual freedom of action and the determination
of meaning within the framework for cooperation constructed by and for
conversation. The empirical starting point for his discussion is the
observation that conversations are inherently ambiguous, and subject to
constant reinterpretation and shifts of meaning. He argues that a critical
difference between the use of AIP by primates and Homo sapiens arises
from 'the way the language transforms the nature of time'. Here Good is
not simply talking about the possibility of referring to past and future
events; by describing again what has already happened, spoken utterances
can give many interpretations, and later reinterpretations, to the same
event. This retrospective reappraisal becomes as important as prospec-
tive planning: 'Foresight becomes focused on what hindsight can do' (p.
140). This creates the need to include in cognitive modelling both future
and retrospective planning, and therefore makes different and much
greater cognitive demands on AIP.
This possibility of restructuring meanings as a conversation progresses
means that participants can slide out of ostensible social contracts, a
Machiavellian realization of AIP in spoken language. It now becomes
advantageous for each participant to maintain sufficient ambiguity to
permit shifting meanings favourable to his objectives; at the same time
each tries to constrain possible future re-interpretations by others.
Good is here delineating the implications of the full potential of spoken
language for cognitive modelling of alternative strategies. When language
has been incorporated into ways of thinking (about time, and about new
understandings of past events), and ways of managing the use of conver-
sation to achieve goals, the complexity of AIP increases dramatically.
While he points out the potential this provides for deception, it is at the
same time true that the whole enterprise of conversation itself rests on the
elaborate cooperative synchronies described by Streeck and Drew.

Part III Genres as tools that shape interaction

We have seen that close analysis of conversation reveals complex pattern-
ing linking cognitive strategies with strategies for interaction. Micro-
routines like prefatory components and repairs often act as bridges
between cognition and action. Indeed they tend to become routinized as
24 Esther Goody

solutions to particular communicative problems such as avoiding con-

flict, and establishing and controlling meanings. Good suggests that
control over meaning can itself be the object of conversational strategies,
which can obscure and may even reverse meanings as talk proceeds.
This management of meanings as strategy in interaction can be seen in
the utilization of communicative genres such as politeness forms (Brown
and Levinson 1978, 1987), questions (E. Goody 1978a) or irony as
discussed in Chapter 7 by Penelope Brown. The Brown and Levinson
analysis of the linguistic and performative uniformities in politeness
across cultures has marked a paradigm shift in sociolinguistics, and
together with the Humphrey paper led to framing the notion of anticipa-
tory interactive planning outlined in this Introduction. This early work
threw into sharp relief the importance of modelling intentions. To be
effective the modelling of others' possible actions must incorporate
assessment of their intentions - 'we positively seek out intentions in what
people say and do. There seems to be a universal premise that other
people's behaviour will be goal-oriented;... Very often this results in the
imputation of intention where no clear intention message is provided' (E.
Goody 1978b: 12). This premise of intentionality makes interaction risky,
since our intentions are always apt to be interpreted in terms of other
people's goals rather than our own. Thus our actions easily are seen as
threatening others' self-esteem or their capacity to reach their own goals.
The genre of politeness is a mode of engaging in interaction (usually, but
not necessarily through conversation) which signals our intention not to
threaten either self-esteem or others' goals. But the power of the Brown
and Levinson model lies in its including factors influencing the ability to
facilitate or impede others' goals: power, social distance and ways in
which self-esteem and competence might be vulnerable. As Brown says
here 'This model ... makes very strong assumptions about humans'
abilities to reason reflexively about each others' desires and intentions'
(p. 154). Speakers have to be able to modify their communicative
expression of intentions accordingly, as hearers must 'be able to read such
modification as evidence that the speaker is inferring particular things
about their own views, desires and intentions' (p. 154). And she notes that
communication of this elaborateness seems so complex as to be imposs-
ible. Use of conventional forms - e.g. for politeness - is clearly one way of
cutting through these layers of reflexive inference. And indeed this is
presumably how such conventional forms emerge in every society.
Because Brown's analysis is sofine-grainedshe is able to show that Tzeltal
irony has become normatively stabilized around three sub-modes, off
record and ambiguous, negatively polite and positively polite; as well as
being able to convey the intention to be impolite. However her analysis
shows definitely that conventional forms in turn come to be used in
 Introduction 25

nuanced ways which depend on the understanding of presumed inten-

tions in this particular context. Further, choice of how irony is used is
strongly influenced by social role. Women frequently use irony to
emphasize shared opinions and understandings, and to elicit sympathy.
Where normally women stress solidarity and suppress anger, in the
context of confrontation in a court the same genre is, as it were, inverted to
convey hostility.
We see at work here the emergence of routines for handling particular
tasks, but the use of these routines then itself becomes an element in the
cognitive and strategy modelling. It is as if where there is a premise of
cooperation, as among kin, friends and other women, irony is employed to
express solidarity; where there is a premise of hostility, as in the
antagonistic confrontation of a court case, irony is used to express
hostility; and where the relationship is ambiguous the use of irony
achieves at least a surface agreement, and 'provides a negotiated end to the
topic' (p. 169). As Good argues, there are situations in which such tools
will be employed to maintain ambiguity, and thus enhance control over
emergent meanings (Brown's 'ambiguous case').
Thus we see that Tzeltal use of the ironic genre does not so much focus
on problems of sustaining communicative interaction, as on problems of
managing meanings and social relationships themselves. In afinaltwist to
the story, Brown finds that irony is unusually elaborated in this Amer-
indian society which, for the 500 years since the coming of the Spaniards,
has been in a subordinate position in its own land; and that within Tzeltal
society it is structurally subordinate women who make the greatest use of
irony. Irony, she argues, is a common strategy for the underdog since it
permits off-the-record complaints that can be innocently disavowed if
necessary. It seems that where a society or social segment is structurally
subordinate, routines which are individually effective may come to be
culturally elaborated as a widely shared genre.24
The fundamental question for Thomas Luckmann (Chapter 8) is how
communication and effective interaction are possible between discrete
individuals. (This is, in another form, the problem of establishing shared
lexical meaning discussed in Chapter 2, and reflects Levinson's puzzle-
ment over the possibility of mutual understanding, in Chapter 11.) The
answer lies, he believes, in mechanisms for the reciprocal adjustment of
perspectives, which in turn both requires and permits patterning of
communication. This occurs on many levels ranging from the 'givens' of
phonology, morphology, semantics and syntax, through linguistic styles
and dialects, to the implicit and explicit shaping by roles, rules and
institutions.
Adjustments of perspective may take place automatically when situa-
tions are clearly predefined. Here one might say that AIP planning has
26 Esther Goody

become routinized. At the other extreme, where background knowledge

is faulty, or doesn't work, 'almost everything . . . must be negotiated
"locally"' through trial-and-error explorations of perspectives of the
participants. Luckmann sees genres as emerging from replicated exper-
iences of solving communicative problems of the same kind falling
between these two extremes of routinization and trial-and-error. Specific
communicative problems are normally solved in communicative repair
procedures (cf. Streeck and Drew, Part II). It is the normal mutual
assumption of actors that social models exist for this. 'When such
modelling reaches a high degree of structural "closure", and when it is
associated with inter subjectively recognized formal constraints, we are in
the presence of comunicative genres' (pp. 179-80). Thus communicative
genres are socially constructed models for the solution of specific types of
communicative problems. They become part of the stock of shared social
knowledge of a particular society at a particular time. The general nature
of such communicative problems is in the form of conversational dia-
logue. 'The elementary function of communicative genres in social life is
to organize, routinize and render (more or less) obligatory the solutions to
recurrent communicative problems' (p. 182).
Genres emerge because the underlying fundamental problem in com-
municative processes is the reciprocal adjustment of perspectives. (This
reciprocal adjustment of perspectives is also a prerequisite for effective
AIP modelling of interaction strategies.) Genres assist in the reciprocal
adjustment of perspectives in two ways: by providing easily recognizable
markers for the mutual adjustment of perspectives (e.g irony, politeness
forms); and because genres help to maintain interlocking perspectives in
the production and reception of communicative interaction.
Just why the attribution of intentions should be so central becomes
clear in the view of human intelligence as essentially social. The model-
ling of others' possible responses to one's own actions (AIP) depends on
judgements of others' intentions. The 'face wants' of the Brown and
Levinson politeness model are fascinating and fundamental. The concept
of negative politeness refers to 'the desire not to be impeded in one's
actions'. This is indeed basic to all AIP strategies; they are about how to
reach goals with minimum interference. In this sense it is prior to
particular strategies.
The desire to be approved of, liked, admired, validated - which is seen
to underlie 'positive politeness' - is surely a universal human characteris-
tic. It is so familiar we take it for granted. We are so accustomed to living
with, and managing, this need to be approved that despite the consider-
able literature on 'face' (beginning with Goffman's classic The Presen-
tation of Self in Everyday Life (1962)) it has not really been addressed. But
why should this matter so much to us? Perhaps because any expectation
 Introduction 27

that others will cooperate with our AIP strategies depends on their
valuing us sufficiently to do so. If we feel approved, liked, then it is
possible to expect that the other will be prepared to recognize our goals as
valid as well as his own. Of course if the other really values us, then
cooperation will become his own goal, and he too will seek the negotiation
AIP strategies that are jointly advantageous. If we sense that we are not
approved of, not taken as valid, then we expect 'our acts to be interfered
with' and the interaction takes on a confrontational form. An AIP strategy
must take account of this. Levinson argues (Chapter 11, this volume) that
cooperation is more complex than competitive, agonistic relations, and
has been more fundamental in the emergence of human society. Indeed it
seems likely that we tend to attribute a general 'cooperative' or 'antagonis-
tic' stance to others first, and that this then shapes more specific
attributions.

Part IV Expressions of a social bias in intelligence

The final part is concerned with ways in which contemporary human
intelligence can be seen to express the social bias originally identified in
work on sub-human primates. The analyses of conversational inter-
actions in earlier parts are clearly also about social bias. They document
finely the intricacies of the interdependence of speakers in the micro-
dynamics of conversation, and the emergence of culturally idiomatic
communicative genres for managing interaction. And indeed the chapters
in this section draw also, in one way or another, on the way in which
spoken language is employed to solve problems of social living. Prayer
and narrative clearly depend on speaking, and express in this sense the
social bias of what Luckmann terms 'dialogic form'. And in a new
approach to understanding divination, David Zeitlyn shows that this too
can be seen to use a dialogue template.
However these authors are also concerned with social bias rooted in the
cognitive modelling of contingent interdependence itself. They argue, in
different ways, that this leads to a concern with understanding and
explaining our experience in terms of our motivations and the intentions
we attribute to others. What happens to us is understood as the result of
purposive action, and we tend to explain what happens by the attribution
of appropriate intentions to the agents - whatever they may actually have
intended (E. Goody 1978b).
Such a primacy of intentionality is consistent with our cognitively
modelling the constraints of social interdependence (AIP). For know-
ledge of what another intends to do can only enhance the accuracy of our
model. Zeitlyn (Chapter 9) and Levinson (Chapter 11) argue that the
social bias of attributing intentionality to events which are in fact random
28 Esther Goody

influences the very way in which we define problems and understand

solutions. Both Levinson and Michael Carrithers (Chapter 12) see human
thought as preoccupied with making sense of experience using social
relations and intentionality as an explanatory model. Carrithers's term
narrativity refers to our need to organize our thinking, as well as our
talking, so as to construct a story which makes sense of our experiences as
this is defined in our particular society and culture. And of course by
telling stories to each other, we establish common understanding of
shared experiences, yet another expression of the Hutchins and Hazle-
hurst model in Chapter 2. Prayers employ the dialogic mode to actually
seek to change the way the world is. We may use prayer to seek to
influence the intentions of spiritual powers; and we may in part do this by
asserting the excellence of our own intentions. Social bias then involves
the use of social interdependence as a model for thinking with.
David Zeitlyn analyses his detailed transcriptions of Mambila
divination sessions as though they were conversations between Nggam
(divination itself) and the local diviner. Since the effectively random
movements of the burrowing crab used in Mambila divination yield
contradictory answers to the questions posed by the diviner, this seems to
support the view that such divination cannot provide rational solutions to
problems. The diviner, however, does not treat the responses as contra-
dictory, but as indicating that he must formulate the questions differently
or include a wider range of factors. In fact there is experimental evidence
that when sophisticated westerners are given contradictory answers to
personal problems they too succeed in rationalizing these in very similar
ways. In both cases the questioner assumes that the one answering must
be giving knowledgeable and rational replies; that the problem lies in the
way he is framing the questions or interpreting the answers. Empirically it
seems that there is a 'premise of intentionality' underlying divination
which matches the constraints which make dialogue possible (Grice
1989).
The premise of intentionality underlying divination is also consistent
with a view of social intelligence based on the modelling of causality in
terms of contingent interaction between actors. It is as though there is an
'other' slot in our cognitive model in which we represent the responses
and plans of those with whom we interact. The underlying question in
Esther Goody's chapter on prayer is 'Why, when there is no direct
evidence of anyone listening, do people pray?' Of course on one level
prayer reflects the cultural construction of cosmologies around deities
demanding prayer. However there are religions like Buddhism which
formally reject prayer, in which, nevertheless, prayer plays a major role.
And there are individuals who say they do not believe in religion, but who
turn to prayer in time of crisis. Within the framework of social cognition,
 Introduction 29

prayer 'makes sense' as a way of solving problems as though they were

caused by a social other. Through speaking/praying to the other we seek
to account for our own actions, to understand His intentions, and solicit
His help. The blissful peace which accompanies deep meditation may
indicate a suspension of the AIP modelling process, with its demands of
representing multiple contingencies and elusive intentions.
A number of studies have reported that otherwise highly intelligent
people demonstrate 'a curious inability to use logical reasoning'. In
Chapter 11, Stephen Levinson brings together a wide range of material to
confront this paradox, arguing that people tend to use forms of reasoning
which are effective for solving problems arising out of social interdepen-
dence, and they prefer these forms, even when presented with 'logical'
problems. This is indeed a 'social bias in intelligence'.
Parallel with this account of social biases in human thought, Levinson
pursues the problem of how it is possible for humans to successfully use
language. How can we possibly know what is meant by gestures, and even
words? (This problem is also central for Hutchins and Hazlehurst, and for
Luckmann.) He considers several previous explanations of this puzzle
andfindsthem inadequate. Instead he suggests that understanding is only
possible through following signals which indicate the speaker's inten-
tions, and that these greatly restrict the otherwise incomprehensible
range of meanings that might apply. If we depend on such heuristic
signals to interpret each other's speech and actions, then in learning to
speak and understand we must learn to think of causation as embedded in
meanings and intentions. It is precisely such assumptions which experi-
mental subjects do in fact make when they fail to 'think logically'. Thus,
Levinson argues, in order to make sense of the actions and talk of those we
interact with, we have had to learn to take account of their plans and
intentions. We have learned, in effect, a premise of intentionality, and we
apply this way of thinking even to those events which are quite random.
Both the discussion of divination in Chapter 9 and of prayer in Chapter 10
offer evidence of the human reluctance to accept that events occur by
chance.
In focusing on the narrative mode of understanding, 'the ability to
create, narrate and comprehend stories' (p. 261) Michael Carrithers may
appear to be abandoning an interest in social intelligence for folklore.
However the cognitive psychologist Jerome Bruner has recently argued
that there are two distinctively different sorts of thought, the logical mode
and the narrative mode; and that 'narrative deals with the vicissitudes of
intention' (1986:17). Carrithers broadens the concept of narrativity by
pointing out that it functions simultaneously on the cognitive and on the
social plane, enabling us to both understand theflowof social events, and
to convey this understanding to others. Further, he suggests that sharing
30 Esther Goody

stories is a 'particularly powerful form of interactive planning' because it

brings together accounts of the past and of present events in a way that
makes possible planning for future joint action. We might add that
narrative brings the individual planning of AIP out into the arena of
shared action; planning becomes a social product. In the case of Kaluli
stories the describing of past events vividly recalls both the social
dynamics - witchcraft, a fight, calling brothers-in-law to plant bananas -
and the associated emotions. It is the combination of the meaning of past
events and the evoking of associated emotions which makes plans for the
future seem to follow necessarily from what has already happened. Like
David Good, Carrithers stresses the power of spoken language to
transform our experience of time. There are also interesting resonances
here with Good's stress on conversational ambiguity, potential reversi-
bility and contests for control over meanings.
To emphasize the socially dynamic nature of story-telling Carrithers
offers the term 'confabulate', for 'together making a story'. However, he
suggests that in making their stories, people also confabulate - make
together - their social worlds. This confabulation creates several kinds of
reality: for the individual it creates an account of personal experience that
makes 'sense', and can continue to structure past and future experience;
for the set of people who together make the narrative it creates a shared
view of past, present and future to which they are all in some measure
committed, since they contributed to its creation; and for both individuals
and the group what is made is social action, perhaps an exchange of pigs,
perhaps a vengeful attack.
This discussion of narrative raises again the issue of how shared
meanings are established. But where Hutchins and Hazlehurst model a
possible basis for naive strangers to develop shared meanings, in our real
worlds narrative is built on 'cultural scenarios', templates for meaningful
relationships between certain actions and certain consequences.25 Car-
rithers also stresses the particularity of personal experiences in narrative.
This not only provides 'news value' but as it were anchors the story in our
immediately shared world; it is our story. If the story is to legitimately
make social action, then it has to be certified relevant. For ultimately it is
the joint participation in 'doing confabulation' which gives it its power to
create shared meaning and plan action.

Notes
1 This modelling ability seems to underlie-what Sperber and Wilson (1986;
Sperber 1990), refer to as inferential communication in which 'communication
is achieved by producing and interpreting evidence' without resort to a
linguistic 'code' (1986:2).
2 See Bickerton 1990, ch. 6; Phyllis Lee, personal communication.
 Introduction 31

3 See especially MacNeilage, 'The evolution of hemispheric specialization for

manual function and language5 (1987); MacNeilage, Studdert-Kennedy, and
Lindblom, 'Primate handedness reconsidered' (1987); Greenfield, 'Language
tools and brain: the ontogeny and phylogeny of hierarchically organized
sequential behavior' (1991).
4 Hominid adaptations of the primate vocal tract for speaking are costly; both
breathing and swallowing are less efficient, choking is not uncommon, and
there is a strong possibility that cot deaths in human infants are linked to these
changes (Lieberman 1991:53-7; Phyllis Lee, personal communication).
5 The evidence from fossil hominid endocasts suggests a radical difference in
the structure of the brains of Australopithecus and H. habilis, with the latter
having an enlargement of the parieto-occipito-temporal junction (POT)
which has been identified as 'the association area of association areas'
(Geschwind 1965, quoted by Wilkins and Dumford, 1990:763). It has been
argued that the presence of a developed POT is the basis of the ability for
spoken language. If this datum and the related arguments are substantiated by
subsequent archaeologicalfindsthen even habilis may have used an early form
of protolanguage (Wilkins and Dumford 1990). Such a protolanguage might
well have combined gesture and vocalization (pace Bickerton and Donald -
neither of whom offers a convincing argument as to why early protolanguage
need have used a single channel). It need not have required changes in the
vocal tract; Lieberman has argued that the ape vocal tract would be adequate
for a simple spoken protolanguage z/they had the necessary neural structures
to support it (1991).
6 Malinowski, writing on language use in the Trobriand Islands in relation to
meaningful communication noted what he termed 'phatic communion' -
conversational forms which actually conveyed little information but which he
saw as critical for managing social relationships.
7 Lieberman has stressed the significance of speed of speaking and of decoding
speech for the transmission of long messages within the constraints of human
short-term memory and information processing (e.g. 1991:38); the linguists'
paradox of the human ability to produce an infinite number of new sentences
from afinitelanguage corpus is due to the recombinative potential of syntax.
8 'The general process of automatization ... allows us to learn to rapidly execute
complex goal-directed patterns ...' It 'converts a series of learned motor
instructions into a "sub-routine" that is stored in the motor cortex and
executed as a complete whole' (Lieberman 1991:48). The concept refers
specifically to the learning of motor control subroutines. Lieberman is
discussing the automatization of speech production which combines planning
of what to say and how to say it. Grammar and syntax similarly involve
combinations of what is said and how it is said, though the 'how' has shifted
from the motor cortex to the areas responsible for planning speech. If learned
motor instructions can be automatized, why cannot learned grammar and
syntax instructions also be automatized?
9 The implications of classification for cognitive complexity are discussed by,
among others, G. Miller (1956), Miller and Johnson-Laird (1976) and
Goodenough (1990).
32 Esther Goody

10 The difficulty with the Sperber-Wilson discussion of the opposition between

inferred and coded meaning lies in the problem of how individuals know what
to infer from others' behaviour. The implication is that coded meanings exist
'out there in language' (but are cognitively decoded) while inferred meanings
exist in others' behaviour and are also made sense of cognitively. However,
most of the clues on which we base social inference are either partly cultural (in
which case they too have a culturally coded meaning, it just isn't in words), or
they take their meaning from shared experience, as it were a private code. The
Chomskian stress on language as code masks the important fact that language
constantly creates a shared focus of attention and shared meanings in a far
more explicit way than where communication is based on inference from non-
linguistic behaviour.
11 The term 'mands' has been used to represent this generic aspect of control in
conversation (Whiting and Edwards 1988).
12 'Phatic' presumably from 'emphatic' meaning 'forcibly expressive' from the
Greek emphaino, an appearance, a declaration; this appears in Malinowski's
contribution to Ogden and Richards's pathbreaking The Meaning of Meaning
(1927).
13 A private totemic world would be a psychotic delusion; but shared by others it
becomes the basis for birth, marriage and death - for social reality.
14 There is a large recent literature on 'activity theory' in relation to learning.
The interested reader may gain an overview by looking at: Vygot sky (1978),
Luria (1979), Wertsch (1985a), Rogoff (1990).
15 Reported by Ryan in the UK and Goody in northern Ghana, cited and
discussed in E. Goody (1978a).
16 The relationship between intentionality and awareness, how these relate to
goal-directed behaviour, and in turn to routinization and automatization is, to
say the least, problematic. The work of Leslie (1987), Velmans (1991) and
Dennett (e.g. 1983,1988) indicates the range involved and some of the issues.
The planning of social intelligence (AIP) is probably best seen as shifting
between levels of intentionality and awareness in achieving socially con-
strained goals.
17 The work of Argyle on gaze and Hall and his colleagues on body language and
response to cultural variations in patterns of proximity and touching are
particularly interesting examples (Argyle 1968, and Hall 1959).
18 There is a huge anthropological literature on 'the gift'. Apart from Mauss's
original article, the single most significant contribution is probably Marilyn
Strathern The Gender of the Gift (1988).
19 This theme is developed in the 'Working papers on implications of a social
origin of intelligence', E. Goody n.d. [1990; 1993].
20 This view draws heavily on S.F. Nadel's discussion of role in relation to the
articulation of individual and social institutions (1957).
21 The recent prominence of vigilanti movements in places as far apart as
Tanzania and New York reminds us that this process is never finished; new
problems of synchronization are likely to lead to the emergence of new
mechanisms. See R.G. Abrahams 1989.
 Introduction 33

22 Bruner's discussion of narrative intelligence begins by proposing that inten-

tion constitutes a primitive category system in terms of which experience is
organized, indeed one even more 'primitive' than causality (1986:18-19). This
is exactly what we would expect with a social intelligence based on modelling
actions of others, which is dependent on their intentions.
23 One wonders what is the significance for this non-framed implicitness of
norms of the total absence of formal authority roles which might back up
assertions of what is right. Perhaps if there is no 'might', there is no 'right'. See
E. Goody 1987.
24 The cultural elaboration of the insult in early modern Italy as described by
Peter Burke is another instance of genre shaped by 'local' constraints - here
class, status and concept of honour (1987).
25 The cultural shaping of templates for narrative is of critical importance for
establishing shared meanings. This can be particularly problematic for
children of different sub-cultures in early schooling. See Heath 1983 and
Michaels 1986.
PART I

Primary processes
 RICHARD W. BYRNE

1 The ape legacy: the evolution of

Machiavellian intelligence and anticipatory
interactive planning

There appears to be a yawning gulf in sophistication and complexity

between the subject matter of social anthropology - ritual, language,
religion, negotiation, taboo, obligation - and the behaviour of a group of
monkeys. But unless one takes the view that the possession of language
utterly changed the mental life and cognitive structures of our ancestors
(and such a view will see any research on the origin of human mentality as
a fruitless task), there must be an evolutionary connection.
In order to know in what spheres of human activity this connection
might become visible, and what form any legacy of our evolutionary
heritage might take, it would helpfirstto know how our ancestors came to
be intelligent: what selection pressures led to the evolution of human
cognition? While the traditional view has emphasized the impact of tool-
making, ever since suggestions were made of an alternative, social,
pressure towards high intelligence (Jolly 1966a; Humphrey 1976), there
has been interest in developing social models of human intellectual
evolution. Perhaps more surprisingly, there is considerable evidence
supporting the basic idea (see papers in Byrne and Whiten 1988).
Primates display remarkable social complexity in their social manipula-
tions. Since the ends of these manipulations are - at least in non-humans -
always selfish, even where the means involve some cooperation, and since
the process often appears impressively cunning, Byrne and Whiten
(1988) named this Humphrey-Jolly hypothesis 'Machiavellian intelli-
gence'.
Distinguishing among the alternative theories depends on finding
phylogenetic differences in intelligent behaviour among animals other
than humans, and this chapter will sketch the evidence that some non-
human primates differ intellectually from other animals. In thefirstplace,
these data make a case in favour of accepting social complexity as the
evolutionary pressure that started the chain of increasing intellectual
sophistication that led, ultimately, to the stuff of social anthropology. But
the same data, when viewed comparatively, can also suggest what mental
faculties were possessed by those species ancestral to humans.

37
38 Richard W. Byrne

55My

25My

13My

5My

strepsirhine monkey orang chimp human

Figure 1.1 Modern evolutionary taxonomy of primates. The diagram is
a cladogram, showing the pattern of shared ancestry of the five primate
groups referred to in the text. The evidence comes largely from
molecular taxonomy, including immunological distances, DNA hybrid-
ization and especially the comparison of detailed sequences of large
fragments of DNA - the genetic 'code'. All these methods agree on the
basic pattern, and there seems no room for any doubt that this is correct.
The dates of each split (and thus of the inferred common ancestor of the
modern forms) are less certain. However, those given here are consistent
with recent sequencing of large strings of both mitochondrial DNA and
of a pseudogene, an unexpressed portion of nuclear DNA. The patterns
thus found have been calibrated with the most reliable ape fossil now
known, an ancestral form of orangutan (Sivapithecus) from 13 million
years.

It was in an animal with these faculties - and in no other - that language

and all other uniquely human skills developed, and so there is every
reason to suppose that modern modes of human cognition required these
ancestral skills, or at least that modern thinking accommodated to primate
skills rather than replacing them. Thus the chapter can also serve to chart
the legacy of intelligence's evolutionary origin in social manoeuvring: the
styles of thought that no being whose intellect was evolutionarily derived
via a non-social route would possess. As such, its aim is to allow any such
'residues' to be identified the more easily, if they do indeed occur in
human thinking.

Classifying primates
It will be useful first to divide the modern primates into three groups, of
differing relatedness to human beings (see Figure 1.1). Firstly, there are
strepsirhine primates such as lemurs, galagos and lorises (for purposes of
using the comparative method to derive the 'best-bet' characteristics of
 The ape legacy 39

ancestral species, this grouping replaces prosimians, a category which

also included the tarsiers - now classified with the monkeys). Strepsir-
hines are separated from humans by around 55-90 million years of
independent evolution (R.D. Martin 1990), and in many ways appear
similar in form to the earliest primates known from fossils. The rest of the
extant primates (which are called haplorhines), can be conveniently
divided into the monkeys and the apes. Monkeys includes tarsiers, New
World monkeys and Old World monkeys, taxa which split off from the
line of human descent at various times, the most recent of which was 25-
35 million years ago.
However, humans themselves are part of the other group, the apes, and
among the apes it is now certain that humans are much more closely
related to some species than others. There are many uncertainties in
attaching even an approximate time-scale to any of the measures of
evolutionary similarity {ibid.), but there is now no doubt that the African
apes (chimpanzees and gorillas) are more closely related to humans than
are the Asian apes (Oxnard 1981; Sibley and Ahlquist 1984) - indeed,
from the chimp point of view, humans are a closer cousin than orang-
utans! As an illustrative example, taking the Old World monkeys to have
diverged from apes at 30 million years and using immunological similar-
ity, gives the last common ancestor of human and chimpanzee a date of 5
million years, not long before the earliest known Australopithecines, while
the orangutan is separated by 10 million years (Sarich and Wilson 1967).

Comparing primate brain sizes

In contrasting these three groups, use will be made of brain size as well as
behavioural data. The reason for using such an indirect measure of
intellectual ability, which tells us nothing about the type of intellect and
anyway itself needs to be validated as a measure, is a certain desperation.
There is a long history of work in comparative psychology, trying to find
out how animals differ in their intellectual abilities by testing them with
problems set in the laboratory (see review by Warren 1973). Over the
years, many of the differences that have been hailed as due to intelligence
have been found merely to reflect differences in the ease of motivating
different species, or in the sensory and motor aptitudes of the animals.
Indeed some would go so far as to say that the general picture shows no
intellectual differences between any animal groups (Macphail 1985).
Certainly, we now realize that intelligence is such a slippery concept to
define and quantify fairly that it seems less unreasonable - and it is often
done - to take the brain size of the animal as an indirect measure of it - one
which at least can be accurately estimated.
To compare brain sizes properly, it is necessary to take account of two
40 Richard W. Byrne

characteristics of growth and form. Larger animals, in general, have

larger brains; this is not particularly surprising, since for most mammals,
much of the brain is taken up with sensory and motor processes.
Secondly, as the absolute size of living things changes, so the relative
proportions of their parts is liable to change. In this case, absolutely larger
animals have relatively smaller brains than one would expect from
linearly enlarging smaller animals. A technique that takes account of both
of these difficulties is called allometric scaling. In it, a double logarithmic
plot of something, in this case brain size, is made against body size for the
given group of animals. This forces the species points onto an approxi-
mate straight line. Then we can see whether any particular animal in the
group lies above the line (has a relatively larger brain than one might
expect), on the line (has average brain size), or below the line (has a
relatively small brain). This technique has limitations (see below), but it is
the one used standardly now to compare animals' brain sizes.
When we apply allometric scaling to brain sizes of mammals, we find
that primates as a whole are larger brained than most other groups
(Jerison 1973). And using the same techniques, wefindthat humans have
brains three times as large as we would expect from even a monkey of
human size (Passingham 1982), which of course tends to give us confi-
dence in the relationship between relative brain volume and intelligence!
But when we separate the strepsirhine primates from the rest, the
strepsirhines emerge as a perfectly typical group of mammals (Pass-
ingham and Ettlinger 1974): in other words, in general they have brains
about the size one would expect from their body sizes. The monkeys and
apes, however, have brains twice as large as normal mammals of their size.
What are they using these large brains for?

Comparing intelligence among primates

Primates have an everyday reputation for greater cleverness than most
mammals, but can we quantify this? Certainly monkeys and apes do not
emerge as outstanding when intelligence is measured in the laboratory
(see review by Passingham 1982). The best that can be said is that when
animals of various groups are trained on the same task, it frequently turns
out that monkeys are able to reach the required criterion very much
quicker than rats, cats, squirrels, or whatever (Passingham 1981). Conver-
sely, when animal curiosity is tested, for instance by presenting novel
objects to animals housed in a zoo, monkeys remain interested in the
objects for longer than other animals (Glickman and Scroges 1966).
These differences in curiosity and mental speed are matters of degree,
and fairly minor degree at that. However in recent years there have been a
number of suggestions that the lack of species variation is an artifact of the
 The ape legacy 41

tasks used (Jolly 1966a; Humphrey 1976). If the increase in brain size of
monkeys and apes was promoted by social needs, and the resulting
intelligence was therefore shaped for social skill, then the gadgetry of the
laboratory will be a very blunt instrument for its assessment. Have we let
our own technological superiority, perhaps largely developed since
language evolved, mislead us into believing that this was what intelligence
was all about?

Socially smart monkeys

In the social arena the distinctiveness of monkeys from strepsirhine
primates (and from non-primate animals) is more striking. This is
apparent in several ways.
Most animals interact in dyads with one other animal at a time, but in
monkeys interactions are frequently triadic, with a third party crucially
affecting what happens in the dyad. Not only may the third party choose
to intervene, but the interactants sometimes visibly take this into account:
for instance, a threatened female hamadryas baboon may run to and sit in
front of her harem-leader male, so that any threat directed to her is
necessarily also directed to the male (Kummer 1967).
While many animals compete with fighting strength or weaponry,
monkeys rely far more on social alliances to give them power and influence
in social groups (see review by Harcourt 1988). These alliances may often
be among close kin, but in addition long-term friendships are found
(Ransom 1981; Strum 1983; Smuts 1985); these friendships persist over a
number of years, and confer mutual benefits - sometimes the rewards are
of different kinds for each participant, as in a barter system. In building
up friendships, or merely putting together a temporary alliance, the 'trade
currency' that is used among monkeys is social grooming. As a result,
monkeys groom each other far more than would be necessary for health
purposes.
Primatologists analyse the social dynamics of monkey societies in terms
of patterns of kinship, friendship, and rank. Recent work has shown that
the monkeys themselves do the same: monkeys are socially knowledgeable
as well as socially complex. In both macaques and baboons, individuals
who have been attacked aim 'redirected' aggression to innocent parties,
but the choice of victims shows that they are well aware who the relatives
or friends of their persecutors are (Judge 1982; Smuts 1983), a sort of
monkey vendetta. In experimental tests, monkeys easily learn and use the
concept of 'mother of (Dasser 1988). And a series of experiments
(Seyfarth and Cheney 1988; Cheney and Seyfarth 1988), playing back
pre-recorded calls to vervet monkeys to study their reactions under
controlled conditions, has shown that they have knowledge of dominance
42 Richard W. Byrne

relations of third parties as well as their kinship, and even know the
membership of vervet groups which they have never themselves entered.
This knowledge shows up in the strength and type of their reaction to calls
which are not specifically directed at them, yet convey information about
other individuals. For instance, individuals normally do not react much
to the calls of vervets who are not members of their group. But when a call
is artificially manipulated to suggest that one of these animals has
transferred groups (so now calls from an unexpected direction), the
hearer's reaction shows its surprise. It should be noted that, while
monkeys show knowledge of one another's kin and friends, they may not
distinguish between these concepts: in most cases, and all those experi-
mentally tested, the relationship could be subsumed under some rubric
like 'looks after', and exactly why monkey A looks after monkey B may not
be clear to the monkey.
Finally, monkeys also differ from strepsirhine primates in the way that
they use tactics of deception with some regularity to achieve goals in their
everyday lives (Byrne and Whiten 1990, 1992). For instance, a female
monkey, living in a small group with a powerful male who prohibits her
social or sexual contact with other subordinate males, may use a number
of tactics to give her the freedom she desires. She may simply 'get left
behind' so that she is fully out of sight of her leader male before she
socializes or copulates; she may carry out her actions with unusual
quietness, for instance suppressing the copulation calls that she would
normally make; or, perhaps most subtly of all, she may remain partly in
the sight of her leader, but adjust her position so that she can carry out
prohibited acts which cannot be seen from his viewpoint. All monkey
groups sometimes use tactical deception, but the African savannah
baboons do so significantly more than others (ibid.).
All this implies that monkeys are in some way 'better' than most
animals, including the 25 per cent of primates classed as strepsirhines, at
certain kinds of learning - and in particular they are better at representing
socially relevant information about conspecifics and using this information to
gain rewards in group living. It seems unlikely to be a coincidence that this
enhanced ability to learn and plan, which largely shows up in social
contexts and in taking account of the social attributes and typical
reactions of other players, occurs in animals whose brain volume is also
much larger than would be expected in animals of their weight.

Empathic apes
But how do humans compare intellectually with their much closer
relatives, the great apes? In terms of relative brain size, apes do not
notably differ from monkeys (Clutton-Brock and Harvey 1980). However
 The ape legacy 43

one must add two caveats to that statement. Firstly, the method of
working out a relative brain size assumes that the extra volume, over that
minimally required to service the bodily needs of sensory and motor
systems - in other words the 'computational part' - is some multiplicitive
fraction of total volume. So, 5 per cent extra brain is equally useful for
intelligence in a 5kg animal or a 50kg animal, even though the extra brain
tissue is far larger in the second case, and contains far more neurons. This
is a very odd assumption for anyone used to computational machines,
since these are chiefly limited in power by the number of their elements.
One can instead make a different assumption, that of additive volumes,
in which what matters is not the percentage of the total that is in excess,
but the absolute volume free for computation (and the number of
neurons, which also scales allometrically with brain volume at a different
slope). This assumption is less tractable mathematically, but on it the
living great apes all have more 'extra neurons' than those of any monkey
(Jerison 1973), since they are much larger animals than monkeys.
(Fortunately for our egos, humans both extant and extinct turn out to
have considerably larger numbers of extra neurons than do modern apes.)
The great apes also have a generally larger 'neocortical ratio', the ratio of
neocortex to the rest of the brain, than do monkeys (Dunbar 1992b), and it
is the neocortex which has chiefly enlarged during primate evolution.
The second caveat is that there are problems in trying to compare
allometrically the brains of animals which differ in diet type: a larger gut
makes the abdomen larger but requires little brain expansion, so it may be
invalid to compare species adapted for different diet qualities. Scaling
against body length instead of weight helps a bit, but the length of primate
bodies is still heavily influenced by gut size. Though all great apes have a
broadly similar type of gut, they do vary in their adaptedness to coarse
food, i.e. in the size of their large intestines. In fact, the gut sizes rank
gorilla > orangutan > chimpanzee > > human. This is exactly the inverse
of their relative brain sizes!
In any case, there is little doubt in most researchers' minds that in fact
apes are at least as adept socially as monkeys. Yet this has been less studied
in the case of apes, apparently because apes are less easily manipulated
into test situations. Many of the tests of monkey social knowledge have
used field playback of calls, a technique which has only been done
successfully with apes in the very simplest and most obvious of ways.
Perhaps the very difficulty of experimentally manipulating apes should
tell us something about their intelligence! Where there is evidence for the
apes, it is consistent with the idea that apes are just as socially smart as
monkeys.
Chimpanzee and gorilla societies show female transfer (and in general
they are patrilocal), so the monkey hallmarks, of power groups consisting
44 Richard W. Byrne

of extended matrilines and the acquisition of female rank by the support

of relatives, seldom occur. In the few cases where chimpanzee and gorilla
females have remained in their natal groups, they have gained rank from
the support of their mothers, just like monkeys do (Goodall 1986, for
chimpanzees; personal observation for gorillas). Male rank is also attained
largely by social support in chimpanzees, but in a different way. In a
number of known cases in the wild, the younger of a pair of brothers close
in age gains the highest rank with the support of his elder sibling. Where
such obvious kin support is lacking, the particular ways that chimpanzees
use grooming to forge friendships, and use friendship and alliances to gain
power (deWaal 1982; Nishida 1983; Goodall 1986), suggest that they have
even greater sophistication than monkeys in social manipulation. Direct
comparisons have been made to the recommendations of Niccolo
Machiavelli, as regards the benefits of rising to power with the support of
many weak allies rather than a few strong ones, and the use of the special
status of a 'kingmaker' to gain the ultimate power without ever possessing
the strongest resources.
So apes too are generally considered to be socially adept; but increas-
ingly, information is emerging that implies more than this. Converging
lines of evidence suggest that the knowledge and understanding of the
great apes (or at least some of them) are of a quite different type to that of
monkeys, going beyond the enhanced learning and complex planning of
monkeys.

Understanding mirrors
Most animals, when confronted with a mirror in which they can see
themselves, give an initial reaction that is appropriate to a strange
member of their species, and then habituate or get bored and show no
further reaction to mirrors. Monkeys are unusual, in that they can learn to
use mirrors, for instance to see round a corner and identify another
monkey there (Anderson 1984). But they systematically fail to under-
stand their own reflection in a mirror, and continue to react to it as if it
were a stranger (Gallup 1970). By contrast, many chimpanzees and
several orangutans (and recently also a gorilla: Patterson and Cohn, in
press) are able to interpret their reflection in a mirror correctly. Under
anaesthetic a coloured mark is made on the animal's face. When it
recovers, as soon as it catches sight of its face in a mirror it immediately
reaches and touches the spot. Gallup has used mirror understanding as an
indication of whether an animal has a self-concept, and it can be argued
that an understanding of oneself is a necessary basis for the understanding
of another individual's mind (Humphrey 1983).
 The ape legacy 45

Imitation
Despite the popular belief that monkeys are the great imitators of the
animal kingdom, and indeed despite a considerable tradition in primato-
logy and psychology of attributing various monkey behaviours to imita-
tion, attempts directly to test monkeys' ability to imitate have all failed
(Visalberghi and Fragaszy 1990). Instead, these experiments have shown
monkeys to be quick at learning, and to benefit from the social context by
enhancement of trial and error learning. This now appears fully adequate
to explain the innovation and spread of the special feeding techniques so
famous in Japanese macaques. Some would even go so far as to deny that
imitation has been well demonstrated in apes (Tomasello et at. 1987), yet
the ability of the captive chimpanzee Viki to understand a command
which meant 'copy the action that I am now doing', shows clearly that at
least one ape species can truly imitate (Hayes and Hayes 1952). The fact
that home-reared chimpanzees, orangutans and gorillas routinely pick up
unrewarded motor skills, such as teeth brushing and tying shoelaces,
would also be very hard to account for in an animal which could not
imitate action patterns. Bruner has long argued that the ability to imitate
requires that the mimic is able to put itself into the mental position of the
model (Bruner 1972): in contemporary terms, to represent mentally the
knowledge and plans of another individual.

Theory of mind
A series of experiments by Premack and Woodruff presented a captive
chimpanzee with short film clips, each of which depicted a human with a
problem (Premack 1988). For instance, a shivering man was seen in a bare
room with an oil heater, but no matches with which to light it. The ape
was given no formal training, but merely shown several photographs, yet
it spontaneously chose the photograph of the object that would solve the
problem - in the example given, a box of matches. The researchers termed
this emphatic understanding of the needs of another individual 'theory of
mind', and the concept of theory of mind is now used in understanding
the deficits of autistic children (Baron-Cohen et at. 1985).

Pretend play
Kittens chase balls of wool as if pretending that they were prey. This can
be most parsimoniously explained by saying that the characteristics of the
object elicit a hunting response, because of an overlap in features between
the appearance of an unwinding ball of wool and an escaping mouse. Such
46 Richard W. Byrne

a simplistic explanation cannot apply to pretend play when the conse-

quences of the pretence are elaborated and used in the play. Leslie (1987)
has argued that this form of pretence requires the holding of simultaneous
mental representations of two conflicting sets of knowledge or belief, and
as such is logically equivalent to knowing that another individual's
knowledge or belief is different to one's own. It is therefore of interest that
several apes have shown pretend play. The classic example was again the
chimpanzee Viki (Hayes 1951), who not only acted as if she were trailing a
pull-toy on a string but showed consternation and appropriate strategies
to free the toy when it became snagged on a (real) object! Other
chimpanzees have apparently invented imaginary monsters with which to
frighten their conspecifics (Savage-Rumbaugh and McDonald 1988),
and publicly availablefilmof these individuals (Savage-Rumbaugh 1986)
also shows them eating imaginary food and one of them having its finger
bitten by a child's doll. Kanzi, a captive individual of the closely related
pygmy chimpanzee, often acts as if he is holding an object which does not
exist: he may hide this invisible object and retrieve it later, all as if it were
real (Savage-Rumbaugh and McDonald 1988). In gorillas too, there are
some data of a similar kind: the home-reared Koko has used signs to refer
to a rubber hose as an elephant's trunk in play (Patterson and Linden
1981), and a wild gorilla, Maggie, was seen to collect up moss and carry
the bundle as if it were a baby, and then, when she had made a nest, to
cuddle and fuss over her imaginary baby for many minutes (Karisoke
Research Centre, unpublished records). Pretend play has not been
recorded in monkeys.

Teaching
After the death of her own baby, Washoe, the chimp taught to sign
American Sign Language, was given an infant chimp whom she adopted.
The human caretakers did not teach the infant Loulis to sign, and stopped
signing at all in her presence. Washoe used both demonstration (with
attention to Loulis's gaze direction) and physical moulding of Loulis's
hands to teach her to sign - with considerable success (Fouts et al. 1989).
Recently, wild chimpanzees too have been seen teaching their infants, in
this case to impart a technique of cracking hard nuts using a hammerstone
with another stone as an anvil (Boesch 1991). Mothers sometimes
perform actions slowly and in full view, paying close attention to the eye-
gaze of the infant and performing the action only when it is watching.
'Scaffolding' has also been seen, the mother setting up the physical
situation so that the infant is easily able to achieve thefinalgoal, at a stage
of development when it could not perform all the necessary constituent
acts or sequence them correctly.
 The ape legacy 47

Table 1.1 Distribution of the behavioural indicators of imagination in apes

pretend theory deceit

imitation play mirror of mind attribution teaching
orangutan Yes* Yes* Yes ? Yes
gorilla Yes* Yes* Yes* ? Yes
pygmy chimpanzee Yes* Yes* ? ? Yes
chimpanzee Yes* Yes* Yes Yes Yes Yes

Note: * Strong evidence comes only from individuals home-reared by humans.

Attribution of intentions
Recent work on tactical deception in primates has thrown up a number of
ways in which animals can in principle show us that they are capable of
attributing intentions to others (Byrne and Whiten 1991). That is, in
certain cases tactical deception can sometimes be shown to be 'lying'. (In
many other cases, the acts match what in humans we would assume to be
lying, but it is also possible plausibly to account for the tactic's acquisition
by reinforcement without invoking an understanding of another's inten-
tions.) As one example, a chimpanzee which evidently suspected (and
quite rightly) that another was concealing the location of food, hid behind
a tree and peeped out to unmask the deception. Here is the use of a tactic of
counterdeception, with no possible opportunity to learn by trial and error
since it relies on an appropriately designed novel action. Other diagnostics
include the righteous indignation that is visible evidence that an animal has
understood that it has been deceived; suppressing the anticipated behav-
iour of another animal before it could happen and so become a nuisance;
and in general, cases of deception where it is known that there has been no
previous opportunity for trial and error learning. Almost unknown in
monkeys, this sort of evidence has been found repeatedly with all great
apes species (Byrne and Whiten 1992).

The origin of imagination

Examples of this collection of phenomena have variously been found in
orangutans, gorillas, pygmy chimpanzees and common chimpanzees (see
Table 1.1), but not in monkeys. What they have in common is that in each
case the animal must imagine another'possible world\ in which things differ
from what it knows to be really the case at the moment. This possible
world varies in whether it is the current belief of another individual,
another version (an untrue one) of the current world as viewed by the self,
or a potential future state of the world that may or may not be desired.
48 Richard W. Byrne

Imagining other individuals' minds has been given various labels, as well
as Premack's 'theory of mind': 'second-order intentionality' (Dennett
1983), 'mindreading' (Whiten and Perner 1991), 'first-order belief attri-
bution' (Wimmer and Perner 1983).
The argument of this chapter is that a single common aptitude
underlies the wide range of behaviours catalogued above, and that the
great apes, but no other group of animals, possess this ability to imagine
other possible worlds than the current perceived truth. This ability is
lacking in monkeys, and - it seems - in autistic children (Baron-Cohen et
al. 1985). The data on tactical deception have shown that, whereas
monkeys can not apparently represent mentally what other animals know,
they are able to react appropriately to what other animals can see, when it
differs from what they can see themselves. At first sight this seems a hair-
thin distinction to make. Yet it turns out that autistic children also have no
difficulty with tasks that require them to anticipate a physical view from
another individual's position (Hobson 1984), and only systematically fail
when the task involves understanding the knowledge of another person.
Whether it is a coincidence that the one ape which has shown that it is
able to understand its reflection, imitate, pretend, empathize with others,
teach its infants and attribute intentions to others, is also the one which
routinely makes tools in the wild, must be left as an interesting unsolved
question for the moment. We simply cannot rule out the possibility that
this is merely a matter of sampling bias, when chimpanzees have been so
much more studied than any of the other apes. Equally, variations across
the apes in the basic ability to imagine possible worlds cannot be ruled
out. If present, this variation could tell us more precisely how and when
the human imagination developed in our evolution. Thus the issues of
clarifying exactly which ape species are capable of which mental skills,
and at what dates their lines of descent diverged from our own, are crucial
ones for future research.

Interpretation of actions: a double standard?

Accepting apes as able and monkeys as unable to attribute intentions to
others and imagine other possible worlds than the current reality, has
consequences that will worry some: a sort of 'double standard'. Consider
a gorilla and a vervet monkey, both described as executing a sequence of
actions that has the effect of socially manipulating a conspecific to the
actor's advantage. The sequence has the same pattern, and in neither case
is a full history of its ontogeny available (as in fact is true for many records
of infrequent but interesting primate behaviours collected on an ad lib
basis by skilled observers). Imagine that in both cases it is possible to
construct an elaborate post hoc account in behaviourist terms - the social
 The ape legacy 49

situation acted as a stimulus array and controlled the behaviour, with no

need to invoke intention attribution - but the accounts are so elaborate
and seemingly contrived as to be dismissed by any social psychologist
whose data was of human action. Then in the monkey case we should still
accept the behaviourist account, but for the ape behaviour we would now
use an intentional explanation even though in the given case intention
attribution was unprovable.
For a real example, consider what we should make of this observation of
gorillas by Dian Fossey (Byrne and Whiten 1990: 72):
The majority of the group were day-nesting within a 25ft radius, with low-
ranking Quince at the edge of the group. After intently gazing at Poppy [a one-
month old infant, Effie's daughter; adolescent gorillas are strongly attracted to
babies] from the side-lines, Quince stares in the opposite direction, circles and
begins bending down branches for a nest. After momentarily sitting in the nest,
Quince, with gaze averted from Poppy, gets up and moves a few feet closer to
repeat the act of 'nest-building'. After roughly 40 minutes and 6 'nests' later,
Quince was sitting next to the dominant female Effie and gazing directly at infant
Poppy.
The record itself contributes no evidence of attribution, yet once it is
accepted that gorillas have such an ability then an intentional interpre-
tation becomes rather compelling. Quince, adolescent and unrelated to
the highest rank female Effie, could not simply walk up to Effie and sit
down to gaze at her new infant: only Effie's own daughters can do such a
thing (present tense, since Effie is still the highest ranking female of this
group today, and is no more permissive with non-relatives than in 1976!).
Yet if at any point Effie had screamed in threat at this approaching female,
the silverback leader who would rush to intervene would see that Quince
had 'only been building a day-nest', and would be expected to intervene
against Effie. The behaviour remained at all times ambiguous and liable to
be misinterpreted - by anyone, that is, except Effie who could hardly have
failed to realize what we call the 'true' motive (see Good, Chapter 6, this
volume, for consideration of related cases where the 'future retrospective'
interpretation is crucial).
If Quince had been an inarticulate and working-class adolescent,
approaching a well-to-do lady's baby in an expensive pram but 'checking
her shopping list' throughout, a similar interpretation of intended
ambiguity would be uncontroversial. The girl could have been asked her
motive for continual re-checking of the same list, of course; but would she
have been honest in her reply, or even (since I described her as
inarticulate) able to explain? In general we find it unproblematic that
people cannot or will not explain what we attribute to be their 'real'
motives (see also Drew, Chapter 5, this volume, for similar dilemmas of
interpretation). Instead, gaining predictive power over their future
50 Richard W. Byrne

behaviour is our everyday criterion of understanding. (If the well-to-do

lady picked up her baby, smiled at the teenager, and offered her a chance
to hold the child, we might predict that the girl would not return to list-
checking afterwards.)
This is the same criterion that any behavioural scientist would adopt,
and the issue is simply one of whether such intentional accounts make
more wrong predictions than other theories do. Since gorillas' non-verbal
communication is less subtle than humans', I suspect that we will make
fewer errors of prediction for gorillas than for humans. A radical
behaviourist would take a different line, on principle rejecting intentional
explanations for humans' and animals' actions even-handedly, treating all
as 'operants' under stimulus control, which is at least fair. But this
approach has yet to explain which stimuli might have controlled Kim
Philby's everyday behaviour, for instance.

Evolutionary implications
The implications of this chapter are as follows. The earliest primates were
no more intelligent or socially sophisticated than other mammals. How-
ever, at some point between 55 and 25 million years ago (or 90 and 25
million years at the other extreme of estimation) the monkey and ape line
of descent was exposed to a strong selection of pressure (presumably a
need for greater social or Machiavellian aptitude). It responded with a
much greater ability at learning and using social knowledge, and conse-
quently an increased brain size - despite the energetic costs of the
metabolically expensive and fragile brain. This probably occurred early
rather than late in the time-band, since most monkey groups show at least
some tactical deception and all are large-brained.
One branch of their descendants, the apes, acquired the remarkable
ability to imagine alternative possible worlds, for instance the knowledge
and intentions of another individual. Although it is obvious that this
ability vastly increases the scope for social manipulation, it is quite
unclear what selection pressure could have promoted it. The problem is
just why such a useful trick should have evolved in apes but not monkeys,
who appear to have equal need of it. It is likely that the basic ability was
present in apes before 10-20 million years ago. By 5-8 million years, at the
time of the last common ancestor of human and chimpanzee, that
ancestor's ability to attribute intentions to others and imagine alternative
worlds would have been sufficiently sophisticated to permit success at all
the purely behavioural tests of it that psychologists have yet been able to
devise.
In the full sense of the term (Goody, Introduction to this volume),
these animals were capable of 'anticipatory interactive planning'. What
 The ape legacy 51

they lacked was language and its consequences. As far as we know, only
the Homo line of their descendants made this final step, but it may be
argued that anticipatory interactive planning was a necessary precursor to
it. Would any species unable to take account of an interlocutor's current
intentions and needs be capable of benefiting from linguistic communica-
tion, anyway? And from an understanding of another's intentions it is
perhaps a small step to that of understanding true communication, in
which a speaker takes account of the hearer's belief that the speaker
intends to pass information to the hearer. Modern apes may completely
lack the formalizing systems of language, but they do not lack the
understanding of what this kind of communication is all about. From this
perspective, we can predict that when the heated arguments over 'ape
signing' experiments havefinallydied down, it will be acknowledged that
apes understand and use true communication, even if they never double-
embed a relative clause.

The idea of Machiavellian intelligence

Does this mean that the social or Machiavellian intelligence hypothesis is
to be accepted? The problem with answering this question is that 'social
intelligence' is not just a little vague, it is multiply ambiguous. At least
three senses (and thus answers) can be distinguished. Firstly, there is the
hypothesis that an overriding selection pressure for the evolution of
primate and human intelligence was the need to live in societies and yet
maximize individual benefits, and that subtle manipulation of conspecific
group members was the best way to achieve this. The finding of
increasing intelligence in the social domain as one looks from strepsirhine
to monkey to ape, is consistent with this idea (and indeed predated it), but
does not prove it. The current popularity of this idea owes more to the
paucity of well-developed alternatives - never a strong basis for faith -
and it is anyway unlikely that a single factor was crucial in selecting
intelligence over a span of 55 million years. The energetic constraints on
large brains, for instance, cannot be ignored: diet must be a part of a full
explanation (see Milton 1988).
Secondly, there is the claim that primate intelligence is more
'advanced' in social domains, that the social factor of intelligence has
advanced ahead of any other. In one sense this is simply redescribing the
data. Monkeys sensitive to demographic changes in other troops of
conspecifics appear not to notice bizarre behaviour by hippos and
waterbirds, as simulated by playbacks, or even obvious signs of the
presence of predators (Cheney and Seyfarth 1988), and children demon-
strate an understanding of concepts like seriation earlier in social inter-
action than they can be shown to possess them in conventional tests
52 Richard W. Byrne

Smith 1988). In another sense the claim is bizarre and untestable. We are
not all-knowing beings able to catalogue the facets of intelligence in the
absolute; we can only notice how monkeys or children deviate from our
own, adult norm. Not only is there no metric on which to compare
intelligences, but we know that all supposed factors of human intelligence
correlate with each other.

The legacy
The third meaning of the Machiavellian intelligence hypothesis, and the
one of most interest for this volume, is the idea that all descendants of the
first anticipatory interactive planners carry a distinctive legacy. Human
intelligence includes abilities which are certainly not unique to language-
using humans. The monkeys and apes, like their human relatives, have
considerable social knowledge and an ability to plan effectively in
complex and socially subtle situations, in order to achieve their ends. And
modern apes have the ability to imagine the intentions, needs and beliefs
of other individuals, and to understand how these differ from their own.
Certainly we still rely on these skills (see Brown, Chapter 7; and Streeck,
Chapter 4, in this volume, for impressive examples). But perhaps human
intelligence is, as a result of its evolutionary origin, biased towards dealing
with all problems as if they were social problems. It is evident that in
everyday life humans are poor at predicting the lawful, probabilistic
results of random processes. Is this because they insist on treating all
processes as intentionally caused? (See Levinson, Chapter 11 in this
volume.) If so, then could this explain the odd tendency to reason with
and see communicative meaning in things which are wholly unpredict-
able and unaffected by human actions? (See Goody, Chapter 10; and
Zeitlyn, Chapter 9 in this volume.) Suggestions like these have been made
before (Humphrey 1976), and it has long been a matter of amusement that
people treat inanimate objects such as cars, computers and houseplants as
interlocutors in a dialogue. Now that we understand a little more of the
background to this peculiar trait, it is time to take the idea as a serious
subject of study.
 EDWIN HUTCHINS AND BRIAN HAZLEHURST

How to invent a shared lexicon: the

emergence of shared form-meaning mappings
in interaction

Recently, we have been exploring a new approach to cognitive anthro-

pology. We've been trying to push the boundaries of a genuinely cognitive
unit of analysis out beyond the skin of the individual. Ever since symbolic
and cognitive anthropology embarked on their ideational odyssey in the
1950s, they have proceeded without the material and the social. Of course,
many people are interested in social cognition where the social world is the
content of cognition. And in fact, there are good arguments for believing
that human intelligence developed in the context of reasoning about social
situations (Byrne and Whiten 1988; Byrne, Chapter 1, this volume;
Levinson, Chapter 11, this volume). This kind of relationship between
the social and the cognitive is important, but it is still centred on the
notion of the individual as the primary unit of cognitive analysis. The
social world is taken to be a set of circumstances 'outside' the individual
about which the individual reasons.
What we intend instead is to put the social and the cognitive on an equal
theoretical footing by taking a community of minds as the unit of analysis.
This permits us to do three things that are not possible from the
traditional perspective. First, we can inquire about the role of social
organization in the cognitive architecture of the system and may describe
the cognitive consequences of social organization at the level of the
community (Hutchins 1991; in press). Second, we can treat symbolic
phenomena that are outside the individual as real components of the
cognitive unit of analysis. This we take to be the position pioneered by
Sperber (1985). Third, we can account for the emergence of shared
symbols or form-meaning mappings; something that cannot be explained
by reference to the processes operating in individual minds alone.
Making this move also presents an opportunity to view language in a
new way. Cognitive science generally takes the existence of language as a
given and concerns itself with the sorts of cognitive processes that must be
involved when an individual processes language; in production and
comprehension. From the perspective of the community of minds
as cognitive system, language - its information-bearing capacity, its

53
54 Edwin Hutchins and Brian Hazlehurst

structural properties, the conventions of its use, etc. - becomes the

determinant of the cognitive properties of the community because it in
part determines where and when different kinds of information move
through the system. This attention to the movement of information in the
larger system necessarily brings the social and material back into play
since, having acknowledged symbols outside the head, we now must take
seriously their material nature and their use for organizing behaviour in
social integration (cf. J. Goody 1977).
The existence of shared language is one of the central facts of human
existence. Language appears to be closely tied to most high-level cogni-
tive activities. It mediates most of the social interactions among the
members of the most social of all species. Once a language exists, it is not
difficult to think of means by which it could be maintained and propa-
gated from generation to generation in a population. But without anyone
to tell individuals which language to speak, how could a language ever
arise? How could something structured come from that which is unstruc-
tured? It's a puzzle.
Here we are not thinking only, or even primarily, of the historical
origins of human language. There is, of course, a vast speculative
literature on the origins of language which we will not attempt to treat
here. Rather, we are thinking more modestly in terms of the development
of sets of local lexical distinctions such as may arise in small groups
engaged in shared tasks. In this chapter we outline a scheme by which a
shared lexicon emerges, the material components of one kind of form-
meaning mapping. Shared form-meaning pairs, which we will call
symbols, emerge from the interactions among members of a community.
The set of public forms used in interaction by the members of this
community constitutes the lexicon.
This is certainly not a model of the development of a human language,
but it does demonstrate how simple shared form-meaning mappings can
arise where none existed before. In the presentation below we will refer to
the forms as though they are words, terms, descriptions, or patterns of
acoustic features. There is, however, no strong commitment to any
particular level of linguistic representation here and the structures
described might just as well be thought of as patterns of denotational or
even relational features.
The model is based on six central theoretical assumptions.
1. No mind can influence another except via mediating structure.
(The no telepathy assumption.)
2. No social mind can become appropriately organized except via
interaction with the products of the organization of other minds,
and the shared physical environment. (The cultural grounding of
intelligence assumption.)
 How to invent a shared lexicon 55

3. The nature of mental representations cannot simply be assumed,

it must be explained. (The 'shallow symbols' assumption - in
contrast with the 'deep symbols' assumption which brings
symbols into the language of thought as an article of faith rather
than as a consequence of cultural process.)
4. Symbols always have both a material and an ideal component.
The material component is what makes forms, structure, and
difference possible. The ideal component is a function of the
stance that organized individuals take toward these material
forms. (The material symbols assumption.)
5. Cognition can be described as the propagation of representa-
tional states across representational media that may be internal to
or external to individual minds. (The distributed information-
processing assumption.)
6. The processes that account for the normal operation of the
cognitive system should also account for its development
through time. (The no developmental magic assumption.)
Below we present a computer simulation that is an implementation of
these assumptions. It turns out to be a very robust procedure by which a
community of individuals can develop a shared set of symbols. It is the
simplest possible scheme that captures the essential properties of the
system being modelled. The models are much too simple to be taken as
representations of human cognition. They simply demonstrate that a
particular kind of process is capable of producing a particular sort of
outcome, in this case, a community with a shared lexicon.

The constraints on a shared lexicon

The central problems of inventing a lexicon can be stated in terms of a
description of the outcome. Consider two individuals, A and B, and a set
of phenomena in the world numbered 1, 2, 3 . . . m. Let the word that an
individual uses for a phenomenon be denoted by the concatenation of the
letter designating the individual and the number of the phenomenon (i.e.
'B5' denotes the word that B uses for the fifth phenomenon). Now, if the
lexicon is to be shared, the word that A uses for any particular phenome-
non must be the same as that used by B. In notation: Al = Bl, A2 = B2,...
Am = Bm. Simultaneously, if the lexicon is to be a lexicon at all, there must
be differences between the material forms of the words used by each
individual for different phenomena. In notation: Al=/=A2=/= . . . Am;
Bl =1=32=1= . . . Bm. It won't do to have a lexicon for m phenomena that
consists of m homonyms.1 These two constraints must somehow be
simultaneously satisfied in any process that is to develop a shared lexicon.
For our purposes, a shared lexicon is a consensus on a set of distinctions.
56 Edwin Hutchins and Brian Hazlehurst

Natural
Internal
Artifactual
Natural
Internal
Artifactual
Natural
Internal
Artifactual
Natural
Internal
Artifactual

Figure 2.1 The relations of environment, internal and artifactual struc-

ture. The arrows represent the propagation of constraints. Constraints
may be propagated by many means. We use the cover term 'coordina-
tion' to refer to the satisfaction of constraints no matter the mechanism
by which constraint satisfaction is achieved.

Before turning to the simulation, we need to say a few more words

about the theoretical stance. The six theoretical assumptions described
above can be assembled into a model of socially distributed cognition as
shown in Figure 2.1. Our inventory of representational structure includes
natural structure in the environment, internal structure in the individuals,
and artifactual structure in the environment. Artifactual structure is a
bridge between internal structures. Artifacts may provide the link
between internal structures in one individual and those in another
individual (as is the case in communication), or between one set of
internal structures in an individual and another set of internal structures
in that same individual (as is the case in using written records as a
memory, for example). Internal structures provide bridges both between
successive artifactual structures and between natural and artifactual
structures. Following Sperber (1985:76) we may say that 'A represen-
tation [artifactual structure] is of something [natural structure] for some
information processing device [internal structure].'
In the past five years, developments in computational modelling have
made it possible to think in different ways about the relations between
structure inside a system and structure outside. Connectionist networks
of a class called 'autoassociators' have particularly nice properties with
respect to the problem of discovering and encoding structural regularities
in their environment.2
Autoassociator networks learn to duplicate on the output layer the
pattern of activation presented to the input layer. Figure 2.2 shows a
simple autoassociator network. It consists of three 'layers' of units. Input
 How to invent a shared lexicon 57

1 0

Inputs
0 0

Input
Layer

Output
Layer

1 0
Targets

0 0

Figure 2.2 A typical autoassociation network and learning scheme. See

text for explanation of network components.

units on the left, output units on the right, and 'hidden' units in the
middle. 'Targets' are real valued vectors which are structurally similar to
the output and input layers but, like inputs, are thought of as information
external to the network - these are part of the environment which the
network is made to learn (see below).
Limitations on space make a full description of this kind of infor-
mation-processing system impossible. The following sentences will
hopefully convey the style of computation entailed, if not the details.
Every unit in the input layer of a network has a unique connection to
every unit in the hidden layer, and every unit in the hidden layer has a
unique connection to every unit in the output layer (see Figure 2.2). The
58 Edwin Hutchins and Brian Hazlehurst

strengths of these connections can be adjusted. The activations of the

input layer are set by external phenomena. The activations of the other
units are determined by the activations of the units from which they have
connections and on the strengths of those connections. The task for the
network is, starting from random connection strengths, to discover a
pattern of connection strengths that will produce the desired output in
response to a given set of inputs. Incremental improvement in accom-
plishing this task is referred to as 'learning'. The networks modelled here
use a procedure called the 'back-propagation of error' tofindan appropri-
ate set of connection strengths. In this scheme, the output produced is
compared to the target3 and the difference between output and target is an
error in the network's ability to perform this input-output mapping. The
connections are then adjusted to reduce this error on future trials at this
task. The problem for the network can be viewed as one offindinga set of
connection strengths which simultaneously meets the constraints
imposed by all of the input-output mappings it is made to perform.
Rumelhart, Hinton and Williams (1986) have shown that under certain
conditions the activations of the hidden layer units of fully trained
autoassociator networks converge on efficient encodings of the structural
regularities of the input data set. That is, the connections between input
and hidden units must produce activations at the hidden layer which can
be used by the connections between hidden and output units to produce
the target, under the constraints of the function which propagates
activation. For example, given any four orthogonal input patterns and an
autoassociator network with two hidden units, the hidden unit activations
for the four input cases should converge on {(0 0) (0 1) (1 0) (1 1)}. This is
because the network must use the activations of the hidden units to
encode the four cases and the encoding scheme attempts to distinguish
(optimally) among the cases. Producing these efficient encodings is
equivalent to feature extraction. That is, the networks learn how to
classify the input data in terms of distinctive features or principal
components.
Imagine an autoassociator folded in half. The hidden units still
produce efficient encodings of the input data set. Since the patterns of
activation on the hidden units encode the features of the input set, they
can be seen as representations of the world of experience. If we make these
representations 'public', so that they become a part of the shared material
world of interaction, then they serve as this individual's descriptions of
the input data.4 These public hidden layer encodings produce one of the
properties we want in a lexicon; distinctions among representational
forms. In Figure 2.3 we have relabelled these units 'verbal input/output'
units.
If we take the remaining parts of the network to be a simple visual
system - capable of classifying scenes in the environment - then the verbal
 How to invent a shared lexicon 59

Verbal
Visual Input/
Input Output
Layer Layer

Visual
Output
Layer

Figure 2.3 A modified autoassociator, with 'public hidden units'. By

'folding' an autoassociator back on itself, we create a system capable of
generating referentially meaningful (i.e. distinct) representations of the
m scenes.

input/output layer is capable of generating patterns of activation in

response to each visual scene encountered, and these patterns are (or
become) maximally different from each other. Regarding network A's
descriptions for the m scenes, this satisfies the constraints: Al f= A2 f= . . .
Am.
Virtually all work in connectionist modelling today models aspects of
the cognition of individuals. Our theoretical stance suggests that it might
be useful to consider the properties of communities of networks. Of
particular interest here is the fact that in traditional connectionist
modelling, the programmer constructs the world of experience from
which the networks learn. In a community of networks the behaviour of
other networks might also be an important source of structure from which
each network could learn. Connectionist programmers refer to the output
patterns to be learned as the 'teachers' for their networks. With a
community of networks, we can let an important part of the teaching be
embodied in the behaviour of other networks. Thus, where traditional
network modelling is concerned only with the relation of structure in the
environment to internal structure, a model of interactions in a community
of networks adds the universe of communicational artifacts to the picture.
It is easy to show that consensus among two networks (say A and B) can
be achieved by taking the output of each as the teacher for the other. This
satisfies the constraints that Al = Bl, A2 = B2, . . . Am = Bra.

Implementation
The simulation proceeds via interactions - one interaction is one time-
step in the simulation. An interaction consists of the presentation of a
chosen scene (from the set of ra scenes) to two chosen individuals, a
60 Edwin Hutchins and Brian Hazlehurst

Figure 2.4 A scheme for evolving consensus on a set of distinctions. By

reciprocally constraining autoassociators at their hidden layers, consen-
sus about the representations used to classify the m scenes can be
achieved.

'speaker' and a 'listener' (from the set of n individuals). The functions

which do this 'choosing' determine what we call the 'interaction protocol'
of the simulation. The typical functions simply implement random
selection from the domains of scenes and individuals, respectively. One of
the individuals chosen (say A) responds to the scene by producing a
pattern of activation on its verbal output layer (A 'speaks'). The other
individual (say B) also generates a representation of what it would say in
this context but, as 'listener', uses what A said as a target to correct its own
verbal representation. The listener (B) is also engaged in a standard
learning trial on the current scene, which means its own verbal represen-
tation - in addition to being a token for comparison with A's verbal
representation - is also being used to produce a visual output by feeding
activation forward to the visual output layer. The effects of this learning
on B's future behaviour can be stated as: (1) in this context produce a
 How to invent a shared lexicon 61

Hidden
Layer
(4 units)

Input Layer
(36 units)

Verbal
Input/
Output
Layer
(4 units)
Output Layer
(36 units)

Figure 2.5 Network architecture for the simulation. (Not all of the
connections between layers are shown.)

representation at verbal output more like what A said, and (2) produce a
representation at visual output more like the scene itself.
By randomly choosing interactants and scenes, over time every indivi-
dual has the opportunity to interact with all the others in both speaking
and listening roles in all visual contexts. The effect to be achieved is for
the population to converge on a shared set of patterns of activation on the
verbal output units that makes distinctions among the m scenes. That is,
we hope to see the development of a consensus on a set of distinctions.
The small network architecture described above is a simplification of
the architecture actually used in the simulation reported below. In this
simulation, each individual is an autoassociator network consisting of
thirty-six visual input units, four hidden units, four verbal output units
and thirty-six visual output units, as shown in Figure 2.5. Notice that an
additional layer of four hidden units appears in these networks. These
additional resources were required by networks in this simulation in
order for the community to converge on a shared lexicon.5 The scenes to
be classified are twelve phases of a moon, represented as patterns in the
6 x 6 arrays shown in Figure 2.6.

Results
Developing consensus on a set of distinctions appears to be a highly likely
final stable state of this dynamical system. Since the initial connection
strengths of individuals are small and randomly assigned, early verbal
representations do not differentiate among the scenes represented. Figure
62 Edwin Hutchins and Brian Hazlehurst

FULL

i U

LAST QUARTER \n NEW

Figure 2.6 The visual scenes utilized in the simulation. These can be
thought of as representations of the visualfieldassociated with sight of
the moon in twelve different phases.

2.7 shows the activation levels of the four verbal output units in response
to the twelve scenes for some typical individuals, early in a simulation run.
It is easy to see that there is little variation in the response of any
individual to the different scenes.
Figure 2.8 shows the same individuals after an average of 2,000
interactions with each of the other individuals in the five-member
community. For the most part, individuals now respond differently to
each of the twelve scenes, and all of the individuals agree with each other
on how to respond. That is, we have consensus on a set of distinctions.
Due to the random starting weights of the networks, and the random
interaction protocol functions which organize their learning experiences,
there is no way to predict which lexicon will develop - but the procedure is
robust in the sense that some well-formed lexicon or another develops
nearly every time.

The need for a critical period of language learning

We have also experimented with adding new individuals with random-
ized nets ('babies') to communities that have already developed a lexicon.
Depending on the size of the community, the addition of the new
individual may have quite different effects. A new individual added to a
large community with a highly shared lexicon will be entrained by the
community and will learn the shared lexicon. A new individual added to a
 How to invent a shared lexicon 63

Figure 2.7 Four individuals of afive-membercommunity at the start of

a simulation run. The surface represents the value of each verbal output
unit in response to each moon scene.

smaller community may completely destroy the previously achieved

solution. After such an event the community may or may not be able to
relearn a lexicon with the new individual.
In running these simulations, we found ourselves wishing that there
was some principled way to reduce the learning rate once individuals had
learned the language. In particular, one would like to reduce the learning
rate at the point where individuals are likely to encounter many interac-
tions with disorganized individuals. This would amount to implementing
a critical period for language learning so that individuals learn less from
the linguistic behaviour of others once they have reached sexual maturity.
Perhaps evolution has engineered something like this into our species. We
did not implement a critical period, however, because to do so seemed
arbitrary and a violation of one of the core premises: that the processes
that account for normal operation of the system should also account for its
development through time. In more complex situations, like that of
biological evolution where adaptive searches are conducted in parallel at
many levels of specification, it may be reasonable to expect violations of
this premise.
64 Edwin Hutchins and Brian Hazlehurst

Figure 2.8 Four individuals of a five-member community after 50,000

interactions. Each individual has had (on average) 2,000 interactions
with each of the other four individuals. Half of these were in the role of
'listener' and half in the role of 'speaker'. The surface represents the
value of each verbal output unit in response to each moon scene.

Discussion
The model explicitly represents the interactions of the three kinds of
structure discussed earlier: natural, internal and artifactual. The patterns
representing phases of the moon are the 'natural' structure. The connec-
tion strengths in the networks are the internal structure that provides
coordination between the two kinds of external structure. The patterns of
activation on the verbal input/output units are the 'artifactual' structure.
We see this as the smallest first step toward a system in which artifactual
structures invoke the experience of that which is not present in the
environment.
As we have seen, no individual can influence the internal processing of
another except by putting mediating artifactual structure in the environ-
ment of the other. However, by putting particular kinds of structure in
each other's environments, they all achieve a useful internal organization.
 How to invent a shared lexicon 65

It would be possible for each individual to achieve an internal classifica-

tion scheme in isolation - after all, that is what autoassociators are known
to do by themselves. But such a classification would be useless in
interaction with others. That is, idiosyncratic distinctions may be useful,
but not as useful as shared ones. By forcing the individuals to learn from
the classification behaviour of others we ensure that each individual can
only become internally organized by interacting with the external pro-
ducts of the internal organization of others. The effects of this kind of
system enable individuals to tap the resources of an entire group (and
ancestors of the group), enabling cognitive performance not achievable
by individuals alone. This is the foundation upon which human intelli-
gence is built (Hutchins and Hazlehurst 1991).
Although this simulation is too simple to claim to address the issue of
symbolic thought processes directly, it suggests a way in which shared
symbols that could subsequently come to serve internal functions could
arise as a consequence of social interaction. Such symbols are outside the
individualfirstas pieces of organized material structure - in the behaviour
of others - before they have explicit internal representations. Undoubt-
edly, such shared public forms can be given internal representations, as
can any structural regularity in the environment whether natural or
artifactual. This perspective in which symbols are in the world first, and
only represented internally as a consequence of interaction with their
physical form is what we mean by the 'shallow symbols' hypothesis. In
this view, symbols and symbolic processing may be relatively shallow
cognitive phenomena, residing near the surface of functional organiza-
tions resulting from interaction with material structures.
The computations performed by the networks are well characterized
by the propagation of the representational state. The universe of inputs is
propagated through the networks and re-represented at the output. This
general notion of computation comes from Simon (1981:153) who says,
'Solving a problem simply means representing it so as to make the
solution transparent.' Simon may not have intended quite so broad a
reading of his definition but it seems to capture well the behaviour of this
system. The structure of the natural world isfixedin this model, but the
internal structures and the artifactual structures co-determine each other
and co-evolve in the development of the lexicon. In the broadest sense,
the solution arrived at was determined by the structure of the natural
world as manifested in the phenomena encountered and in the random
initial configurations of the internal states of the individuals in the
community. The process of developing a lexicon in this model is a process
of propagating transformed representations of naturally occurring struc-
ture throughout a system that contains artificial structure as well. At the
outset, the structures of the 'minds' of the individuals are 'natural',
66 Edwin Hutchins and Brian Hazlehurst

having been given by nature. At the end of the simulation, these same
structures are 'artificial' in the sense that they are organized by structure
created by other individuals.
Finally, even when two networks are in complete agreement with each
other about the structure of the lexicon, each has a unique internal
structure. Through learning from each other the individuals become
functional equivalents, not structural replicates, of each other. That is,
constraints on public form do not uniquely specify internal structure so
long as shared form-meaning pairs are established.
Clearly, this generation of shared form-meaning pairs is not limited to
acoustic forms of symbolic behaviour - what we have characterized in this
simulation as words of a lexicon. Words are a particularly salient vehicle
for symbolic processes, as they are representations which are easily
recorded and objectively characterized. What of the meanings we attach
to other social behaviours? Or the meaningful activities of the incumbent
of a recognized social role? The same sorts of processes are certainly at
work in establishing the sharedness and consistency of these forms and
meanings as well (cf. Barth 1966; Goffman 1959). Roles and the meanings
of those roles - including norms which legitimate and motivate constel-
lations of social behaviours - may come into being together as the result of
constraints on the building and maintenance of consensus through social
interactions. Although the current simulation does not address this level
of complexity of social interaction, it does suggest possible candidates for
the foundational mechanisms involved.

Acknowledgement
Research support was provided by grant NCC 2-1591 to Donald Norman and
Edwin Hutchins from the Ames Research Center of the National Aeronautics and
Space Administration in the Aviation Safety/Automation Program. Everett
Palmer served as technical monitor. Additional support for the first author was
provided by a fellowship from the John D. and Catherine T. Mac Arthur
Foundation.

Notes
1 This model does not deal with either homonyms or synonyms.
2 The best background work in connectionism is the two-volume set Parallel
Distributed Processing by Rumelhart and McClelland (1986). The behaviour of
autoassociator networks is thoroughly analysed in Chauvin (1988).
3 For an autoassociator, the target is identical to the input, thus reducing the
problem to an identity mapping on the input set.
4 We thank Elizabeth Bates (personal communication, February 1991) for
coining the term 'public hidden units' for this construction.
 How to invent a shared lexicon 67

It is a well-known result from connectionist history that two layers of weights

are required to perform nonlinear mappings from input to output (Rumelhart,
Hinton and Williams 1986). The range of verbal representations that indivi-
duals are attempting to map in this simulation apparently constitute a nonlinear
set, which requires the extra hidden layer to perform properly. Another reason
for the required extra layer has to do with the large compression of data from
input to verbal output layers. This compression tends to 'swamp' the verbal
output layer with large values reducing the amount of useful information here,
and limiting the usefulness of these values as targets for other individuals.
 NURIT BIRD-DAVID

Hunter-gatherers9 kinship organization:

implicit roles and rules

This chapter offers an analysis of the emergence of explicitly institution-

alized roles and rules based on kinship (Kinship roles and rules) among
Nayaka.1 They provide a case of 'hunter-gatherers',2 with the simplest
known social organization called band-organization. While analysis of
this kind of organization has been traditionally informed by an ecological-
evolutionary paradigm, the present analysis is informed by Berger and
Luckmann's The Social Construction of Reality (1966). The analysis was
inspired by Goody's AIP project, which calls for examining human
practices in relation to evolutionary programmed anticipatory interac-
tive planning (AIP). The analysis aims to contribute to the project, in
reciprocity, a perspective on hunter-gatherers that can raise hypotheses
about the possible contribution of AIP to the emergence of human roles
and rules.
I shall first outline the conceptual links that lead from the AIP idea to
the analysis. Then I shall examine the Nayaka social scene, and their
kinship system (terms, cooperation, and avoidance). Finally, I shall offer
a comparative perspective, and reflect back on AIP concerns.

Theoretical links
The AIP idea started from primate research. Challenging the traditional
view that goes back to Darwin and Wallace, students of high primates
recently suggested that the principal selective pressure behind the
evolution of 'higher intellectual faculties' lay not in technical dealings
with the physical environment but, instead, in everyday dealings with
social associates (Byrne and Whiten 1988; Humphrey 1976; Jolly 1966a).
They argued that living within social groups, early hominids had - like
chess players - to foresee others' actions in order to preempt or exploit
these actions to their own advantage. Furthermore, they had to adjust
their plans to the actual deeds of others, who equally planned their actions
strategically. This complex situation generated selective pressure for
intellectual prowess.
As Goody recognized, pursuing further this primate-based evolution-

68
 Kinship organization 69

ary argument, it follows that anticipatory interactive planning is evolu-

tionarily programmed into human intelligence. If so, it should have
bearings upon cultural practices and institutions and be expressed in
them. Goody explored this hypothesis at great depth in her work on
questions, politeness and praying (1978a,b; Chapter 10 in this volume).
She briefly noted other areas in which it could be explored - among them,
the emergence of roles and rules.
It would have been best to follow this last-mentioned avenue in relation
to hunter-gatherers: seen as ethnographic examples of the simplest
human social organization, or - much more problematically - as a source
of insight on evolutionary processes. However, hunter-gatherers have
been traditionally viewed within a (naturalistic) ecological-evolutionary
paradigm. Conventionally, this paradigm views their institutions in
relation to their dealings with the physical environment (rather than their
dealings with each other); in terms of functional subsistence needs (rather
than inter-personal interaction and communication). Therefore, the
existing ethnography would not have done for the purpose.
A fresh perspective on hunter-gatherers is required in order to pursue
the AIP hypothesis - a perspective informed by a paradigm that does view
institutions in relation to inter-personal interaction and communication.
Berger and Luckmann's The Social Construction of Reality (1966) is an
obvious choice. Its first chapter, based on Alfred Schutz and Thomas
Luckmann's work (1973) - separate from the rest of the book and less well
known- is particularly useful. It deals with the 'prototypical case of social
interaction' that takes place in the 'face-to-face situation'. In this situa-
tion, it is argued, individuals know each other 'vividly' and as 'fully real'.
Their interaction is predominantly directed towards the concreteness of
the other. As a result, typificatory schemes are more 'vulnerable' to
personal interferences than they are in 'remoter' forms of interaction
where there is a greater amount of anonymity. Relationships are highly
flexible. It is comparatively difficult to impose rigid patterns on them.
The dialectic development of institutions is held at bay (Berger and
Luckmann 1966:28-34).
Phenomenologically complex, this argument can be read more simply
and then brought to bear on what anthropologists commonly called 'band
organization'. The latter involves living in very small groups - twenty-
five individuals on average - under material conditions that preclude
privacy. However, 'face-to-face' has been traditionally used by anthro-
pologists to describe situations wherein individuals interact with each
other in diverse roles (almost the opposite of Berger and Luckmann's
'face-to-face situation'). Moreover, under Goffman's influence, 'face' has
come to be associated with a representational mask (again, almost the
opposite of Berger and Luckmann's 'face-to-face'). Therefore, instead of
70 Nurit Bird-David

'face-to-face', yet paraphrasing it, I use 'person-to-person'. This termi-

nological change also signals a departure from the full phenomenological
connotations of the concept (discussed at length by Schutz: see 1970).
To conclude, The Social Construction of Reality's perspective,
especially the discussion of the pre-institutionalized 'person-to-person'
situation, provides a framework within which to analyse hunter-gatherers
in a way which will make it possible to use their important case for
exploring AlP-related ideas, especially with reference to the emergence
of roles and rules.

A case study
This perspective will be brought to bear on kinship among hunter-
gatherers with 'immediate-return systems' (Woodburn 1980, 1982).
These are hunter-gatherers with the simplest known social organization,
comprising among others, the !Kung of the Kalahari, the Hadza of
Tanzania, the Pygmy of the Ituri forest in the Congo, the Negrito Batek of
Malaysia, and the Hill Pandaram, Paliyan and Nayaka of South India. In
general, they have few and simple roles and rules (Woodburn 1980).
Division of labour, if at all, is by sex and age alone. As for their Kinship -
on which the analysis will focus, not least because its institutionalization
is argued to have been crucial in the genesis of human society (e.g., Levi-
Strauss 1969; Fortes 1983) - Woodburn has already argued that it
'regulates so little' (1979:257). However, he left many questions open -
why? how? can it really be so? - questions which could not have been
explored within the conventional ecological-evolutionary paradigm. My
argument will be that this state of kinship has to do with the 'person-to-
person' situation which dominates their social life; a situation which
hinders the dialectic development of Kinship roles and rules.
The analysis refers specifically to Nayaka. They inhabit tropical forests
in the north-western slopes of the Nilgiris in South India. This is a
border-area between Tamil Nadu, Kerala and Karnataka, and their
composite dialect reflects the three languages spoken there: Kanada
(being traditionally dominant); Malayalam (rapidly increasing its
influence); and Tamil (having a steady minor impact).
The total population of Nayaka is estimated at about 1,000. In practice,
they are distributed in enclave local communities, each of which is almost
autonomous. My own work was done among one local community -
whom I call the Gir Valley group. I stayed with them during 1978-9, and
re-visited them in 1989.3 They numbered (in 1978) 69 individuals (22
men, 24 women, and 23 children), who were further dispersed in five
residential clusters ('hamlets'), at a distance of two to ten kilometres from
each other, one tofivefamilies living in each. For a living, they gathered
 Kinship organization 71

wild fruits, nuts, tubers and the honey of wild bees; they fished and
hunted occasionally (mainly deer and monitor lizard); they engaged on a
casual basis in wage work and in trade in minor forest produce; and
occasionally grew some fruit around their huts.

The social scene

What is social life like in a band organization? One's imagination is
frequently coloured by one's own conditions of sociality (in itself attesting
to Goody's AIP hypothesis!). Therefore, westerners often find it difficult
to imagine. To my students I often suggest that they imagine themselves
spending the twenty-four hours of the day, each day, from birth to death,
almost exclusively, with the neighbours (who are relatives as well) in a
sixteen-flat building, without the material partitions which provide
privacy. Here, I rather sketch the daily routines of an 'average' Nayaka
(arbitrarily female living with a man and young children).
She lives in a hut constructed in a small forest clearing surrounded by a
thick tropical jungle on the sloping side of the Gir valley. The hut stands
next to several other huts (ranging from one tofive)at a distance of two to
five metres from them. She occupies in this hut - with her family - a small
living space (at most two by two metres); and the hut may also contain
other living spaces (at most two), each for another family. The internal
partitions, if any, are clumsily built, and often are crude signals of
partitions. The external walls are made of strips of bamboo, interwoven
either thickly - in which case they block out vision but not audition, or
thinly - in which case they bar neither. Under these conditions, she
overhears almost everything uttered in the hut, and much of what is said
in the hamlet, whether it is said to her or to others. She also sees a great
deal of what her neighbours do (in the hut and within the hamlet
generally).
At any rate, during the dry seasons (in this local monsoon climate) she
spends most of her day outside the hut, at leisure, eating or sleeping by the
side of an external fireplace. Her fireplace is only a few metres away - in
the open - from other fireplaces, where other people spend their time.
None of them pretends to be outside the sight and hearing of the others.
They often talk to each other across space (and in the hut, across
partitions), each remaining by their own fireplace (or in the hut, in their
own living space).
Our Nayaka bathes and draws water from a nearby stream, going
always to the same stretch of water together with the other Nayaka in her
hamlet - frequently during the same early and late hours of the day. She
goes on forays, normally with her family, but they are often in close
proximity to other families, since all follow the same spatial trajectory of
72 Nurit Bird-David

seasonal variations. Occasionally, she goes to work in the plantation, or

the local timber company, but there she is designated tasks commonly
given to Nayaka, and therefore works alongside her Nayaka neighbours.
Occasionally she goes on visits to other hamlets in the locality, staying
there for weeks or several months at a time; while others come for visits to
her hamlet. However, since there are only five hamlets in the close
locality, with fewer than fifty adults altogether, she gets to know them,
through repeated visits over the years, in almost the same immediate way
I have just described.
The local community itself is not totally cut off from other local Nayaka
communities and, occasionally, single young persons leave for, or come
from, other communities. The 'immigrants' settle in, often upon mar-
riage to a local person, and within a year or two our Nayaka get to know
them pretty well, in an immediate sort of way. Or they leave after a while
and, remaining outside the zone of everyday social life, they are rarely
mentioned, if at all. It seems that when they are out of everyday social
reach, they are also out of everyday concerns and expressed conscious-
ness. Likewise, the 'emigrants' either remain away, effectively abolished
from collective memory, or they return after a while (with or without a
spouse) and become re-absorbed into the intimately known local
community. At any rate, these young persons constitute a negligible
proportion of the local community, at any given point in time (for
example, only two came, and none left during 1978-9).
All in all, the Nayaka woman spends the greater part of her life in the
way I have just described, sharing the 'here and now' in all domains of life
with the same few adults who, at diverse degrees of relatedness, are also all
her relatives. She also grew up with many of them.
Thus, we can conclude4 that our Nayaka's (inter-Nayaka)5 interaction
is largely confined to people with whom she has ongoing relationships.
Furthermore, and this is important, she constantly witnesses their
ongoing interactions with each other. In other words, this is a highly
close-meshed intimate network. Moreover, she is familiar with their
routines in most domains of life, their idiosyncrasies, and most of what
has happened, and is happening, in their lives. In short, her principal and
regular field of (inter-Nayaka) social interaction comprises people who
she knows vividly, immediately, and in a manifold way. We can say that
her social world is predominated by 'the person-to-person' situation.
In this situation, how is kinship expressed and what does it do?
Through the two-stage detailed description which follows, I want to
show that while people classify kin and use kinship terms extensively in
everyday communicative acts, they neither formalize nor discuss Kinship
roles and rules. They do not expect distinct categories of kin (spouse
excluded) to behave in any distinct way.
 Kinship organization 73

Kinship and communication

Kinship terms are used extensively in everyday communicative acts. One
can constantly hear Nayaka using them interchangeably with personal
names. They use twenty-seven Dravidian kinship terms (most of which
come from Kanada, but a few are from Malayalam). They also use
twenty-seven names (commonly used in the region by people of other
ethnicities as well), twelve of which are clearly of recent origin (see Bird
1983a).
This intermingled use of kinship terms and personal names varies
according to the life-cycle. Young babies do not receive names at all. My
host explained it in a simple, if somewhat pragmatic, way. One woman,
for example, said to me: 'He [her baby] cannot respond to a name, so I do
not call him.'
Young children are then called either by the name Kungan (for a boy)
and Kungi (for a female), or by the kinship terms maga(n) (son) and
maga(l) (daughter). All of them are called by these same names or kinship
terms - by any adult, including the parents. It seems that, familiar with
individuals' voices and idiosyncratic routines in this 'person-to-person'
situation, Nayaka easily identify by context which of the children - few in
any case - is being referred to or hailed in any instance.
Adolescents, compared with all others, are frequently addressed and
referred to by distinguishing nicknames. These are names drawn from the
limited list of twenty-seven, with prefixes which relate to diverse attri-
butes, for instance, relative age (e.g., Cik Mathen; 'cik' means junior),
place of origin (e.g., Munderi Mathen; 'Munderi' is a place) and
biographic detail (e.g., Chikkari Mathen; 'chikkari' is a hunting guide).
The nicknames are frequently changed. Furthermore, any given indivi-
dual can be simultaneously called by diverse names. I remember asking
informants in 1978 to list Nayaka by name. To my surprise - and at the
time also suspicion - they hailed my subjects of inquiry and asked them
'how they were called now' then passed the replies on to me. During my
follow-up survey in 1989, some of them then asked me what they were
called in 1978. (Other local non-Nayaka - for example, employers and
traders who keep accounts of transactions with Nayaka for their own
needs - also stabilize the use of names in out-Nayaka communication).
The use of nicknames almost ceases in adult life, whereupon kinship
terms are used almost exclusively. Although the kinship terms are
Dravidian, Nayaka use them in a distinct way, sometimes in stark
contradiction with the logic of the Dravidian kinship system. They learn
which kinship terms to use - as the examples which follow illustrate -
within the context of everyday communicative acts. My hosts always
explained to me how they arrived at any particular kinship term in these
74 Nurit Bird-David

ways. Firstly, 'He (or she) [often a person from ascending generation]
calls me "x" so I call him (or her) "y": for instance, he calls me tamma(n)
[younger brother] so I call him anna(n) [older brother].' Secondly, 'My
so-and-so relative calls him (or her) "x" so I call him (or her) "y": for
instance, my appa(n) [father] calls this man tamma(n) [younger brother]
so I call him cikappa(n) [father's younger brother - literally, junior
father].' Finally, 'My spouse calls this woman "x" so I call her "y" (or
sometimes "x" as well): for instance, my spouse calls her tanga [younger
sister] so I call her nadini [younger sister in-law] - or sometimes tanga as
well.' Thus, they learn kinship terms by hearing older people using them
to address and refer to other present people. They learn them from
persons who directly address them by kinship term. And they learn them
from their spouses.
In the obverse, the width and depth of Nayaka genealogical knowledge
is confined to the zone of everyday person-to-person interaction and
communication ('the zone of bodily manipulation' in Berger and Luck-
mann's terms). My hosts, for example, could not specify genealogical
relatives beyond the third ascending generation. They said matter-of-
factly that 'they were not there'. Likewise, they could not specify
genealogical relatives of the second and third ascending generation when
the individuals concerned had died or emigrated when they themselves
were young. They said 'they were not here when I came to know things so
I did not know them'. Thus, they learned about their kin - who they were
and how they related to them - through interactive use of kinship terms in
place of personal names.
It is noteworthy that, in the first year or so, my hosts did not address or
refer to newly arrived Nayaka, who married someone locally, by kinship
terms. They could have reasoned out the appropriate kinship
through the ways specified above, but chose not to in favour of names. As
Mathi explained it, when referring to a young 'immigrant' who had
married her sister's daughter: T do not know how to call him. I am shy to
call him because I do not know him'. This suggests that they use kinship
terms only in reference to Nayaka who, in the course of time, they come to
know in a 'person-to-person', 'vivid' and 'fully real' sort of way. (And it is
then, in fact, that we can understand how a general kinship term - like
tamma(n) (younger brother) - being called across the hamlet, reaches its
exact addressee.)

Kinship and cooperation

Using twenty-seven kinship terms so extensively in everyday communi-
cation, the importance of Nayaka linguistic classification of kin cannot be
doubted. However, I shall now attempt to show that this linguistic
 Kinship organization 75

classification does not go with a behavioural classification. My hosts did

not talk about linguistically distinguished relatives in terms of what they
'should do' and 'do' in interaction with them. Conversations with them
about how one should behave with a particular category of relative, say
cikappa(n) (father's younger brother), or even about how they them-
selves behave towards a specific person who is a cikappa(n), always came
to a short end. It seems they do not associate clusters of rights and
obligations, or even kinds of habitualized cooperation, with linguistically
distinguished kin. But can this really be the case?
Even when the same is reported for other comparable groups (Wood-
burn 1979, 1980; see below), it can still be counter-argued that the
ethnographer (and her colleagues as well) simply failed to find out
everything there is to find. To fend off such criticism one can substantiate
the claim by further ethnographic observations - which I shall do next -
and, since even further data cannot conclusively make the point, one can
at least explain why roles and rules are not fully institutionalized in the
'person-to-person' situation - which I shall do subsequently.
Starting then with the former, I describe life-cycles and special events
which in most other societies, constitute a stage for demonstrating and
affirming Kinship roles. However, in this case they do not: it can be seen
from the following examples which are concerned, respectively, with
marriage, burial, the annual celebration, sharing of large game, disputes,
and the division of labour - all stages par excellence for Kinship
performances.

Marriage
Though couples often simply start to live together, they sometimes
celebrate it with a meal. The meal is rarely planned in advance, and is
given to people who happen to be around at the time. For example, in the
most elaborate case I have on record, a messenger went to another hamlet,
the evening before the meal, and returned the next day with a few
additional guests. The father of the bridegroom, a widower, was in
neither of these hamlets, and he did not attend the celebration.

Burial
Burial is carried out on the day of death. Those who happen to be in the
hamlet carry the body to a place, several walking-hours away from the
hamlet, where all Nayaka are buried. The spouse and the children, and
then other attendants, take what they want for remembrance, and the rest
of the possessions are buried with the deceased. A person, even a close
relative, who happens to be away, does not participate in the event, nor
'inherit' anything from the deceased.
76 Nurit Bird-David

The annual celebration

A celebration is held almost every year in each of the hamlets, staging
communitas with local ancestral and natural spirits. Lasting twenty-four
hours, it involves shamanistic conversations with the spirits, dances, the
playing of music, and a communal meal at the end of the day. The
preparation of the communal food is done by a resident couple - not a
group of relatives. Furthermore, when the food is dished out, each
nuclear family takes its share and sits to one side to eat it. Moreover, the
families spend most of the twenty-four-hour communal occasion stand-
ing by themselves at some distance from others. Only occasionally do
'married' men and women part to join single-sex dancing circles, kept
going by young 'unmarried' dancers.

Sharing large game

Large game, hunted by dogs and killed by knives, or rescued from the
claws of other forest predators, is distributed among all who happen to be
present in the hamlet. Everyone present during the distribution gets a
share, irrespective of particularistic kinship ties, or duration of stay in the
hamlet.

Disputes
Individuals do not intervene in the internal affairs of their close kin. For
example, in an exceptional case I have on record of a husband who
maltreated his wife (a very uncommon occurrence), her brothers did not
interfere, although they lived in the same hamlet.

Division of labour
Conjugal spouses (accompanied by their young children, if any) cooper-
ate with each other - not with respective relatives - in most subsistence
and domestic activities. They go on most subsistence pursuits together;
they jointly build and repair their hut; they carry firewood together; they
sometimes cook together; they often share care (and carriage) of their
young children; they look after each other during times of illnesses (see
Bird 1983a for a fuller description).
All in all, it is clear that Kinship does not play a major role in regulating
Nayaka life: Kinship roles and rules are neither discussed and formalized
by Nayaka nor manifested in everyday and kinship-prone occasions.
Relationships between kin are 'weak' in material and moral content in the
sense that they do not involve kinship-located obligatory economic
 Kinship organization 77

transactions, nor conspicuous instances thereof. Furthermore, they do

not involve kinship-located obligations to help, avenge, or take part in
life-cycle events.
Following The Social Construction of Reality'sfirstchapter, I now want
to argue that the 'person-to-person' situation impedes the dialectic
development of Kinship roles and rules. Nayaka know 'vividly' and 'fully'
most (if not all) of the people in their (Nayaka) social world - including
most (if not all) of their own and others' relatives of any given kinship
category. For example, they know in a 'person-to-person' way, say, all
their own cikappa(n), and all the cikappa(n) of anybody else. There is no
cikappa(n) they can think about in typical terms divorced from concrete
personality. Their interactions with diverse cikappa(n) are flexible and
influenced by idiosyncrasies and shared 'here and now', respectively.
Thus, to put it shortly, kinship does not dialectically develop into
Kinship.

Kinship and avoidance

Hopefully, there is a more convincing case now for the claim that 'kinship
regulates so little' - at least as far as cooperation between kin is concerned.
Is kinship associated with something else in this 'person-to-person'
situation? I want to suggest now that it is: while Nayaka neither discuss,
nor formalize cooperation with kin in relation to linguistically distinct
kinship categories, they do so - at least in a quasi-manner - in relation to
avoiding kin. Furthermore, on occasions they actually avoid interaction
with close kin.
My hosts commonly mentioned nachika. Translatable as shyness, or
reticence, nachika, they said, would bring them to avoid certain interac-
tions, for example, direct physical contact, frontal encounter, joint stay in
enclosed space, and joking. They suggested a rough correspondence
between the degree of avoidance and the closeness/proximity of the
relative. For example, strong nachika, they said, is felt with respect to
one's parent-in-law of the opposite sex who lives nearby, and, to a lesser
degree, with one's own parent. For example, one avoids joking with either
of them; but one steps out of one's path only when one sees one's mother-
in-law coming towards one.
Though somewhat faintly, nachika-related behaviour resembles what
the literature on tribal societies commonly associates with the rule-bound
non-joking relationship (distinguished from the joking one). However,
my hosts insisted that they do not follow any Kinship rule but behave this
way because, to diverse degrees, they feel nachika towards these kin.
Indeed, the nachika-related behaviour they speak about was neither
stylized, nor conspicuous in other ways.
78 Nurit Bird-David

Concern with habitualized avoidance behaviour is seen elsewhere:

Nayaka say that one must not refuse requests for sharing: if someone asks
one for something they say one 'ought' to give it. The closer the kinship
tie, the more this applies; one should refuse nobody, but least of all close
kin who live next to one.
All this means that cooperation with close kin in domestic and
subsistence activities is sensitive and trouble-prone. And indeed, conju-
gal pairs - who mostly pursue domestic and subsistence activities jointly -
rarely cooperate with other couples. Occasionally, they only cooperate
with single (unmarried) persons - and then avoid those who are close kin
(see detailed analysis in Bird 1983a).
Why do Kinship roles and rules develop - or quasi-develop - with
respect to avoidance rather than cooperation? A commonsensical expla-
nation first offers itself. Given a social scene which is dominated by the
'person-to-person' situation - rather than one characteristic of large
societies, which are dominated by alienated, anonymous situations -
concern with avoidance is understandable. The band organization throws
people into involuntary proximity, immediacy and intimacy with each
other. Therefore, their major concern lies with ways of maintaining inter-
personal space.
Over and above this, it can be argued that there are fewer personal
'interferences' to the dialectic development of roles and rules than there
are in the context of cooperation; for, by definition, avoidance involves
disengagement from the other person. Therefore, idiosyncrasies and
personal attributes do not interfere as much. In simple words, it does not
matter much if one avoids this cikappa(n), who has this particular
personality, or that cikappa(n) with his idiosyncratic personality - the
avoidance remains the same. Whereas the cooperation with each one of
them can be significantly different, depending on their personalities.
However, whatever the reason may be, we can conclude that while
Nayaka neither formalize, nor discuss Kinship roles and rules in relation
to cooperation, they do so to a modest degree in relation to avoidance.

A comparative perspective
Notwithstanding differences in degree (see below), Nayaka are similar in
broad terms to other hunter-gatherers on all the major points that
emerged from the analysis above. Most importantly, the 'person-to-
person' situation dominates social interaction among hunter-gatherers
generally. They are band societies - this is one of their most distinguish-
ing common characteristics (Lee and DeVore 1968:8) - who live dis-
persed in small groups. The common size of the band - called the 'magic
number' by Lee and DeVore (1968) and never since contested while
 Kinship organization 79

almost everything else has been - is twenty-five. Its members are attached
to a particular location - much more than was realized in the early 1970s -
but at the same time frequently visit others. Huts are casually built - often
more crudely than Nayaka huts - and they are also constantly recon-
structed and relocated, placing people in relations of immediate neigh-
bourhood with diverse others. Thus, on the whole, people live within the
'bodily manipulative zone' of a significant number of their fellows (I
would say, even, the critical mass thereof), far more, in any case, than they
would in settled communities.
Secondly, many hunter-gatherers use kinship terms extensively and
universally within the group (Barnard 1981; see also Woodburn
1980:105).
Thirdly, nevertheless, kinship 'regulates so little'. As Woodburn
further elaborated, kinship relationships are typically not 'load bearing';
they 'do not carry a heavy burden of goods and services transmitted
between the participants in recognition of claims and obligations'
(1980:105). 'Kinship . . . [only provides] a broad idiom for friendly rather
than hostile relations and a set of rough and ready expectations for
appropriate behaviour' (1979:257). More generally, 'people often do not,
at least explicitly, seem to value their own culture and institutions very
highly and may, indeed, not be accustomed to formulating what their
custom is or what it ought to be' (1980:106).
Finally, hunter-gatherers are commonly distinguished by a lack of, or
minimal, cooperation between people in general, and kin in particular
(P. Gardner 1966; Woodburn 1980). Families and individuals are highly
autonomous (P. Gardner 1991). They often sanction refusal of explicit
requests to share (e.g. Myers 1982; Ingold 1987). Many of them deal with
disputes by 'voting with their feet', not through kinship-related channels
(e.g. Turnbull 1968).
Though similar in broad terms, there are differences in degree between
hunter-gatherer groups on all these four points - and the differences are of
equal interest to us. Take for example the !Kung, as reported in the
detailed studies of Marshall (e.g. 1976) and Lee (e.g. 1979).
Firstly, they are larger in number than Nayaka. With an estimated
population of 6,500 (Lee 1979:35), the number of !Kung is more than
sixfold that of the 1,000-strong Nayaka. Compared with the 69 Nayaka of
the Gir locality, the Dobe !Kung (counted by Lee in his 1964 study)
comprise 379 'residents' (who mostly, but not always stay in the Dobe)
and 87 'marginals' (who spend less than six months in the area, and in
some cases only a day (1979:43)). Each !Kung hamlet (or camp) is larger
than the Nayaka equivalent. While the Gir Nayaka are distributed in five
residential clusters, the much larger Dobe !Kung are distributed around
nine water-holes. Moreover, being confined within highly populated
80 Nurit Bird-David

South India^, Gir Nayaka have little contact with Nayaka in other
communities, while Dobe !Kung maintain extensive links with other
!Kung communities, and occasionally come into contact with relatively
anonymous and even stranger !Kung. All in all, it can be said that
anonymity is more apparent in !Kung life - though, on the whole, they
also mostly interact with other !Kung in the 'person-to-person' situation.
Secondly, !Kung use kinship terms and personal names in everyday
communicative acts, much as Nayaka do, but have a more elaborate
system (see Marshall (1957) and Lee (1984) for detailed descriptions).
Names are given, each to a number of people (of the same sex), and
establish between them namesake relationships. In a unique way, these
name-relationships implicate kinship relationships: for example, a
woman who bears the same name as one's sister is referred to as a sister -
even if she is a stranger. Indeed, when a 'Kung arrives at a new unknown
group, he locates bearers of names identical to names of close relatives in
his home group and establishes corresponding kinship relationships with
them' (Marshall 1957:24-5).6
Thirdly, !Kung have significantly more kinship-located rights, obli-
gations and economic cooperation than Nayaka do (though still few in
comparison with non-hunter-gatherer societies). For example, men
customarily work for their wives' parents for some years after their
marriages, and are, later, obliged to always give the in-laws specific shares
of hunted game. In addition !Kung maintain extensive networks of hxaro
relationships - involving exchange of gifts and hospitality - and fre-
quently (at times weekly) gather for communal medicine dances.
It is most interesting that kinship-located behaviour goes in tandem
with kinship terms; the latter are applied to others - even strangers -
through the name-relationship practice and entail a corresponding
behaviour. Thus, a !Kung who, through similarities of name, calls a
stranger 'sister' also behaves towards her as towards a sister.
Finally, !Kung practice various avoidance behaviours with respect to
close kin, including (among others) parents, parents' siblings and
spouses' parents (see Marshall 1957: 19-20; Lee 1984: 65). These
avoidance practices seem to be more elaborate and more fully formalized
than among Nayaka.
The differences between Nayaka and !Kung, I suggest, can be
explained within the terms of this analysis adapted from The Social
Construction of Reality. It can be argued that, with the greater anonymity
that exists among !Kung, relating in turn to social and demographic
scales, the dialectic development of Kinship roles and rules goes further,
impeded less by the 'person-to-person' situation.
 Kinship organization 81

The emergence of roles and rules: AIP and evolutionary reading

Hopefully, the analysis shows how hunter-gatherers' kinship practices
are embedded in inter-personal social interaction and, furthermore, how
the emergence (or lack thereof) of Kinship roles and rules is influenced by
patterns of social interaction. Kinship terms, it has been shown, are learnt
within the context of social interaction; and roles and rules 'rise and fall'
depending on the predominant mode thereof.
Over and above this conclusion - which, in a way, achieves the purpose
of this chapter by itself- the analysis invites thoughts about the possible
evolutionary emergence of human roles and rules. Of course this is highly
problematic. We cannot overlook the considerable problems that are
involved in reading contemporary hunter-gatherers evolutionary, and
comparative ethnographies precessually. However, this has been tra-
ditionally done by scholars seeking insight and information on evol-
utionary questions - for lack of better alternatives as much as for other
reasons - and the present chapter, notwithstanding the problems, will
follow this tradition. It will depart from it in one significant respect,
however: the rationale of the exercise will not lie in presuming continuity
of a deterministic mode of production (between contemporary hunter-
gatherers and early humans) as the orthodoxy maintains, and is currently
heavily criticized for.7 Rather, it lies in presuming a suggestive similarity
in demographic and societal scale (at least relative to other contemporary
examples) which can help us to imagine what social life in emergent small
societies could have been like. But it has to be emphasized that in this
exercise we no longer talk about the Nayaka, the !Kung or other
contemporary tribal minorities we collectively call hunter-gatherers. We
'think with them' on evolutionary questions, trying to say something -
conjectural at that - on evolution.
Based on primate research, the naturalistic scheme for the evolution of
roles and rules would have been this: roles and rules evolved because it
was advantageous to make intents explicit, and behaviour predictable.
Once behaviour standardized, it was labelled and subsequently institu-
tionalized (see E. Goody 1978b; n.d.). Does human-based research
suggest otherwise?
Given a language-able human, three hypotheses can be drawn from the
above ethnographic examination (and comparison):
Firstly, linguistic objectification of kinship links (i.e., kinship terms)
existed before fully developed Kinship roles and rules. The Nayaka show
us how it could have been. What could have been the evolutionary
rationale of kin classification? Its use - a universal phenomenon after all -
could have demonstrated (and reproduced) inter-relatedness, which was
important for social group-based life. Furthermore, the terms could have
82 Nurit Bird-David

served for human individuals to initiate interaction with others from a

distance. One can imagine how, wanting to attract someone's attention,
they turned to verbal signs for biographic relationships (meanwhile
usefully emphasizing bonding inter-relatedness). Or in the dialectic
obverse, how distinct sounds they used for the purpose came to be
interpreted as symbols of kinship links. It is harder to conjure up how,
instead, they started to use objective, detached names.
Secondly, the emergence of roles and rules started with avoidance (not
cooperation) rules. For as individuals became aware of their individu-
ality, so they became concerned about inter-personal social space. At the
same time, they would be concerned to avoid disruptions within the
group. Group disruption was dangerous and, at the same time, highly
likely in group-based social life.
Thirdly, as groups grew and dispersed, anonymity entered the world,
yet people remained aware of kinship relatedness, expressed in the
communicative use of kinship terms. Thus, kinship terms played an
instrumental role in the dialectic development of Kinship roles and rules.
It is probable that general social interaction, in which kinship terms were
applied to group-recognized but personally unknown individuals (as
among the !Kung), were modelled on the interaction with personally
known kin. With this sort of demographic dispersal, the 'person-to-
person' situation would no longer impede the dialectic development of
Kinship roles and rules.
Compared with the naturalistic scheme, this three-stage sociological
one is, at the very least, more plausible. For the naturalistic scheme is
logically inconsistent. It posits an overall selective pressure for pre-
dictable behaviour - yet predictable behaviour is advantageous to others,
not to oneself; it is, in fact, disadvantageous to oneself. Furthermore, the
naturalistic scheme implies that (human) roles and rules are biologically
programmed, yet history and personal experiences, no less than the
ethnographic record on hunter-gatherers, speak strongly against this. In
contrast, the sociological scheme, following The Social Construction of
Reality, posits that language, social interaction and group-awareness
suffice for the dialectic development of roles and rules. The potential was
released and realized with demographic growth and the decreasing
dominance of the 'person-to-person' mode of social interaction.
The sociological scheme adds a subtle yet consequential role to AIP
thinking in the evolutionary emergence of Kinship roles and rules.
Firstly, language, which strongly reflects on AIP (see other contributors,
this volume), was imperative in the evolutionary process, not just a label-
provider. Secondly, kinship terms, particularly crucial in the process,
evolved within an interactive context involving anticipation and plan-
 Kinship organization 83

ning. Finally, minimizing group disruption concurrent with maximizing

private gains (as posited by the scheme) called for more sophisticated AIP
thinking. It called for anticipatory interactive planning in view of two
aims which often contradict each other and, therefore, had to be weighted
against each other. Such a situation generated selective pressure for a kind
of intelligence which enabled humans to arrive at novel solutions by
playing variations of old and socially accepted ones. It generated their
ability, as the phrase goes, 'to play within the rules'.

Conclusions
This chapter has, I hope, provided an AlP-compatible perspective on
hunter-gatherers, with particular reference to Kinship roles and rules. It
also illustrates, I hope, the scope of spin-offs from the AIP project. The
latter led, in this case, to a fresh perspective on contemporary hunter-
gatherers, and to an explanation of their otherwise enigmatic Kinship. In
turn, to add to The Social Construction of Reality, it offers an ethnographic
illustration of a 'prototypical case of institutionalization'; a case which
Berger and Luckmann themselves left unexplored, being concerned with
the developed form, and, more important, unfamiliar with band societies.

Acknowledgements
I am indebted to Esther Goody for recognizing the seeds of the ideas expressed
here in my doctoral dissertation; and for her comments on earlier drafts. I also
thank G. Kunda, Y. Shenhav, Y. Peres, A. Gopher and E. Illuz for their
comments on early drafts.

Notes
1 Since 1992, when this chapter was written, I have developed some of the ideas
contained in it. The results appear in Bird-David (1994).
2 The term is problematic, not least because it confuses early and contemporary
populations, both called indistinguishably by the same term.
3 For financial assistance during the first spell of fieldwork I am indebted to
Trinity College, Anthony Wilkin Fund, H.M. Chad wick Fund, Smuts
Memorial Fund, the Wyse Fund and the Radcliffe-Brown Fund of Cambridge;
and for the second spell offieldworkto the Horovitz Institute for Research in
Developing Countries, Tel Aviv, and to the Jerusalem Foundation for Anthro-
pological Studies.
4 A few Nayaka occasionally live in the fringe of the locality, in huts constructed
near villages or places of employment. They visit close friends once or twice a
year for several days each time.
84 Nurit Bird-David

5 Some non-Nayaka people live in and around the Gir valley. Nayaka maintain a
regular contact with them, however this lies outside the concerns of the present
chapter (but see Bird-David 1988).
6 !Kung also occasionally change names - like Nayaka - and then the name-based
network of kinship relationships change accordingly.
7 See Solway and Lee (1990) and Wilmsen and Denbow (1991) for summaries of
the opposing views on the issue.
PART II

The interactive negotiation of meaning

in conversation
 JURGEN STREECK

4 On projection

The topic of this chapter1 is a varied class of phenomena, most of them

small if not tiny, some elaborate and large, which occur in all kinds of
visible and audible shapes in all kinds of contexts in human talk and
interaction. What is common to all of them is that they occur as prefatory
components to bigger things to come. Prefaces range from rather minimal
units such as uh> well, or micro-moments of silence, to fully developed
pre-sequential utterances such as can I ask you a question? (Schegloff
1980). Gestures also are quite often performed in prefatory slots. The role
of prefaces - or pre's (as conversation analysts have fondly nicknamed
these pet phenomena) - is to 'foreshadow' or 'project' (Sacks et al. 1974)
something that comes after them, to bring it into play and 'prepare the
scene' (Schegloff, 1984b). They allow other participants a certain pre-
monition as to what this actor might be up to next.2
Vague as it is, this description is not likely to yield a neatly bounded set
of phenomena. The collection I describe is eclectic at best. The chapter is
loosely organized around a sequence of talk between two nurses from
Thailand who discuss weather conditions and proper attire in Germany.
This sequence was chosen because it nicely illustrates the theme that runs
through all of the examples, namely that interactional units foreshadow
one another: moment by moment, the speaker's gestures prefigure the
next moment, allowing the participants to negotiate joint courses of action
until, finally, a communication problem is solved collaboratively. Pre's, it
is suggested, are at the very heart of social collaboration in talk and
interaction. The various projections made within this sequence lead us
radially to other interaction samples, each of which instantiates a similar
type.
The interest in these small phenomena with respect to the investigation
of social intelligence and anticipatory interaction planning is that many
of the behaviour artifacts that humans have created to be used in
interaction (deception and other secondary stratagems excluded) are
shaped in ways that enable 'foresight' (see Good, Chapter 6, this volume).
In all their diversity pre's seem to point to an underlying organization, a

87
88 Jiirgen Streeck

design feature of the human interaction order itself (Goffman 1983). It is

argued here, in line with Drew (Chapter 5, this volume) and Heritage
(1990/91) that 'socially intelligent' action formats - ones that give reci-
pients immediate and maximum opportunities for social adjustment - can
not be attributed to actors' pro-social motives. Rather, social intelligence
resides as much in the cultural resources that interactants routinely and
automatically employ. They are among the 'prothetic devices' that make
up culture (Bruner 1990). The issue of where social intelligence lives -
whether it is 'in the mind' or 'around us', will be taken up at the end of the
chapter.

Action projection: pre's and possible trajectories

The term 'action projection' was introduced by Schegloffin an article on
'preliminaries to preliminaries' (1980).

Extract 1
-> B I like tuh ask you something 11
A Shoot. 12
B Y'know Fad my license suspended fuh six 13
munts, 14
A Uhhuh 15
B Y' know for a reason which, I rathuh not mention tuh you, in 16
othuh words - a serious reason, en I like tuh know if w'd talk 17
tuh my senator, or- somebuddy, could they help me get it back, 18
In this extract, B announces a question. But he then does not ask it, but
gives a report, which then turns out to be preliminary to the question. The
main action (question) is 'doubly displaced' (p. 106). The preface-
question indicates that what is to follow will be a preparation; everything
said between the preface and the action will be understood to be part of
this preparatory work.
There are other ways in which preface-questions are used; often they
are indeed followed immediately by the question. In these cases, the pre
marks the question as 'delicate':

Extract 2
PAM H']lo3 9
VICKY HK. Vicky. 10
(0.4) 11
VICKY You rang? 12
PAM jOh hello there yes I di^d 13
 On projection 89

.hh um I nee:d tuh ask you a 14

questio:n? 15
(0.4) 16
PAM en you musn't (0.7) uh take 17
it personally or kill me. 18
(0.7) 19
PAM I wan to know (0.7) 20
whether you: will (b) would 21
be free:, (.) to work o:n um 22
tomorrow night. 23
(0.4) 24

A preface question, then, does not unequivocally show its character as a

pre or pre-pre. Recipients may mistake one for the other. Pre's only
project possible courses of action. Not only is the meaning-in-context of
the preface underspecified - a 'prefiguring' rather than a 'figure' - but the
course of action sequences is always open to unforeseen contingencies;
participants can take unanticipated actions that alter the projected
sequence course. While pre's enable a bit of foresight, occasionally only
hindsight will reveal the prefatory role of an utterance or
utterance-component.
Preface questions can engender pre-sequences (Drew 1984).

Extract 3
-> j So who'r the boyfriends for the week. 1
(0.2) 2
M .k.hhhh- Oh: go::d e-yih this one'n that 3
one yihknow, I jist, yihknow keep busy en 4
go out when I wanna go out John it's nothing .hhh 5
I don' have anybody serious on the string, 6
-• J So in other words you'd go out if I:: askedche 7
out one a'these times, 8
M Yeah! Why not. 9
J's question turns out to be a preinvitation, and M may or may not have
heard it this way in the first place. This kind of ambiguity is essential to
the design of prefaces: they are designed to foreshadow possibilities,
which are often enough cancelled or redefined. Pre's are noncommittal for
good reasons. Their role is to 'point forward to possible patterns of
perception', to give a 'prescription of the path to a more precise determi-
nation' (Husserl, quoted in Liberman 1985:182).
Let us now take a first look at the mundane sequence of Thai
conversation at the Berlin Goethe Institute: who could have known this is
what they call summer, not a chance to wear these clothes. We will look
90 Jiirgen Streeck

especially at the gestures. Each one seems to foreshadow something and to

give the hearer an orientation, until eventually, a rapidly delivered
movement triggers the hearer's response. She becomes active and helps
out in a word-search.

Extract 4
((wiggle))

A Mai koi dai pra yo:t. (.)

You can rarely use it.
((scratches face))
I
Arai la man uh- (
What it is uh-
((clapping))

r
man mai mai mai mai mai moh- =
it's not not not not not suit-
((folds hands))
B Mai moh gab agart.
Not suitable in this weather
A Mai moh gab spab tang ni loei.
Not suitable in this climate.

Figure 4.1
 On projection 91

Figure 4.2

Figure 4.3
92 Jiirgen Streeck

Figure 4.4

Figure 4.5
 On projection 93

This sequence culminates in joint action. The collaborative action is the

'shared saying' at lines 4 and 5. It is brought about via a series of steps
which typologically fall under the category of'repair activities'. The term
'repair' in conversation analysis designates all those activities by speakers
and recipients which deal with (actual or virtual) problems in speaking
and understanding: speaking errors, lack of memory, failure to hear an
utterance or identify a referent, etc. 'Repair actions' are those that address
- or are made out to address3 - these problems: self-correction, word-
searches, asking for and providing clarification, etc. A useful distinction
separates 'retrospective' repair (e.g., correction) from 'prospective'
repair, i.e. repair on a (potentially) upcoming problem.4
Speaker A initiates repair on her turn at line 5, but at this point, repair is
already in play: the turbulences in line 3 mark her problems. Throughout
this sequence, the speaker makes a series of gestures, each of which
addresses the trajectory of the sequence in a different way: first she
foreshadows 'vagueness'; then she indicates that there is a 'problem'; next
she produces a gesture indicating a 'fit' - arguably a spatial representation
of the concept she is searching for, 'suitable'; finally, when the search is
completed, she folds her hands.
Let us examine a few other instances of joint action in which the
mechanics of this format are more transparent. A rather transparent
example is extract 5; it is a 'unison'. Three children, two girls and a boy, in
an American elementary classroom argue over an assignment. At this
point, one of the antagonists (Leola) is about to present her opponent
(Wallace) with evidence: she is about to read out loud the instructions that
are printed on the work-sheet. But while Leola initiates the moves, she
and her friend Carolyn end up reading these instructions in unison.

Extract 5
1 2 3
LEOLA You see. Here it say. Ho:w man. How many words
[
CAROLYN How many words
LEOLA can you make out of thosefiveletters.
[
CAROLYN can you make out of thosefiveletters.
The unison is prepared in a series of three prefaces (marked by numbers
above the text). First, Leola suggests a particular cognitive orientation for
the reception of her subsequent utterance (you see); then she projects
reading (here it say) - this is the action projection 'proper', finally she
begins to read (3), but immediately self-interrupts to then resume: she
thereby transforms the beginning of the action into an action-preface and
thus not only informs her friend that she is going to read, but also affords
94 Jurgen Streeck

her another opportunity to join her. In fact, restarts often solicit co-
participation (Goodwin 1981). The reading-in-unison, then, is a carefully
prepared-for event.
Extract 5 has the format
[action, action abandoned-• joint action]
This format can be observed in many, rather diverse interaction domains.
For example, pre-enactments often occur in 'juncture activities', i.e.
activities accomplishing or taking part in a more embracing 'recali-
bration' of an encounter. Among the actions typically engaged in during
such moments are posture shifts and activities dealing with a participant's
bodily needs (e.g. drinking, self-touch, clothing adjustments etc.).
A posture-shift is frequently pre-initiated by one party - by a small
'pre-shift' - and then carried out collaboratively and simultaneously by
two or several parties. A frequent format for drinking is that one party lifts
the cup or glass but then puts it back down. Thereafter, both parties
simultaneously lift their cups and drink. One of the social functions of
such joint activities at moments of juncture may be that by engaging in
symmetrical and identically constructed actions, the participants sustain
a state of engagement while no particular framework for the interaction is
selected. Yet another kind of action that is frequent during moments of
interactional juncture is self-touch: people in interaction become peri-
pherally involved with themselves when involvement with the other
decreases. Often, however, another participant will 'mirror' the self-
involvement, thereby turning it into a display of mutual engagement.
Responding to self-touch with self-touch in return enables a participant
to demonstrate to another that, although currently not involved with one
another, they are nevertheless involved in the interaction in the same way.
And again, frequently a participant who initiates self-touch without
observing self-touch in return by the other, will discard the activity to
then resume it; and the restart of self-touch is then done jointly, in
synchrony, by both parties. In other words, self-touch - considered to be
an instance of self-involvement - is often constructed in a way that secures
joint engagement in it.
The 'shared talk' in the sequence of Thai conversation, then, is brought
about via a very routinized, very widespread interaction design.

Possible repair pre-initiation

In contrast to body actions, talk is generally not carried out by two parties
simultaneously, but turn-by-turn, with a minimum of overlap. Extended
simultaneous talk is rare. However, collaborative talk does occur in repair
segments and sequences, in particular in word-search repair, of which the
Thai sequence is an instantiation.
 On projection 95

Repair is another domain of conversational organization in which

action projection can be observed. Self-initiated repair is often initiated
only after a pre-indication of 'trouble ahead' has been given. Within the
repair-segment proper, the speaker, by manipulating sounds and other
language units, can indicate what kind of audience-participation is
sought, whether or not the audience's support in dealing with the trouble
is wanted. Typically, there is a shift from non-collaborative to collabora-
tive organization: at first, self-repair is preferred; audience collaboration
is invited only after one or more attempts at self-repair have failed.
ScheglorT (1984b) has described the initiation of repair in American
conversation and specified 'the first signs in an ongoing flow of talk that
repair is upcoming' (p. 268). Most common are 'cut-offs' (glottal stops)
when a word is in progress, and 'uh's' when not. Sometimes there are
repair 'pre-initiators', indications of trouble ahead, e.g. hitches or sound-
stretches before the actual repair is initiated, as in extract 6.

Extract 6
1 2 3
ROBIN ^he hadda wait up the:re fo:r u-she:s been there
since eight uh' clock this morning
Two sound-stretches (1 and 2) pre-indicate the initiation of repair (at 3).
Schegloff points out that the more removed these 'harbingers' are from
the 'trouble source', the more problematic it is whether they are, in fact,
harbingers: even when repair is prefigured, it can subsequently be
'cancelled' (p. 269). In other words, 'harbingers' are only possible repair
pre-initiators.
There are interesting cross-linguistic variations in repair-initiation and
pre-initiation (and, by implication, in the organization of collaborative
talk). In conversations among speakers of Ilokano in the Philippines, for
example, 'cut-offs' and 'uh's' are unlikely repair-initiators, 'uh's' being
virtually absent from speech; by far the most likely initiators are sound-
stretches (Streeck 1989b).

Extract 7
A Ngem dagsan gasat ta::- ti sumagmamano nga aldaw nagarirulos 1
But after several days had passed
ti:::: Disyembre. 2
in:::: December
[
B Disyembre 3
December
Thus, sound-stretches are not only used as repair-initiators, but also - as
in American-English conversation - as 'harbingers' of such initiations.
96 Jiirgen Streeck

Ilokano sound-stretches, thus, are structurally ambiguous things: they

can be heard as harbingers of later repair, or as initiations of current
repair, or, when the talk quickly progresses, as (for all practical purposes)
'nothing at all'. What the sound-stretch amounts to, on any given
occasion, is to be determined by 'recipient's work'. These sound-
stretches can go to very considerable length, creating larger and larger
'opportunity spaces' for such negotiation; the longer the stretch, the more
compelled the recipient may feel to treat it as a 'repairable', thereby
securing the 'contiguity' (H. Sacks 1986 [1973]) or 'progressivity'
(ScheglofT 1979) of the conversation.
Similarly, speakers who have initiated repair can then actively pursue
the audience's co-participation and turn the sequence into a collaborative
one. This often takes place in word-searches.

Trajectories of a word-search
Once a speaker has initiated repair on a current utterance, he or she can
overtly mark the repair type as a word-search. The 'parts' with which this
is commonly done include pauses and 'recycles'; the element recycled is
the last unit of talk prior to the trouble source.
When entering into a pause, speakers engaged in a word-search
commonly withdraw their gaze from recipients. They thereby display
that they are currently 'non-listeners' in the moment's conversation and -
by implication - indicate that co-participation is dispreferred.5 When no
solution is found, gaze is returned to the recipient, and the 'preference for
self-repair' is relinquished or relaxed in favour of the 'progressivity' of the
conversation: co-participation is now invited. In the Thai sequence, the
speaker withdraws her gaze from the listener just before she enters into
the pause; she returns it just before she resumes speech.
Speakers of trouble sources can actively pursue collaboration. One way
in which this is done is to recycle the last morpheme prior to the trouble
source. These units project grammatical features of the 'target'.
In the following extract from a conversation between two speakers of
German, the recycled unit is son (so + ein\ a fusion of a deictic and an
indefinite article, indicating that the target is a noun and that information
relevant to the solution can be gleaned from the environment, in this case
from the gesture (Figure 4.7).

Extract 8
B Also ich hatte- ich hatte letztens w- irgendwo a ah ah ahm aufm
Well, I had- I had- the other day wh- somewhere- uh uh uhm at the
ArbeitsamtPoder (.) was das war. Da hatte ich ge- ah ah
Labor Office or (.) whatever that was. I had (ge-) uh uh
mal son son son son son son Merkblatt ah inner Hand?
one time this this this this this flier uh in my hand?
 On projection 97

Figure 4.6

Figure 4.7
98 Jurgen Streeck

In extract 9 - an Ilokano morpheme (in) projects grammatical infor-

mation which is taken up by the recipient in her attempt to help with the
word-search.

Extract 9
A Inkabil koti telepono idiay kwa:;-idiay sa:la, kunak a. 1
I had put the telephone in the what- in the living-room, I said
Sa in- kwak- in- in- in- in- in- uhh- 2
And then (in-) I what-
[
B Hmm. 3
B Insaram didiay ruangan? 4
You had closed the door?
Recipients may or may not opt for and/or be able to volunteer candidate
solutions. Similarly, candidate solutions can be treated by the primary
speaker as appropriate or as non-solutions. Recipients who volunteer
candidate solutions, in constructing them, make use of the materials
provided by the pre. The pre becomes a component of the solution - like
in the Thai sequence (extract 4) where the negation morpheme mdi is
incorporated in the solution.
man mai mai mai mai mai moh- = 3
it's not not not not not suit-
 On projection 99

B mai moh gab agart. 4

not suitable in this weather
[_
A mai moh gab spab tang ni loei. 5
not suitable in this climate.

Gestures as trajectory components

I now want to extend the range of phenomena subsumed here under the
category 'projection' yet a bit further, by addressing the incorporation of
visual components - gestures - into turns-at-talk. For 'projection' also
describes the communicative effects of many visual utterance compo-
nents: they often precede speech-components and thereby 'prepare the
scene' for them (see ScheglofT 1984b). Gestures - especially iconic
gestures - are often components of word-search sequences. For brief
moments, sense-making is primarily accomplished by the speaker's
hands. Both speaker and recipients attend to these manual symbol
constructions; that is, they temporarily shift their attention to them. This
shift in orientation is itself prepared.
Four successive gestures appear in the Thai fragment (see Figures 4.1-
5 above). Each gesture is constructed differently, and not all occur at the
same spatial location. The first is carried out in the neutral space where
most descriptive gestures are done: in front of the speaker's torso. It is a
slight horizontal rotation (Figure 4.1). It conveys a sense of ambiguity,
indecisiveness, vagueness. The second visual act (at the beginning of line
2) involves self-touch: the speaker looks away and rubs her face (Figure
4.2). Interestingly, as Goodwin (1986) has shown, facial touch tends to be
disattended by recipients - it 'drives their gaze away'. Thus, by rubbing
her face in the course of a word-search, the speaker in effect discourages
the listener from attending to or even becoming involved in her current
action; she thereby secures the possibility of self-repair (the 'preferred'
repair solution). The speaker then returns her gaze to the listener and
brings her hands into a position from which she can execute a gesture
(Figure 4.3), and immediately produces a series of claps (Figure 4.4).
These are synchronized with the recycling of the function word mai. I
regard these motions as, on the one hand, vaguely iconic displays of a 'fit',
and thus as projections of 'suitable', and, on the other, as acts of self-
stimulation (which are not uncommon during moments of speech-
production problems). Finally, when the recipient begins to offer a
solution (line 5), the primary speaker freezes the gesture and, once the
search comes to completion, she returns her hands to her lap and folds
them (Figure 4.5). These gestural trajectory components all instantiate
'types'; their organization is not idiosyncratic. We will describe how these
100 Jiirgen Streeck

gestures work by looking at a few more examples; they represent 'iconic

gestures' and what one might call 'action projectors'.

An iconic gesture's trajectory

Iconic gestures commonly precede their 'speech-affiliates'; they unfold -
and often decay - before the word arrives. This makes them interesting
phenomena for a study of projection in interaction.
Extracts 10 and 11 are from a conversation between two Japanese
women who tell each other about car-accidents in which they were
involved. In extract 10, the speaker verbally encodes a motion: supin (the
loanword 'spin'), lexicalizes a rapid circular movement; the gesture
visualizes the same conceptual profile, but further specifies it as a motion
on a plane.

Extract 10
1
T .h nichiyoobi datta shi ne (—) kuruma suiteta shi h 1
But, well, it was a Sunday and, you know, the car was empty and .h
(2.0)
2
supin shite 2
it spun around
The speaker begins the gesture during a pause in her turn (1), before
she utters the verb. She looks at the gesture. She then returns her gaze to
the recipient, after uttering the verb (2). Thus, while gesture and speech
share a semantic profile, the gesture nevertheless 'projects' that profile
prior to the speech.
In the second instance (11), taken from the same compound utterance,

Figure 4.9
 On projection 101

Figure 4.10

Figure 4.11

the speech unit - the noun gaadoreru (the loanword 'guardrail') profiles a
long, horizontal object ('rail') as well as its function ('guard'). The
gesture, in contrast, profiles a thin, round, vertical object, a part of the
object - the pole. There is thus a remarkable division of labour between
speech and gesture; each supplies a different partial conceptual profile. In
this instance, the part of the object selected for representation by gesture
is not a salient semantic component of the lexical unit guardrail. But it is
salient in the course of this story, because it is the pole that was hit by the
car (Figures 4.12-13).

Extract 11
T watashi no hoo no seki ga ano hora ( ) are
the seat on my side, you know, look ( ) there
aru ja na (—) gaadoreru
was a guardrail
102 Jiirgen Streeck

Figure 4.12 Figure 4.13

Again, the gesture is initiated far before the speech-unit to which it

'belongs'. This is the characteristic place of gestures in turns-at-talk: they
preface speech units and prefigure the concepts communicated by them.
They thus enable recipients - in varying degrees and depending upon
contextual information provided by prior talk - to anticipate conceptual
profiles of subsequent talk. The semantic relationships between the
profiles supplied by the gesture and those encoded in lexical units are
manifold. In these two examples, one gesture (in extract 10) overlaps with
the lexical concept 'spin' but additionally encodes a horizontal plane; the
gesture in the second example (extract 11) highlights a feature of the
object which is not a prominent feature of the lexicalized gestalt. The
gesture thus supports the story by prefiguring a component (the guard-
rail's pole) that will be relevant in the further course of events. The
speaker then moves on to show, by gesture, how the car hit and wrapped
around the pole. By virtue of both their placement and their specific
conceptual profiles, gestures afford recipients opportunities for anticipa-
tory understandings.
Iconic gestures are symbolic constructions, equally motivated by the
speaker's analysis of the signifie and by her analysis of the local require-
ments for understanding. They make use of established symbolic conven-
tions and of prefabricated parts; but gestures at the same time allow local,
idiosyncratic elaborations. Iconic gestures are designed to communicate;
they provide imagery, kinaesthetic profiles. But given their regular place
of occurrence — as prefaces to speech - they only 'point forward to possible
patterns of perception' give a 'prescription of the path to a more precise
determination' (Husserl, quoted in Liberman 1985: 182; my emphasis).
While iconic gestures prepare the scene for upcoming linguistic
representations, they, too, arrive on a prepared scene. In order for
gestures to receive the attention of the listener and to thereby become
 On projection 103

components of conceptual understanding, listeners must be pre-alerted

to the gesture's occurrence. A common way in which speakers solicit the
audience's attention to a gesture is the use of a deictic particle which points
to the movement. This item is incorporated into the talk at the moment
when the gesture begins. In the following example (from Goodwin 1986),
the 'pointer' is the demonstrative this.

Extract 12
A Ma::n she's this wi::de.
In extract 13 it is the demonstrative som.

Extract 13
A Sie stand unten? Mit som weifin Rock nur?
She stood down below? In this white skirt only?

Words, thus, establish the gesture's relevance to the talk.

Secondly, speakers who produce iconic gestures withdraw their gaze
from the recipient and briefly focus it on their own hands. Depending
upon the gesture's placement in the listener's field of perception, the
recipient also shifts her gaze. Speaker and recipient thus share in a
momentary orientation to the gesture, which thus becomes the primary
medium of communication (Streeck 1993, 1994).
This is also the design of gestural communication during word-
searches. When the speaker has difficulty finding an appropriate word, he
or she can opt for gesture as subsidiary symbolism, and the visual image -
along with other contextually available information - can then be used by
the recipient in formulating a solution. The speaker uses deictic parti-
ciples and gaze to orient the recipient to the gesture and, characteristic-
ally, recycles a grammatical morpheme while he or she produces the
gesture.
Extract 8 above - although completed by the primary speaker - has
design features similar to the Thai sequence. Two German postgraduates
discuss the difficulties of finding employment. B describes a visit to the
employment office. He wants to relate how he read something in a flier;
flier (Merkblatt) is the target of the search. As he moves into the search,
the speaker lifts his hands and shows the shape of a booklet. He looks at his
hands, thereby displaying 'reading'. The recipient's gaze follows his
hands. When the speaker completes the search he - like the Thai speaker -
folds his hands (see Figures 4.6-8 above).
There is thus a very systematic, apparently transcultural organization
104 Jurgen Streeck

to the use of iconic gestural symbols in human communication. Speakers

and recipients utilize the projective potential of manually produced
images, and they use yet other communicative modalities (gaze and
speech) to highlight the local relevance of the manual mode.

Action projectors
Not all gestures in interaction are of the iconic type, representing visible
features of real-world objects and events. Another way in which gestures
are used is the prefiguration of linguistic action, e.g. of the construction
features of a subsequent utterance, the speech act about to be performed,
the type of response proposed, etc. Only little research has been done on
these uses of gesture, and the following remarks are tentative.
One place in unfolding interaction sequences where gestures project
upcoming speech actions are 'transition places' between turns-at-talk.
Participants who intend to take the turn sometimes demonstrate this
intent gesturally, while another participant is still talking. Intending next
speakers can also prefigure aspects of the action-type for which they
solicit the turn (i.e. a gesture can be a recognizable 'pre' to a 'telling').
Turn-transition is a rather specific environment in conversation. The
very circumstances of the task - turn-taking - occasion the use of the
visual mode. Using gestural displays, intending next speakers can make
their claim to the floor known and show what they plan to do with it -
without interrupting the current speaker and without subjecting their
own premature talk to overlap. Preliminary research on gestures used
during turn-transition (Streeck and Hartge 1992) suggests that these are
highly conventionalized, less formally elaborated than iconic gestures,
and often metaphorical.
Turn-exit is another place where gestures that project actions are used.
The task they are occupied with is the specification of the response-type
sought. Gestures are occasionally appended as 're-completors' to turns.
For example, shoulder shrugs are sometimes appended to the end of a
story-telling. Story endings are interactionally under specified: audiences
have considerable leverage in selecting a type of response. Shoulder
shrugs are commonly done when no immediate response is given, and
display that no specific type of response is proposed.
Another place for visual displays during turn-endings is before the
'utterance-completor'. Often, a micro-pause makes room for the enact-
ment. Gestures or displays thus positioned can solicit an 'early response',
e.g. applause. In extract 14, a talk-show guest frames the 'keyword' of a
punchline by a facial display. The guest, the comedian Richard Jenni,
talks about his childhood and youth in Brooklyn.
 On projection 105

Extract 14
And plus I had- no, I was into different things, I wa:s- 1
1
and my particular neighborhood wa:s, you know (.) 2
2
very into: uh (.) crime I guess is the word. 3
In this extract, the arrival of the keyword (crime) is projected by a 'broad
smile' (2) which proposes 'laughter' as an appropriate type of response;
the smile, however, is itself projected by a brief eyebrow-flash (1). This
sequence is a good illustration of the ways in which projection contributes
to the rapid fine-tuning that is characteristic of human interaction.
Speakers consistently foreshadow their next moves, and while this
foreshadowing does not specify all features of this move but only projects
a silhouette, it nevertheless prepares the audience and often enables them
to initiate a response before the act to which they respond is actually
completed. Action images and bodily displays of human states are
frequently used as dramaturgical devices in narratives. In extract 15, the
speaker talks about her always depressed roommate; here she describes in
what state she finds her roommate when she comes home.

Extract 15
1
A und Sibylle. (.) mit Grabesstimme und so. 1
and Sibylle with a voice like from like a grave
((voice shift)) .hhh Was ganz Schreckliches passiert. 2
Something terrible happened.
The facial display (1) is a 'mouth of sadness'. By making the display the
speaker enables her listener to 'meet' Sibylle before 'listening to her'. The
speaker recreates the temporal structure of the event she describes.
We can now return to the sequence with which we began this survey of
action projection and summarize the different 'paths' or 'trajectories' that
eventually culminate in a collaborative completion. The path towards a
word-search initially comes into play when turbulences appear in the
utterance. The possibility of imminent trouble that is projected by these
hitches is then specified when the speaker makes a first gesture which - in
this context - can quite readily be identified as a projection of vagueness.
At this point, a formulation problem is clearly in play. The speaker then
addresses the trouble with a face-rub, a 'monitor' (Scheflen 1972) not
uncommon in troubled interaction environments. By touching her face
and at the same time withdrawing her gaze from the listener, the speaker
invites the listener to disattend the problem, that is, not to become
106 Jurgen Streeck

involved in the search. In other words, the speaker's visible actions

project self-repair as the preferred trajectory for this sequence. But the
speaker then reorients to the listener and initiates a series of iconic
gestures. She thereby invites the listener to aid her in the search. Once the
word-search is collaboratively completed, the speaker displays that
successful completion by folding her hands. Each move thus prefigures a
possible path for navigating through the sequence.

Language and such: socially intelligent shareware

In this chapter, a haphazard collection of interaction units has been
described; each item in the collection in one way or another foreshadows
shapes of things to come. 'Things to come' includes varied appearances
ranging from posture shifts to nouns (which can be premarked for gender,
number, case and so on). The term 'trajectory' has been used to describe
the unfolding of symbolic and interaction units from a first indication that
something is 'in play', to the unit's completion and ratification as a social
act. The underlying theme has been the suggested need for a description
of language and behaviour as phenomena in progression, unfolding towards
(possible) completion.6 The hallmark of pre's is that they pre-indicate
default courses of action. It appears that this design is an imprint of a
rather generic social intelligence.
But where does this intelligence live? As far as pre's are concerned,
when we pre-decline an offer, do we each exercise our individual
sensibility, or do we just do 'what everyone does?' In this case the social
intelligence of our action is owed to the common resource, the socially
available conventional 'format for declination'. So when we ask, 'where
does social intelligence live, in our heads or in the world around us?', the
answer must obviously be: in both (and in other places as well). Reviewing
only the odd collection of moments illustrated here, it appears that social
intelligence resides partly in our bodies: a tendency to go along with
another organism's 'intention movement' (as in the posture shift ex-
amples), or a tendency to inhibit certain acts (and to restrain associated
affects (D'Andrade 1984)), as in the case of delayed antagonistic actions.
To what extent this intelligence is not only embodied but also innate, and
to what extent it is traditional and learned, is not completely clear. The
boundary between these classes of phenomena might well be fuzzy.
Certainly, innate impulses are generally shaped into traditional mean-
ingful forms. In either case, the effect is a tendency within interaction
processes towards avoidance of antagonistic action and towards maximiz-
ing opportunities for collaboration and solidarity (Heritage 1984: 277).
Much embodied social knowledge needs other, interlocking 'bodies of
knowledge' to become alive and to function. (This is true of all competen-
 On projection 107

cies relating to entrainment phenomena such as shared speech and

movement rhythm (Stern 1977; Trevarthen, 1979a, b).)
Then there are the many kinds of knowledge that we have acquired as
competent individual members of a society or culture; much of this
knowledge presumably resides in our separate minds/brains. (The latter
is what is commonly studied under the terms 'knowledge' and 'cogni-
tion'.) As far as pre's are concerned, some also may occur because their
users possess certain items of propositional, cultural knowledge, for
example that 'one doesn't say no to a request for a favour'.
And finally (to simplify matters drastically), social intelligence is
embodied in the symbolic resources that we share, in particular our
languages, in their structural design. Simply by using their normal forms,
the ones that the grammar wants us to use (by making them easy to
master), we gravitate towards civility in our interactional dealings with
one another. Not acting upon the consensus presupposed by the normal
forms takes a little bit of extra constructional effort. This share of the
corpus of knowledge belongs to the traditional, evolved interaction/brms,
not to individual users. All that people have to do is 'plug in' to the public
shareware in a routine fashion and take it from there. To an extent at least,
one can be fully unconscious and still act intelligently. Symbolic and
interactional resources are always open to clumsy and refined, to naive
and self-reflexive uses. But just participating in conversational interac-
tion has a mildly civilizing effect.
Hence, to be able to describe even the most trivial and mundane
moments of symbolic interaction, we must acknowledge that knowledge
and intelligence are distributed across a variety of 'locations': bodies,
individual minds, other minds, cultural symbolisms and external
memory systems. How bodies of knowledge are represented depends on
where they reside, how they are used, and what purposes they have.7 But
any modelling of intelligent activity as a mere externalization of indivi-
dual mental knowledge resources is false and misleading; even when we
silently talk to ourselves our communicative intentions are shaped by the
conceptual schemata that our common languages offers us (Bruner 1990).
While the computer metaphor for the human mind has perhaps caused
more confusion than enlightenment, it is aptly used by Donald (1991:17)
when he suggests that 'the cognitive architecture of humans ... is similar
to providing ... a computer ... with a link to a network'.
If a computer is embedded in a network of computers, that is, if it interacts with a
'society' of other computers, it does not necessarily retain the same 'cognitive
capacity' ... As part of a network, the computer can now delegate computations
beyond its own internal capacity ... It can store its outputs anywhere in the
network... The point is, in a true network the resources of the system are shared,
and the system functions as a unit larger than any of its individual components.
{Ibid.: 310)
108 Jiirgen Streeck

The link to a network is for us humans primarily provided by language,

which is not only a code of communication but also an external memory
store. Thus, we cannot properly point to a location in space when asked
where knowledge is located.
In his well-known essay on speech genres, Bakhtin (1986 [1952-3]: 88)
wrote:
The words of a language belong to nobody, but still we hear those words only in
particular individual utterances ... Any word exists for the speaker in three
aspects: as a neutral word of a language, belonging to nobody; as an other's word,
... filled with echoes of the other's utterance; and,finally,as my word,... already
imbued with my expression.
Communal ownership extends not only to words as form of self-
expression, but also to the meanings inherent in them, to the conceptual
matter or 'conventional imagery' (Langacker 1987). Language, including
grammar, is conceptualization, a vast symbolic resource for making sense
of our experiences in the terms in which generations before us have made
sense of theirs (D'Andrade 1990; LakorT 1987).
Grammatical structures... are inherently symbolic, providing for the structuring
and conventional symbolization of conceptual content. Lexicon, morphology,
and syntax form a continuum of symbolic units, divided only arbitrarily into
separate 'components' - it is ultimately as pointless to analyze grammatical units
without reference to their semantic value as to write a dictionary which omits the
meanings of its lexical items (Langacker 1987: 1-2).
But linguistic structures are not only responsive to cognitive (concep-
tual) and communicative demands. In interaction contexts, the places
where languages come to life, they also meet with the demands of rapid,
open-textured, sometimes risky social situations. The highest chances of
surviving in the primary forests of human interaction belong to those
conventionalized formats that meet the two kinds of requirements
concurrently. But for them to reveal themselves, a descriptive language
will be needed which evokes the progressional character of language in
talk-in-interaction (Lerner 1987). Perhaps a language of 'gestalt' and
'appearance' is best suited to the emergence of grammatical-semantic
units over time. Whatever the eventual shape of the descriptive language,
'projection' cannot be explained unless we assume that grammatical
constructs are structured wholes, rather than bottom-up products of
compositional syntax (Langacker 1986). In the origin and deployment of
language schemata out of which interactional turns can be constructed,
conceptual and social intelligence meet.
Pre's are simple but artful upshots of forward-looking utterance
design. Clearly, they are part of langue, and any specification of their
proper use must take account of the social and syntactic fact that pre's
belong at sentence and/or utterance beginnings. To properly describe
 On projection 109

language units we must constantly remind ourselves that 'the natural

environment of language use is talk-in-interaction ... The natural home
environment of clauses and sentences is turns-at-talk' (Schegloff, 1992c).
It is irrelevant whether we are dealing with items such as uh or well, or
schematic constructions (e.g., 'left dislocation'): the phenomena clearly
fall into the domain of traditional grammatical description, but they are
equally clearly interaction-motivated phenomena, some of them exclusi-
vely so. Abstracting them from either force deprives them of their beauty:
the density, economy and transparency of natural symbols.
Pre's are humble things, and they may come from humble beginnings.
They could well be 'vestiges' from pre-human stages of evolution
(Darwin 1964 [1859]).8 They can be traced to an initial recognition by
social animals that 'individuals can benefit not just from obtaining
information from each other, but also from making it available to each
other', because 'the information ... might... reduce the ambiguity in a
developing social interaction' (W. Smith 1977:9). Mead (1967 [1934]) has
defined his concept of 'gestures' similarly: a gesture is an early phase of a
social act, 'that phase ... to which adjustment takes place on the part of
other individuals in the social process of behavior' (p. 46). The gestures
'serve their functions [by] calling out the responses of the others, these
responses becoming themselves stimuli for readjustment, until the final
social act itself can be carried out' (p. 44). In Mead's view this 'conver-
sation of gestures' was the breeding ground for symbolization and the
human mind.
Social intelligence is very, very old, much older than humanity. Some
ancient, pre-human intelligence has survived from times before our own
species emerged. It is alive and well in our modern interactions, distri-
buted across various kinds of locations. Some of this intelligence is
certainly deeply buried in our bodies, inherited by nature and unbeknown
to us, but experienced, nevertheless, in the form of feelings (embodied
social cognitions (M. Rosaldo 1984)). But a lot of it, interestingly enough,
lives in our artifacts, our symbolic inventions. Its place is the 'network',
the 'shareware', the 'tribal encyclopedia' (Havelock 1963). Exposed to the
stormy weathers of moment-to-moment interaction, miniscule objects
such as pre's 'acquire the precision of continuously refined exterior
devices' (Donald 1991: 11).
Language - any language, in effect, any word - is entangled in an
unbroken tradition of talk in interaction. Like life, language is always in
progression, a 'transitory [yet] enduring thing' (Humboldt 1988 [1836]).
That is why we moderns, in our urban, self-reflexive ways, still draw upon
the wisdom of Homo erectus when it comes to the nitty-gritty of moment-
to-moment interactional life. Luckily, there has been no gap in the oral
tradition.
110 Jiirgen Streeck

Notes
1 This chapter owes a lot to conversations with Robert Hopper, about 'trajector-
ies' and 'possible trajectories'.
2 Among the many conversation analysis studies devoted to forms of pre's and/or
action projection are Drew (1984); Goodwin (1986); Heritage (1990/91);
Jefferson (1992); Kendon (1976); Pomerantz (1984); Schegloff (1979, 1980,
1988b); Streeck (1992); Streeck and Hartge (1992); Terasaki (1976).
3 See Jefferson (1975) about repair as social camouflage.
4 Schegloff, Jefferson and Sacks (1977) give a typology of repair.
5 See Goodwin's (1986) analysis of the role of gaze in the negotiation of
participation in word-searches.
6 Conversation analysts make much of the need to say 'possible completion'
because, in order to capture the progressional nature of talk and language, it is
necessary to envision and conceptualize everything from the reference point of
an idealized now. The future appears as a horizon of possibilities which, the
more apparent they become, the more likely they are to be preempted.
7 See the beautiful study of 'The dialectic of arithmetic in grocery shopping'
(Lave et al. 1984). See also D'Andrade 1981; Goodwin 1993; Hutchins and
Klausen 1990; Lynch and Woolgar 1988.
8 This is how Donald (1991:3-4) summarizes his modern variant of Darwin's
thesis: 'The essence of my hypothesis is that the modern human mind evolved
from the primate mind through a series of major adaptations, each of which led
to the emergence of a new representational system. Each successive new
representational system has remained intact within our current mental archi-
tecture, so that the modern mind is a mosaic structure of cognitive vestiges from
earlier stages of human emergence. Cognitive vestiges involve the evolutionary
principle of conversion of previous gains and are similar in principle to the
many other vestigal behaviors we possess - for instance, baring the teeth in
anger, or wailing in grief.'
 PAUL DREW

Interaction sequences and anticipatory

interactive planning

In this chapter I shall consider some relationships between sequential

patterns or organizations in talk-in-interaction, notably in conversation,
and that aspect of social intelligence which Goody has termed 'anticipa-
tory interactive planning' (hereafter AIP). Social intelligence consists, in
part anyway, of cognitive models of action which might underlie the
production and interpretation of 'meaningful' communicative behaviour
among social beings (humans, and possibly some primates; see Byrne,
Chapter 1, this volume). AIP highlights the reciprocity of communica-
tion in social relations, incorporating as it does the mental representation
of alter's responses to ego's actions (Goody, Introduction to this volume).
Goody is thereby suggesting that AIP concerns the possibility that, via
their cognitive models of action, interactants have ways of predicting that
if they make a given verbal 'move' (and from now on I shall refer
exclusively to those actions which are verbal), this will engender or
facilitate a subsequent verbal action by either themselves or their reci-
pients. The predictability of some contingent subsequent action(s) is part
of 'mentally modelling' the likely behaviour/responses of their co-
interactants, which hence underpins the strategic nature of selecting a
current action.
The ability to anticipate subsequent verbal actions, performed in turns
at talk, and to select or design a current action (turn) accordingly, would
seem to rest on the cognitive representation of sequences of actions. I mean
by 'sequences of actions' the discernible shapes, patterns, organizations
or regularities which may be associated with, or which may be the
products of, the progressivity of participants' decisions about 'what to say
next' in response to what was just said. (An extended account of the
progressively unfolding character of contributions to talk, from the
perspective of those participating in the talk, in real time, is given in
Streeck's chapter in this volume; hence it is unnecessary to say more
about that here.) Thus, underlying AIP is some mental representation of
interaction sequences. For this reason I shall focus here on the ways in
which sequential patterns might be associated with AIP, and particularly
on whether a current turn-at-talk might be produced with an eye to an
anticipated future slot or move in a sequence.

ill
112 Paul Drew

I am going to be considering whether a case can be made for the

possibility that AIP may involve the conscious use by speakers of
knowledge of sequential patterns, in attempts to generate subsequent
environments or slots in a sequence in which favoured or unwanted
actions may be, respectively, performed or avoided. However, it should
not be supposed that the relationship between social intelligence and
manifest verbal behaviour is one which necessarily involves consciously
planned action. Communicative strategies do not need to be located at the
conscious level of intentionality of speakers; nor do the other concepts
associated with AIP such as 'planning', 'anticipation', 'predictability' and
'control'. It is perfectly consistent with AIP to regard the 'mental
imaging' of interactive sequences as cognitive procedures for action and
interpretation, and not as part of the overt social knowledge which
speakers may possess in order consciously to model the contingent
responses of their co-interactants. It is thus not necessary to account for
sequential patterns in terms of these being the products of speakers'
conscious knowledge of the organizations that inform their conduct
(Heritage 1984: 241; 1990/91). Whilst participants may orient to stable
patterns of talk-in-interaction (and hence recognize the planned, predic-
table and rational character of communicative actions), this does not
imply that particular speakers on particular occasions engage in con-
sciously planning their utterances. Social intelligence may consist of
social procedures for action and interpretation, without these procedures
becoming the objects of conscious articulation (in the mind, at least) on
the occasions of their use.1 We are not, then, required to find evidence of
conscious intentionality in order to sustain the view that 'anticipatory
planning' involves interactants' representations - at a cognitive level - of
the organized contingencies generated by the selection of actions.
Indeed, the very business of considering what evidence there might be
for the conscious or intentional use of sequential strategies might appear
to risk confusing interactants' attributions of intentionality to one another
with an analyst's version of participants' intentions (in something like a
reconstructed narrative conversation). When participants engage in a
conversation, any move or turn in that conversation is accountable; that
is, inferring what an utterance means and selecting an appropriate
response involves ascribing intentions to our co-interactants. The case for
the centrality of intention-attribution in making 'rapid interactional
moves in an ongoing sequence of actions structured at many levels' is
elegantly argued in Levinson's contribution to this volume (Chapter 11,
p. 221). Indeed, Levinson defines social intelligence as 'just and only the
core ability to attribute intention to other agent's actions, communicative
or otherwise, and to respond appropriately in interdigitated sequences of
actions' (n. 1). Thus it might appear that the focus of enquiry ought to be
 Interaction sequences and AIP 113

fixed on co-participants' attributions of intentionality to one another's

actions. To do otherwise might risk missing the domain of interactional
accountability for participants in the conversation itself.
Certainly our investigations of the social organization of talk-in-
interaction should continue to focus primarily on participants' analyses of
one another's conduct, their understanding of what each other means/
intends, and on the consequent sequential progressivity of talk. Never-
theless, we might also consider whether participants, under certain
circumstances and for certain sequential patterns, may consciously
'orient to' knowledge of those patterns and employ that knowledge in an
intentionally strategic sense. Whilst the planning, anticipation, control,
strategy and prediction which are associated with AIP may be cognitive
operations, there might be evidence that on occasions these are conscious
operations of sequential management.
There are plainly certain respects in which conversationalists con-
sciously orient to sequential patterns. For example, the structure of
greetings exchanges is well enough known for people to anticipate what
they should do in a particular slot when they come to greet someone (e.g.
should they kiss or shake hands?), or in the course of a greeting, to choose
the appropriate moment for the performance of an anticipated move in
the exchange.2 Knowledge about such 'ritualized' exchanges, and how to
behave with propriety during them, is of course so much part of the
vernacular that they may be taught to children as rules of social conduct.3
Furthermore, such knowledge is a resource for deception: for example, in
a telephone greeting a caller may design his/her opening turns in the
greeting exchange so as playfully to deceive the person s/he has called into
believing that s/he (the caller) is not known to the called, when in fact s/he
is (for an analysis of which see Drew (1991)).
But there are more complex sequences which might manifest the
intentionality of conduct. For instance, Jefferson (1993) reports that in
circumstances where a recipient has apparently misunderstood a
question, and replied about something which was not asked, the speaker
(i.e. the one who asked the question) may refrain from correcting the
recipient; instead the speaker first responds appropriately to the answer,
and then re-asks the original question. So the speaker allows an error to go
uncorrected: but the subsequent move of re-asking the question (in a
manner which has a 'first time' appearance) demonstrates that the speaker
understood that an error had been made (i.e. the error was observably
relevant to the speaker, and did not simply go unnoticed) and puts things
to rights. Such instances involve more than the (cognitive) orientation to a
procedure: the sequence is evidence of the speaker recognizing the
recipient's error, refraining from correcting it, but remedying the error
by choosing to ask the question again - all of which manifests the speaker's
114 Paul Drew

consciousness of the error, and the intention to pursue the original

question.
So what I am beginning to consider here is whether there is evidence, in
the sequential management of talk, for participants' conscious orientation
to sequential patterns or procedures. If there were such evidence, then the
procedures or organizations which underlie what Levinson refers to as
'interactive intelligence' could be employed at a conscious level of
anticipatory planning of subsequent moves or turns in a sequence. If so,
we could add a level of consciousness in actors' representations of their
own actions and the contingent responsive actions of others - envisaged
by AIP - to the cognitive imprinting of the procedures for producing and
interpreting conduct. The issue is whether the knowledge of sequential
patterns which underlies mental representations of action may be con-
sciously exploited in interaction.
A couple of caveats, very briefly. First, undoubtedly there are many
other areas of research into the organization of conversation which have a
close bearing on social or interactional intelligence and to AIP, but which
will not be touched upon here. Such areas as 'recipient design' and the
organization of self- and other-repair are, for instance, quite evidently
related to anticipatory planning: Streeck touches on these phenomena,
especially self-repair, in his analysis of 'projection' (Chapter 4, this
volume). I shall not be considering the details of the design of particular
speaker turns, except insofar as the action which a speaker selects to do in
a turn has sequential consequences. Second, I shall not be concerned with
the larger or more diffuse goals which participants might have in
conversing (such as the order of priority in the day's business, cementing
a friendship or securing the other's co-operation), or in choosing to raise a
particular topic in the conversation. Although the pursuit of such goals is
certainly to be considered part of AIP modelling, I do not deal here with
how far such goals may be evident in particular conversational sequences
(but see particularly the various contributions to Tracy (1991) for a
consideration of goal-directedness in conversation). The discussion
which follows is focused just on the management and planning of local
sequences (or types of sequences), involving local sequentially con-
strained goals.

Some types of sequential patterns in conversation

It was noted above that certain sequences are so conventionalized as to be
part of the vernacular, for instance in instructing children how to behave
properly: so children are taught to say 'thank you' when given something,
to answer questions and to say 'hello' when greeted by another. Such
 Interaction sequences and AIP 115

sequences as greetings-greetings and questions-answers are instances of

what have been termed in conversation analysis 'adjacency pairs': this is a
large class of sequences in which the production by one speaker of a first
action (e.g. a question, request, offer, greeting, invitation, accusation etc.)
occasions the relevance by the recipient of a paired next action (e.g. an
answer) or one from alternative paired actions (e.g. acceptance/rejection,
granting/rejection). The conventionalized expectation that if a speaker's
turn or utterance is hearable as the first part of such a pair, then the
recipient should enact the second paired response, is a demonstrably
'strong' normative sequential constraint on what a recipient should
appropriately do in the next turn (Atkinson and Drew 1979: 46-61;
Heritage 1984: 245-60; Levinson 1983, ch.6; H. Sacks 1986; Schegloff
and Sacks 1973). The projectability of relevant next paired actions is a
quite elementary but perhaps fundamental mechanism for anticipating
alter's response (i.e., class of response), and hence is most important for
AIP.
However, in sequences of adjacently paired actions, only one move
ahead - the adjacent next action by speaker B, in response to the first pair-
part by speaker A - can be projected or anticipated. Whilst the conditional
relevance associated with adjacency pairs provides for strong constraints
controlling 'next position', it appears that the sequential organizations for
conversation are relatively weaker in terms of there being structures
which control or project actions beyond that.4 Research in conversation
analysis (and from other perspectives for the analysis of discourse) has
identified very many sequential patterns of three or more turns or moves:
but these patterns are, I think, not all equivalent forms of organization
with respect to participants' sequential management of talk, and in
particular to the projectability of future turns in the sequences.
This can be illustrated by considering three kinds of sequential
patterns which have been discerned in conversation.

Teases
The first is a pattern associated with 'po-faced receipts of teases', a
phenomenon in which recipients of teases overwhelmingly respond to the
tease in some serious fashion (Drew 1987). Even in cases where recipients
do respond to the humour in a tease, for example by laughing, they almost
always do so either as a preliminary to or in the course of making a serious
response.5 The following is an instance; Nancy has called Emma to tell
her about a man she met the previous evening at the home of a close friend
(Martha).
116 Paul Drew

Extract 1 (NB:II:4:14)
NANCY: But he's nixe looki:ng a::_nd ah just a rea:J_nice:
~=PERsonable, VERY personable, VERY SWEET, .hhh
VE:RY: (.) CONSIDERATE MY GOD ALL I HAD
TO DO WAS LOOK AT A CIGARETTE AND HE
WAS OUT OF THE CHAIR LIGHTING (h)IT
YhhhOU KNO(h)OW =
EMMA: =I:KNO:WIT
NANCY: .hehh.hh One of those kind .hhhhh =
EMMA: =Yes
NANCY: A::nd so: but we were
[
-• EMMA: THEY DO THAT BEFORE AND A:FTER
NANCY: [ [
eeYhhehee AHH
EMMA: THEY DO:n't.
NANCY: HAH HAH.hhh
-• NANCY: NO:?e-MARTHA HAS known Cli:ff5 ... ((a good 30 years
and he's an absolute boy scout))
In response to Emma's somewhat sexually laden tease 'THEY DO
THAT BEFORE AND A:FTER THEY DO:n't'5 Nancy begins~by
laughingly agreeing: but she then rejects the teasing proposal with
[No] + [Serious Account], cNO:?e-MARTHA HAS known Cli:ff
This phenomenon of serious responses to humorous teases (in only a very
few instances did recipients respond to the humour alone) is partly
accounted for by the sequential environment in which teases occur. In all
the instances I collected, in the turn(s) immediately preceding the tease
the person who is subsequently teased has been complaining, extolling/
praising or bragging in an exaggerated or overdone fashion. Space
prevents illustrating the extent to which in extract 1 Nancy had been
extolling the attractions and virtues of the man she met. But aflavourof it
is conveyed in her first turn in the extract; and this comes at the end of an
extravagant report of this man's comparative youthfulness (Nancy is well
into middle age), that he was a senior officer in the marines, that he has a
responsible job, his easygoing manner, his intelligence, how well they hit
it off and so on.
The sequential context in which teases occur is, then, typically one in
which the recipient has gone on extolling something or someone, has gone
on complaining or been self-pitying. The subsequent tease, conveying
some scepticism about what the teased has said, has a social control
function insofar as it may be considered a mild form of reproof for a minor
conversational transgression. The sequential pattern associated with
teases, and which is illustrated in extract 1, can be summarized thus:
 Interaction sequences and AIP 117

A: Exaggerated Extolling/Praising/Complaining
B: Tease
A: Serious Response

Preference for agreement

The second kind of pattern that I want to consider is associated with an
aspect of the organization of the preference for agreements with assess-
ments, and a dispreference for disagreements (Pomerantz 1984; see also
H. Sacks 1986). Pomerantz documented how the dispreferred character
of disagreement is evident in such actions being withheld or delayed: a
recipient may delay disagreeing with the first speaker's assessment either
by the design of a turn (by prefacing the upcoming disagreement with an
agreement component) or sequentially - or the recipient may withhold a
response. In the following two cases recipients withhold their response to
the first speakers' assessments.

Extract 2 (NB:III:6)
FRAN: Oh:::. Wih gee isn'at funny gee I'm going down t'see
somebody they're going do:wn the end a' this month et
twunty seven hundred- .hhh Ocean Fro::nt.
(0.7)
FRAN: IS that a^iiffrent plaxe then Newpo:rt?
TED: M-hm I gue:ss, this is, Balboa Penninsula.

Extract 3 (G.L.2: from Frankel 1983)

PATIENT: This- chemotherapy (0.2) it won't have any lasting effects
on havin' kids will it?
(2.2)
PATIENT: It will?
DOCTOR: I'm afraid so.

In extract 2 Fran's assessment is conveyed in 'gee isn'at funny', a

claimed coincidence that people she's going to be visiting are staying in, as
she thinks, the same vicinity as Ted. In extract 3 the patient's question
conveys an evaluation (by selecting 'It won't have ...', rather than asking
'Will it have ...'). In each case, following the pause in which the recipient
does not respond, the speaker (respectively, Fran and the Patient)
apparently recognizes that she had it wrong, and offers instead a revised
position, with which now the recipient straightaway agrees. Thus silences
may be oriented to as potential disagreements, with the interactional
consequence that 'It is not only that what would be a disagreement might
118 Paul Drew

not get said, but that what comes to be said may be said as an agreement'
(Pomerantz 1984: 77). A simplified version of the sequential pattern
which emerges from this orientation to recipient silence post-assessment
is:
A: Assessment
Silence
A: Backdown/Modified Position
B: Agreement

Pre-sequences
The third kind of sequence relevant to a preliminary consideration of the
possible relationship between AIP and sequential organizations will also
be familiar: it is conversational pre-sequences, such as pre-invitations,
pre-requests and such like. These are questions or enquiries which are
made on behalf of a next action, contingent upon the response to the
enquiry (Atkinson and Drew 1979: 141-8; Levinson 1983: 345-64;
Schegloff 1980). The following instance is particularly transparent, and -
what is most to the point here - is particularly transparent to the recipient
(Jim) of the pre-sequence enquiry.

Extract 4 (Holt:2:14)
JIM: J.P.Blenkinsop good morning,
(•)
SKIP: Good morning Ji:m,
(0.5)
SKIP: Uhit's Skip.
JIM: ffiyuh,
SKIP: YOU ^oming past the dooj:,
JIM: Certainly?
(0.8)
-• JIM: What time wouldju like the car Sah. =
SKIP: -Uh well ehhh hhehh hhhehh hhehh .hh^h that's
m:ost unexpected of you hhh::: n(h)o it's y(h)ery
nice'v you to offer huhh uh-heh heh-u-hu-.ehh
Thanks very much, .hhh
t
JIM: Eh:m I wz planning tih leave
here at just about twenty ...
The enquiry 'You coming past the doo:r,' is treated by Jim as a prefatory
enquiry, as leading up to something. His recognition that Skip 'wants
something' is particularly evident in his humorous response to the
enquiry: 'What time wouldju like the car Sah' humorously mimics what a
 Interaction sequences and AIP 119

chauffeur might ask - but it is also an offer (of a lift to work). As such, it
displays Jim's understanding that Skip's enquiry was made in the service
of requesting a lift, a request which does not now need to be made. Thus,
whereas the usual pattern or 'standard sequence' associated with pre-
sequences is
A: Pre-request enquiry
B: Response
A: Request
B: Granting,
here the sequence is attenuated in a way which commonly happens when
the recipient, recognizing what the enquiry is leading to, opts to address
the projected action directly (i.e. without waiting for the request to be
officially or formally made).
A: Pre-request enquiry
B: Offer
A: Acceptance
These then are three examples of the kinds of sequential patterns or
organizations which have been uncovered in conversation analytic
research. They are each an instance of what might be regarded as
'conversational routines' (Coulmas 1981a). But they have, I think, rather
different potentials for the role of projectability in 'anticipatory interactive
planning'. I have selected these to illustrate or represent, in a preliminary
fashion, some of the ways in which sequential patterns may or may not be
associated with conscious 'strategy', or the ability to plan for subsequent
moves in a sequence.
The first sequential pattern, associated with teases and po-faced
responses to them, does not appear to involve projectability at any stage in
the sequence. The exaggerated complainings, praising etc. only 'initiate'
the teasing sequence insofar as the recipient responds (sceptically) to the
overdone version. It is plainly no part of the design of the complaint/
praise to generate or provide the occasion for a tease. Likewise, the tease is
not designed to generate the third stage in the sequence, the serious
response/denial (indeed there's every reason to treat teases as designed not
to be taken seriously). And both the tease and po-faced response are
'backward-looking' responses to prior actions;6 they are not themselves
designed to project any next actions. So whilst there is a discernible
sequential pattern associated with teasing, no individual move in the
sequence seems to be part of any manifest strategic repertoire. At each
stage participants are oriented to a prior turn, and not to some projected
subsequent turn. The overall 'shape' of the sequence is not what either
was aiming to exploit.
120 Paul Drew

There are, I think, many examples of sequential patterns of this kind, in

which the overall organization of the sequence is not being oriented to by
co-participants. Their shape is discernible to the analyst; and of course
the regularity of the sequential pattern is the product of common
interactional problems or goals which participants manage, or find
solutions to, in typical ways/moves. But the project of participants is not
to design a current turn with an eye to any projected action in the
sequence. The sequential pattern is the product of a turn-by-turn
progression through a series of 'typical' interactional contingencies (e.g.
dealing with an exaggerated version of a complaint). It is not the product
of participants' mental representation of a projected sequence, and the
exploitation of a move within such a sequence.7
The other sequences illustrated above, associated with speaker modifi-
cations/backdowns post recipient withholding, and with pre-requests,
seem to be rather different organizations, inasmuch as they do manifest
participants' orientation to projectable next slots or moves.8 In the
former, that orientation is quite implicit; and the turn projected is that
adjacent next turn. In the latter, in pre-sequences, that orientation may be
explicit; it need not be explicit but, as extract 4 (above) illustrates,
participants can and do overtly display their understanding that the
business of a current turn is some projected action in the sequence. And in
this case the projected action is not restricted to the adjacent next turn: the
pre-sequence enquiry is designed - and recognized by the recipient as
designed - with a view to both the next-but-one turn (the request), and
the turn after that (the response to the enquiry indicating the likelihood of
the request being granted, and the making of the request being contingent
upon such a 'promising' response).
Pre-sequences seem then to sit opposite the kind of patterns illustrated
in teasing sequences. Pre-sequences are manifestly initiated with an
interactional goal in view, to see whether a request - if it were made -
would be received favourably, or (as in extract 4) to elicit an offer. And the
fact that the shape of the sequence - what the pre-sequence enquiry is
recognizably leading to etc. - is part of the speaker's conscious design or
strategy is evident in the playful way Jim exposes that design.
The pattern of speaker backdowns post recipient silences lies some-
where between 'non-projectable organizations' and the designedly stra-
tegic nature of pre-sequences. Although recipient withholdings/silences
are generated by the speakers' initial assessments, those assessments are
not, of course, designed to generate the sequence resulting in the
speakers' backdowns. And whilst the speakers' backdowns display their
recognition that the recipients disagree, we can only speculate that
recipients hold off from responding (remain silent) in the first place in an
 Interaction sequences and AIP 121

unspoken effort to get their co-participants to modify their position, and

thereby provide the opportunity, in a subsequent slot, for agreement. It
seems very likely that recipients do so: based on what Heritage (1984:241)
refers to as the symmetry between the procedures for the production of
verbal conduct, and for its interpretation, recipients may withhold a
response and hence 'produce' silences in order to have the speaker
recognize their (implicit) disagreement. Recipients may thereby simul-
taneously orient to the avoidance of a projected action (i.e., their disagree-
ment), and turn the sequence into one which will end instead in
agreement. But the evidence for this orientation/management lies impli-
citly in the sequential pattern.
The points which emerge from this very preliminary consideration of
three kinds of conversational sequences are, quite briefly, these. By no
means are all sequential patterns or organizations evidence of interactive
planning by participants. Whilst the orderliness of the pattern is pro-
duced in the first instance by participants, they may be acting in response
to interactional contingencies, without any view to the overall organiza-
tion of a sequence, or to any particular stage in it. What may be
characteristic of sequences which appear to be the product of partici-
pants' interactive planning is that the sequential organization is suffi-
ciently part of members' conscious linguistic/interactive repertoire that
future moves in the sequence can be projected, or anticipated: and that
such a projection can be exploited in order to pursue some interactional
goal. (Though note that that goal may not necessarily spring from
personal motivation or be strategically self-serving; it may be, as in
withholding disagreement, a matter of social 'preference', 'face', or as
Brown and Levinson have it 'politeness'.) By 'exploit', I mean that a
current action/turn may be designed so as to avoid or to facilitate a future
projected move in a recognizable sequential routine.
This introduces the possibility of speakers consciously exploiting
routines, where the routines or patterns are sufficiently recognizable and
predictable to participants for them quite explicitly to anticipate subse-
quent actions in a projected sequence. This brings us back to my earlier
remarks concerning whether sequential patterns may become the overt
objects of interactional mapping and planning, in the conscious attempt
to achieve sequential goals. In this respect Heritage (1990/91) dis-
tinguishes oriented-to procedures which he terms strategy (cog), and
procedures which are consciously employed, or strategy (cs). (In their
consideration of intentionality and goals in conversation, Mandelbaum
and Pomerantz (1991) make distinctions along similar lines; although
their terminology differs from that of Heritage, many of the problems and
themes being addressed in their paper are common to those in Heritage,
122 Paul Drew

and in this chapter.) The kinds of structural organizations which may 'fit'
behaviour, for example the sequence illustrated in extract 1 above, are
perhaps
'driven into the organism'. They are (or have become through experience) part of
its 'software' or even its neurobiology. Thus strategy (cog) is a property
predicated of an organism (or its program). It is not something available to the
organism's (or the program's) unaided inspection. Thus it is relevant to dis-
tinguish between a strategy (cog) which fits behaviour and a strategy (cs) which
guides behaviour. (Heritage 1990/91: 315.)
Comparing the pattern of speakers' modifications/backdowns in the face
of recipients' withholdings in extracts 2 and 3 with the pre-requests in
extract 4, one gets a sense of the fluctuating borderline to which Heritage
refers between a strategy (cog) and a speaker employing a strategy (cs).
The use of a withholding/silence in order to implicate (but not state)
disagreement is oriented to a sequential procedure, the evidence for the
consciously strategic use of which is only implicit (in a sequential
pattern); by contrast, the conscious use of a sequential routine is quite
manifest in extract 4, conspicuously so in the humorous treatment of the
pre-sequence enquiry.

Some aspects of conversational strategy

'Avoidance' strategy
Few cases of the conscious exploitation of an anticipated or projected
sequential routine will be as demonstrably clear as pre-requests. Most are
likely tofluctuatearound this borderline between (cog) and (cs) strategies,
that is in the space between the second and third types of sequential
pattern discussed above. I'd like to review some other patterns in which I
think it can be demonstrated that, at the least, a move is made in an
interactional sequence which exploits the anticipated development of the
sequence. Whether such sequential routines can be consciously or
mentally modelled by speakers, and hence whether a speaker is, on a given
occasion, consciously exploiting a particular routine in making a particu-
lar move, may sometimes be less easily demonstrable — as will become
evident, for good reason. Nevertheless, the following kinds of conver-
sational patterns are among those which come closest, I think, to speakers
exploiting what they anticipate will follow, in the sequence ahead, from
making a given move in a current turn.
The first arises from the conversational routine of greetings exchanges.
It involves an observation which Sacks made about telephone calls to a
Suicide Prevention Centre, and which, as Schegloff (1992b: xvi) recalls,
was a critical step in the development of what was to become conversation
 Interaction sequences and AIP 123

analysis (and it is the observation with which Sacks began his lectures
about conversation analysis (see H. Sacks 1992, Vol. I, lecture 1)). Sacks
noted that the emergency psychiatric hospital which operated the SPC
line was concerned about the regularity with which people who called the
centre would not give their names. And when Sacks looked to see 'where
in the course of the conversation could you [sic] tell that somebody would
not give their name?' (ibid.: 3), he noticed this:

Extract 5 (SPC)
(A is the SPC staff member answering calls, B is the caller.)
A: This is Mr Smith may I help you?
-+ B: I can't hear you.
A: This is Mr Smith.
B: Smith.
Sacks discussed this in the light of the procedure in greetings exchanges
whereby the form of address which is adopted by the first speaker may be
reciprocally adopted by next speaker. So that if the first speaker (the SPC
staff member) identifies himself by name, so might the other (the caller9).
Sacks observed about extract 5 and cases like it, that in the slot where a
reciprocal identification by a caller would be conditionally relevant, the
caller instead claims not to be able to hear. In reply to the caller's claimed
trouble, 'I can't hear you', the staff member repeats what he said. In the
sequence that therefore ensues - in which the staff member again gives his
name; the caller repeats that name, to check that he has it right; the staff
member confirms it and asks once more if he can help - the slot in which
the caller's reciprocal self-identification would go never occurs. The
caller has thereby managed to skip the move in which he might recipro-
cate with his name.

Extract 6 (SPC: from H. Sacks 1992, Vol.1: 7-8)

A: . . my name is Smith and I'm with the Emergency Psychiatric
Centre.
-• B: Your name is what?
A: Smith.
B: Smith.
A: Yes.
A: Can I help you?
-> B: I don't know hhheh I hope you can.
So there's a quite standard conversational routine, the procedure for
greetings exchanges, which provides for a slot in which the caller should
identify himself: it happens that for these callers (i.e., potential suicides)
124 Paul Drew

this slot is unwanted. The caller replaces that slot with an action (a claim
not to be able to hear) which generates an extended sequence in which the
slot in which he should make a return greeting with his name does not
recur.
The question is, though, whether the caller really cannot hear, or
whether he is using the claim not to hear as a strategy to avoid giving his
name. Well, there is a glimmer of evidence in B's repeat after A has
repeated his name. Certainly that's a regular thing to do after there's been
a hitch; you ask them to repeat their name, they do so, you repeat it to
check you have it right. But then there's the opportunity to latch onto
your repeat the reciprocal self-introduction: that is, the turn can be
managed in such a way that the opportunity to say who you are isn't lost -
if you want to say who you are. For instance, compare B's repeat in extract
6 with the caller's (erroneous) repeat of the receptionist's name in extract
7.

Extract 7 (Minne: Cmas:X:l-2)

CALLER: Who am I talking with. =
WENDI: = .hhhh This is Wendi the receptionist?
-> CALLER: Linda the receptionist. Linda?hh This is =
="((click ciick))
WENDI: Hello?
Evidently there is some difficulty about hearing in this extract (the caller
hears 'Linda' instead of 'Wendi', for example). But notice that having
repeated her name, the caller is then about to go on to self-identify with
'This is' before being cut off by some interference with the phone (the call
ends after Wendi's 'Hello?'). So there is no difficulty in latching a self-
identification onto a repeat if you want the other to know your name. This
isn't proof that the caller in the SPC data is consciously using a repeat
device as a strategy to avoid giving his name: but what happens, or does
not happen, later in a sequence is perhaps relevant internal evidence for
anticipatory planning. In this case the possibility is that he's projecting a
(repair) sequence, the completion of which will be co-terminous with the
end of the greetings exchanges; and hence the slot for his returning with
his name will have been 'missed'.
Another case in which, again, a speaker avoids doing something in a
sequential position where that action was relevant is illustrated in extract
8. This is from a call in which it is clear that on the coming weekend
Charlie is going to drive down to Syracuse, and has arranged to give Ilene
a ride. He's calling to tell her that, because the girl with whom he was
intending to stay will be out of town, he is not going after all.
 Interaction sequences and AIP 125

Extract 8 (Trip to Syracuse: 1)

CHARLIE: I^poke to the gi:r- I spoke tih Karen.
(0.4)
CHARLIE: And u:m:: (.) jt wz rea:lly ba:d because she decided of a:ll
weekends for this one tih go awa:y
(0.6)
-> ILENE: Whajt?
(0.4)"
-• CHARLIE: ^he decidih tih go away this weekend.
ILENE: Yea:h,
CHARLIE: .hhhh =
ILENE: =.khh
t
-> CHARLIE: So tha::t_yihknow Ideally don't have a place tuh
sta:y.
ILENE: .hhOh:::::.hh
ILENE: .hhh So yih not g'nna go up this weeken'?
CHARLIE: Nu-hJ don't think so.
One can notice, first off, that it isn't Charlie who says that he's not going
this weekend: the explicit version of the 'bad news' is articulated by Ilene
('.hhh So yih not g'nna go up this weeken'?'), much in line with
Terasaki's observations about pre-announcements of bad news eliciting
news deliveries from the erstwhile new recipient (Terasaki 1976; Scheg-
loff 1988b). There is, of course, a similarity between this case and extract
4, insofar as the pre-request in extract 4 worked to elicit an offer, without
Skip having formally to make the request: and here Charlie manages the
sequence so as to avoid saying that he is not going, but instead to put Ilene
in the position of figuring out and 'announcing' the bad news. The way
the sequence runs off, Charlie is left just to confirm the bad news ('Nuh:: I
don't think so.'). The speaker who initiated the sequence has managed it
in such a way as to avoid doing an unwanted, dispreferred action, i.e.,
delivering bad news.
But a more intricate 'negotiation' can be discerned in this sequence,
involving each of the participants avoiding making certain actions and
thereby extending the projected sequence. I'll give only a skeletal account
of this negotiation. Charlie's first version of the 'bad news' (in his first
turn in the extract) is only that Karen will be away; he doesn't state the
consequence this has for the arrangement he has with Ilene. In the next
slot, Ilene might have given some form of acknowledgement of that
upshot (e.g., along the lines of the disappointed 'Oh:::::.' she expresses
later). Instead she makes the kind of claim not to have heard, or the repair/
repeat device which Sacks noted the caller uses in extracts 5 and 6, when
she asks 'Wha:t?'
126 Paul Drew

CHARLIE: And u:m:: (.) jt wz reailly ba:d because she decided of a:ll
weekends for this one tih go awa:y
(0.6T
-> ILENE: Whajt?
(0.4)
CHARLIE: j>he decidih tih go away this weekend.
Charlie's partial repeat, 'She decidih tih go away this weekend', elects to
treat the problem only as one of hearing, and not that what he said needs
clarifying or elaborating (which might of course have put him in the
position of making explicit the consequence of Karen's being away, as in
'Karen decided to go away this weekend, so I'm not going up after all'). In
the next slot Ilene passes over yet another opportunity to acknowledge the
upshot of the news; she fills that slot with a continuer, 'Yea:h,', which
leaves Charlie back in the position of announcing his news:
CHARLIE: j>he decidih tih go away this weekend.
-» ILENE: Yea:h,
CHARLIE: .hhhh =
ILENE: =.khh
[
CHARLIE: So tha::t_yihknow l£eally don't have a place tuh
sta:y.
Charlie does now reveal a consequence of Karen's being away: but it's a
version in which the upshot for Ilene's ride still remains implicit, his not
having a place to stay being only another step towards the news that he's
not going. Charlie has thereby managed that slot to extend the sequence in
such a way that it's still left to Ilene, in her next slot, to state the upshot/
bad news.
In each slot after Charlie'sfirstturn, some acknowledgement or explicit
formulation of the bad news was possibly relevant. Each of them,
however, manages their turns in those slots so as to avoid performing that
action; each turn is designed to project a sequence in which it is left to the
other, if they will, to state or to acknowledge the upshot for their trip.

'Facilitating' future sequential moves

The emphasis in the cases considered above has been on participants
avoiding performing some action in a slot in which that action might have
been relevant, and extending the sequence in order to provide a slot for
the co-participant to perform an equivalent action instead. I want to move
to consider sequences in which a speaker designs a current turn so as to
generate or facilitate a subsequent action in the sequence. The distinction
is sometimes more a matter of emphasis: cases in this set, though, don't
involve the degree of avoidance that was evident in extracts 5 to 8.
 Interaction sequences and AIP 127

Perhaps the clearest instances of anticipatory planning are sequences in

which, like the pre-requests discussed earlier, a speaker designs a turn
which projects that, after recipient's response, another turn of a particular
kind will be forthcoming. This is often the case in story telling, which can
be initiated thus: A: 'Did I tell you about such-and-such?' B: 'No.' A:
'Well, [story].' A's first turn/enquiry is designed to project a subsequent
telling, though that telling might be contingent upon the recipient's
response. T w o kinds of environments with which these projected action
sequences are associated are what Schegloff has called 'preliminaries to
preliminaries', and stories.
Schegloff (1980) observed that enquiries such as 'Do you mind if I ask
you a question?' are, paradoxically, already questions; and that quite
regularly what follows (i.e., after the recipient replies 'No, go ahead') is
not a question, or at least not the question. This is one of the instances he
cites.

Extract 9 (BC: 12:18-19)

(A is the compere in a radio phone-in programme, B is the caller.)
-• B: I like tuh ask you something.
A: Shoot.
B: Y'know I 'ad my license suspended fuh six munts,
A: Uhhuh
B: Y'know for a reason which, I rathuh not, mention
tuh you, in othuh words, - a serious reason,
-> en I like tuh know if I w'd talk to my senator,
-» or- somebuddy, could they help me get it back,
Schegloff (1980: 106) notes about this and similar instances that there is a
kind of double displacement: 'Not only has the question that the speaker
plans been "displaced" by his "projection" of its occurrence, but it is also
not asked next.' Instead of the projected question, the thing that B would
like to ask, B reports something (losing his driving licence): and only after
that does he ask his question. Hence, B's initial turn projects an action
after the next (i.e., a question, after a reporting). This is no place to review
Schegloff's analysis of such 'pre-pre's' (but see Streeck, Chapter 4, this
volume), except to highlight his point that 'They serve to exempt what
directly follows them from being treated as "produced in its own right".
They make room for, and mark, what follows them as "preliminary"'
(1980: 116). What they are designed to achieve is the 'space' to give a
report as a preliminary to something else, i.e., the enquiry - and thereby to
restrict the recipient to not responding to the report as such, but to wait for
the projected enquiry. Note that after the first stage of B's report in extract
9, A restricts himself to merely acknowledging it, and letting B continue.
128 Paul Drew

Evidently, then, there is a symmetry between what the pre-pre is

designed to achieve, and the recipient's understanding of that task; that is,
his understanding that the subsequent action (the report) is not the one to
deal with, but that instead he should respond to the action after that (the
enquiry). The pre-pre is a shared inter subjective device for anticipatory
planning. In the pre-request in extract 5 the evidence for the shared
character of such a device was the humorous recognition by the recipient
of what was anticipated: in extract 9, the evidence is that the recipient
refrains from responding (other than with a continuer) until after the
projected action. Either way, the co-participants' mutual orientation to
projected action sequences is manifest in the recipients' behaviour, in a
manner which connects closely with the issues Goody raises about the
reciprocity of social relations in AIP.
Schegloff relates his analysis of 'preliminaries to preliminaries' to the
other environment associated with projected action sequences, that of
stories - and of course 'story prefaces'. These have been discussed
extensively in the literature of conversation and discourse analysis, and
are perhaps too familiar to need much elaboration here. Schegloff
summarizes their employment thus:
The story preface is a device by which a prospective teller can display an intention
to tell a story and yield a next turn to another, with the possible outcome that that
other will reselect the prospective teller to talk again, that is to tell the story, in the
course of which others will not treat each possible sentence/turn completion as a
point at which possibly to take a next turn for themselves. (Schegloff 1980: 146.)
We might note here that in 'displaying an intention', the speaker does also
intend to tell a story; so that the procedure, and the consciousness of the
procedure's purpose, are indistinguishable.
I want to show just one example, in order to highlight the point
concerning the shared orientation to projected action sequences: that is,
whilst the prospective story teller initiates the projected sequence with a
story preface, the recipient collaborates in the anticipatory planning by
holding off until the projected action has been completed. This is thrown
into relief by an instance where this goes awry.

Extract 10 (F:TC:1:1:18)
(The Warehouse is evidently a bar in which Shirley has a part-time job.)
SHIRLEY: . hhh Listen, something very very: cute happened
las' night at the Warehouse.
 Interaction sequences and AIP 129

GERI: What.
[
SHIRLEY: .hhhhhYihKNOW Cathy, (.) Larry Taylor's ex
girlfriend,
[
GERI: Yeeah.
SHIRLEY: jOkay._Cathy came in las'night.
(1.0) ~~
SHIRLEY: Whenever she comes in she always wants me t'do
something for her.
GERI: M-hm,
SHIRLEY: Either ^iddown'n ta:lk, whatever, .hhhhh So she
came in en she starts asking me if I'd seen Gary.
Gary Klei:n, .hhhh I s'd yeh he's here t'night
.hh she sz well wouldju go find him please'n tell
him t'give me my ten dollars thet he owes me,
-> GERI: Whaddiyou haftih get in on that for,
[ [
SHIRLEY: .hhhh Wai:t. I
started lau:ghing I looked at her en I said
((story continues))
It is fairly clear that Geri's 'contingent response' to the story telling, turns
out to have been premature. Having restricted herself to responses which
'fit' the course of the projected sequence, Geri responds (in her arrowed
turn) with the appropriate outrage to the report that Cathy asked Shirley
to run an errand for her, thereby treating that as the 'something cute that
happened at the Warehouse last night', and therefore the completion of
the story projected in the preface. Now as it happens she's wrong, because
that's only the beginning of the story. Shirley sanctions the premature
response ('Wai:t.'), and does not respond in turn to Geri's outraged
'understanding' of the story. Instead she goes on to recount not only how
she refused to get Cathy's money, but also a later incident in which Cathy
was discovered to be drinking alcohol (underage), to which Shirley was
deputed to put a stop. That's the real story, elaborately told over three and
a half pages of transcript, during which Geri's responses are restricted
once again to continuers.

Control
Thus story prefaces designedly 'control' a projected sequence (the story
telling) in which the recipient's speaking role is restricted. But notice that
this 'control' is only achieved through anticipatory planning on the part of
both participants, recipient as well as story teller. Geri's premature
incoming in extract 10 is not of course a deviant case (i.e., not an instance
of non-collaboration, or evidence that story prefaces don't control the
130 Paul Drew

projected telling). She comes in with her appreciation or understanding of

'the story' at just the point where what has been told fits, and is a possible
completion of, what the preface is designed to have her anticipate, the
'something cute that happened'. So the sequential control which is
achieved through story prefaces is collaboratively managed. Again, to
emphasize; the intentionality of sequential control is plainly an intersub-
jective, oriented-to property of story telling. For this reason, story
prefaces, as with other kinds of prefatory actions (such as pre-sequences)
are important evidence for AIP.
Finally, it may be worth noting in this context that moving onto or off
topics of conversation may involve some degree of 'control' over the
projected talk - and that this control can have a strategic character. This is
perhaps evident in such instances as these, in which a speaker opens up a
'new' topic, 10 in a fashion which has the appearance of a stepwise
connection with what was being talked about.

Extract 11 (NB:II:2:5)
(They have been talking about the assassination of Robert Kennedy a few
days before, and to which Nancy refers when she says 'everybody is
talking about it'.)
NANCY: Yeah_it's been a rough week an everbuddy is (.) youknow
(0.2)
EMMA: Mmhm
[
NANCY: _ta:lking about it en everbuddy: course Ij_don't
know whether it's that or just that we're js:t (.)
-> completely bo:gging down at work,h .hhhhmh
(•)
-• NANCY: Er whatta WIH: WITH ME^ with my finals?hhh
[
EMMA: Oh: well£ v r y buddy's sa::d
NANCY: hhuhuh:::
[
-• EMMA: Oh ho:w?d you do with yer finals.

Extract 12 (NB:II:4:10)
(They have just been talking at length about an operation Emma has had
on her toe, and whether they'll both go shopping.)
EMMA: A:nd uh I just am not gunnuh walk around a LOT
bec:u:z uh: Ah:::,hh (.) It's not worth it tuh be
 Interaction sequences and AIP 131

on my fee:t. Yih know

[
NANCY: Ya:h, h ri:ght.
(0
NANCY: Ah huh? .t.hhhhhh OH I WZ JUST OUT WA:SHING
windo:ws: a:nd uh my mother cabled so I ca:me in
I thought w'l while I'm here 'n I looked the
cloxk'n uz eleven thirty en I thought wul: (.)
ther .hhh .hh ther uhm (,) surely they're UP -•
yihknow I knew it wz kind'v a:sleep in da^:y_
[ [
EMMA: Yes Awee-
NANCY: but uh: I didn't get home till (.) .hhh two_last
-» night I met a very: very n:j_:ce gu:y.
In each case a new topic is opened, in extract 11 the university finals
examinations which Nancy has taken, and in extract 12 the announcement
that she met a man last night.11 But in each case the new topic is not
marked, disjunctively, as a new/next topic. It's led into in a stepwise
fashion; and in extract 11 the speaker designs her talk to build the new
topic out of what they had been talking about up to then. 'It's been a rough
week' begins as a reference to the Kennedy assassination: then she
speculates whether it's been rough because of the assassination, or
because they're bogged down at work - or because of her finals. Thus a
transition to the topic of herfinalshas been managed in a stepwise fashion
through the reference to a 'rough week'. Additionally, Nancy manages
that topic introduction in such a way that she can tell Emma about her
finals in response to an enquiry by Emma ('Oh how:'d you do with yer
finals'): that is, her 'listing' herfinalsas one of the things which has made it
a 'rough week' elicits (and may have been designed to elicit) an enquiry
from Emma.
Nancy's news about meeting a 'very very nice guy' in extract 11 is
similarly built out of a stepwise connection, consisting of an account of
how she came to call Emma (see Jefferson 1984: 195) - coming in to
answer a call from her mother, noticing the time, figuring that they'd be
up by now even though it's a 'sleep in day', through to 'didn't get home till
two last night', ^.s in extract 11, then, the new topic is contrived out of
other materials. The new topic is evidently one which the speaker really
wants to talk about (in fact, in extract 11 the news about the man she met
seems to have been the purpose behind suggesting a shopping trip/getting
together). And in each case the stepwise move leads to a pivotal or bivalent
segment, having the potential simultaneously to connect back and to
introduce the new topic (I adopt this notion of pivotal utterances in
stepwise topical transitions from Jefferson (1984)). So it's interesting that
'sleep in day' is selected as a way of describing the weekend (which is how,
132 Paul Drew

in overlap, Emma seems to be going to describe the day, but cuts off). It
connects back to figuring that they would be up by now, and forward to
'late night' (i.e., late rising after a late night - even if the late rising is
Emma's, and the late night was Nancy's). So too 'bogging down at work'
connects back to why it's been a 'rough week' for everyone, and provides a
warrant for introducing herfinals(that being a form of work, and a form of
labour which might be described as 'rough').
Parenthetically, it may be noticed that these pivotal segments, 'bogging
down at work' and 'sleep in day' have an idiomatic quality. This may not
be incidental to the use of idioms to 'legitimately' or accountably close
down a prior topic (Drew and Holt 1988). That is, they may be being used
here by a speaker who wants to close down one topic in order to open up
something she's keen to talk about.
At any rate, in the light of what the speaker goes on to talk about, it looks
as though the pivotal segments are designed to project a next and valued
topic. But the speaker thereby contrives to manage the introduction of the
new topic as a 'natural progression' out of a sequence of other materials.
This looks very much like a strategy for disguising just how keen the
speaker is to tell about it: to take a point from Sacks on invitations,
managing the introduction of a topic in this way displays the topic as
having been occasioned by other materials, and not as the reason for
calling.
Discussion: sequential patterns and 'conscious' strategies?
There can be little doubt that whilst structures of sequences of turns/
actions are part of the cognitive processes through which the coherence of
those sequences was produced in the first place in the course of interac-
tion, the mental models of such structures may not be conscious,
articulated resources. Sequential structures are some of the procedures
whereby co-participants discover the meaning in, and goals behind, one
another's utterances: that is, they are procedures which lie behind
participants' analyses of meaning/action. The adjacent next turn is a basic
structural position in interaction because it is there that participants'
analyses or understandings of what they took the other to be meaning/
doing are displayed (understandings which may be ratified or repaired in
the next turn after that, i.e., third turn (see ScheglorT 1992a)). But
although participants' analyses of one another's behaviour are certainly
part of the 'conscious' level of conversation, the procedures which lie at
the back of those analyses or understandings may not be.
Brown and Levinson's (1978: 90) caveat about their use of strategy
arises, I think, from just that distinction between participants' analyses of
one another's utterances, and the (cognitive) procedures which underlie
their analyses. Participants' understandings of what the other means, or is
up to, are indeed the conscious rational products of routines which
 Interaction sequences and AIP 133

themselves may not generally be subject to conscious modelling or

deliberation. The social intelligibility and perceived rationality of talk-in-
interaction is the product of the procedures - including sequential
structures - which are part of the learned programmes which underlie
social intelligence.
However, I have been reviewing some instances in which sequential
routines are perhaps being employed strategically - in its sense of the
conscious management of slots in a sequence whose organization can be
anticipated by speakers. It may be worth highlighting some aspects of
how I think this sense of conscious 'strategy' might be attributed to the
kinds of cases I have described.
It should be stressed that none of the sequential devices considered
above is intrinsically strategic. Sacks makes the following remark about
the use of 'I can't hear you' in extracts 5 and 6 above:
a device like 'I can't hear you' - the repeat device, providing for a repetition of the
thing that was first said, which is then repeated by thefirstperson who said 'I can't
hear you' - is not necessarily designed for skipping a move. It is not specific to
providing a way of keeping in the conversation and behaving properly while not
giving one's name. It can be used for other purposes and do other tasks, and it can
be used with other items. That's why I talk about it as an 'occasional device'. (H.
Sacks 1992: 7.)
Similarly 'what?', as in extract 8, pauses as in extracts 2 and 3, and
enquiries like 'What are you doing?', such as that in extract 4, may all be
used - indeed may generally be used - in other sequential environments or
to manage sequential tasks quite different from the possibly strategic
goals illustrated above. For example, just as someone may respond with
'What?' or 'I can't hear you' in circumstances where indeed they didn't
hear what the other said, so also someone might enquire 'What are you
doing?' just out of interest. It's not necessarily designedly a preliminary to
something else, such as an invitation.12
Strategy in talk exploits the ordinary properties of ordinary sequential
devices: thus the particular strategic work which a device may occasion-
ally do is done at an unofficial level (on this, and many of the other issues
being discussed here, see Mandelbaum and Pomerantz (1991)). The
official business of 'I can't hear you' or 'What?' is to repair a problem to do
with hearing (or possibly, in the latter case, understanding/clarification);
and such objects are responded to by recipients in terms of their official
business (for analyses of the 'official' and 'unofficial' work that the same
object can do in conversation, see Pomerantz (1980) and Drew (1984)).
The very matter of the talk being apparently organized in terms of the
official business of some sequential objects has the consequence that its
strategic but unofficial business may not show through on the surface of
the conversation — indeed, it may be properly designed not to intrude or
be recognized at that level (for the sakes of both participants). The
134 Paul Drew

ambiguity which results from an object's official and simultaneously

possibly unofficial business is a resource whereby participants can
(sometimes collaboratively) manage to disguise an object's strategic use.
It is for this very reason that finding evidence in linguistic analysis for
speakers' conscious strategies of anticipatory planning has proved so
intractable. If it is to be hidden from co-participants, it may remain
hidden from analysts.
Hence the rather special character of cases like extract 4, in which there
is explicit acknowledgement by the participants that the enquiry was not
'innocent' but was leading up to something. Mostly, though, the analytic
evidence that a turn has been designed to exploit other ('unofficial')
properties and the anticipated sequential consequences of those proper-
ties is much less easy to discern. In the kinds of cases reviewed here, my
best shot at what that evidence might be is that speakers exploit a
sequential routine, in which the slot for a given action can be projected in
the next turn, or the turn after that (i.e., the third turn). The anticipated
slot in the sequence may be one which is valued/wanted, or unwanted/to
be avoided. A current turn exploits such anticipated slots, by initiating the
routine sequence which will hence generate the projected and valued slot.
Alternatively, in cases in which the projected slot is unwanted/to be
avoided, there is evidence that the sequence is being 'misshaped', by
being either expanded or attenuated. In expansions, the next turn is a slot
in which the projected but unwanted action is relevant: the speaker avoids
that action by instead filling that slot with an alternative action, the
occurrence of which is occasioned by different and, thus far, unantici-
pated relevancies or contingencies. In this respect, troubles (in hearing
etc.) are especially exploitable, because troubles are quite freely occurring
in conversation; they may inhabit any turn-at-talk, and hence a 'next turn
repair initiation' may always be mobilized in order to avoid doing
something else (the unwanted action) in that turn.
But devices for avoiding relevant actions in the next slot are not
restricted to repair devices. Extract 13 is an instance of something which
may be quite frequently done: a speaker makes a sequential move which
avoids an inauspicious environment, and instead delays a planned next
move until a more auspicious environment can be produced. This extract
illustrates just such an avoidance/delay: Emma's invitation to Nancy for
lunch follows a pre-sequence enquiry 'What are you doing?' (not shown).

Extract 13 (NB:II:2:14)
EMMA: Wanna come down'n av a bite'a^unch with me:? =
I got s'm bee:r en stu:ff,
(0.2)
 Interaction sequences and AIP 135

NANCY: Wul yer riljsweet hon:, uh::m

(0.9)
NANCY: l e t I: ha(v)
[
-* EMMA: or d'yuh'av sump'n^lse (t')
[
NANCY: NO:, I haf to uh call RoPs
mother, .h I told'er I:'d ca:ll 'er this morning-1
gotta letter from 'er en (.) .hhhhh A:ndum
(1.0)
NANCY: p.SO sh- in the jetter she said if you can why (.)
yihknow call me Saturday morning en I just
haven't_hh.
EMMA: Mm hm
[
NANCY: .hh It's like takin' a beating. (.) mhh heh.
heh heh hh.
[
-> EMMA: Mm:: 'N yuh have'n heard a word huh.
((Nancy then reports and talk continues for some while
about her ex-husband's failure to get in touch with her or
any of the family.))

There are two positions in which Emma forestalls a rejection, or rejection

implication. The first is when, after Nancy's appreciation and then
evident hesitation, Emma anticipates the possibility of an upcoming
rejection, and heads it off with 'or d'yuh 'av sump'n else' (it might do so
only in the sense of now providing a slot for an account, in place of a slot in
which Nancy might have been going to reject the invitation). The second
occasion is her response to Nancy's account of what she has to do, and
Nancy's complaint that 'It's like takin' a beating'. In that slot Emma
might have displayed an understanding of the 'bad news' implication (as
did Ilene in the previous extract) of the account for her lunch invitation,
for example through a marker of disappointment. Instead Emma focuses
on Nancy's ex-husband and what he's not doing - "N yuh have'n heard a
word huh.' - rather than on the call to her mother-in-law. Emma thereby
avoids making explicit any upshot, for her invitation, of Nancy's having
to make a call. In other words, by picking up on that particular
implication of Nancy's account Emma manages to divert the talk away
from the invitation (opening a 'new' topic, rather in the manner that
Nancy did in extracts 11 and 12), until another and more auspicious
opportunity arises to re-introduce her invitation - which she does several
minutes later (for an analysis of which see Drew (1984)).
Sequence expansions therefore appear to be one means for avoiding
unwanted actions. Sequence attentuations appear to be rather less
136 Paul Drew

common; indeed they might be restricted to sequential objects like pre-

sequences, in which the pre-sequence enquiry or noticing is designed
with an eye to the possibility that the recipient will make an offer, thus
absolving the speaker from the necessity of having to make a request.
Hence, the slot in which a request might have been made, after the
recipient's response to the pre-request, is displaced by the recipient
replying right away with an offer.13 Of course such attenuations cannot be
guaranteed; they rely on co-participants' collaboration, and that is not
always forthcoming.
Sequence initiations, attentuations and expansions do not by them-
selves constitute evidence that speakers consciously anticipate some
future move/action, and attempt to facilitate or avoid that projected
move, depending on whether the action is valued or unwanted. The
advantage of treating sequences and sequential slots/objects as procedures
is that it avoids mentalistic attributions; and it avoids the seemingly
intractable analytic task of deciding whether a speaker 'knows' that by
doing something in this slot, they'll avoid having to do something else in a
subsequent slot.
The disadvantage is that any sequential object can be regarded as a
procedure: and so the distinction between kinds of sequences that I was
trying to make at the outset of the chapter would count for nothing. All
sequential patterns would be equivalent. Some routines are perhaps part
of the vernacular: by contrast, the organizations of other sequences are
not projectable by participants. So this disadvantage of referring only to
'procedures' folds into another, that that may fail to capture the projec-
table, anticipated moves which are contingent upon a current move.
When, in designing a current turn, a speaker selects which activity shall
go in that turn - in many of the cases reviewed above, departing from and
expanding the 'standard' sequence - that selection must involve a degree
of'consciousness'. There seems no reason in principle to deny a commen-
surate degree of consciousness to constructing a current turn with an eye
to a subsequent anticipated turn. Of course it's not the principle but the
empirical substantiation of conscious strategy in some of the cases
reviewed here, and in other interactional sequences, that causes us to
vacillate between regarding them as empirical instances of procedural,
programmable cognitive strategies, to the possibility that they are
instances of speakers' conscious strategies in anticipating how to achieve
certain outcomes in projected sequences. That empirical substantiation
may involve the kinds of issues I have been discussing here concerning
sequential initiation, attentuation and expansion.
 Interaction sequences and AIP 137

Notes
1 This is close, I think, to Byrne and Whiten's middle level (level one) of
'intentional behaviour' in animals, which includes 'behaviour that is convinc-
ingly intentional, in the sense of goal directed, ... [although] the category is
agnostic as to the animal's mental states ... [Hence] to qualify as tactical
deception, an action must therefore indicate at least Level One evidence of
intention to achieve a goal which can only be reached if an individual is
deceived {not the same thing as "evidence of an intention to deceive")' (Byrne
and Whiten 1991). The agnosticism of this category as regards the intention-
ality of action (for example, to deceive another) parallels that of 'oriented to
procedures' in conversation analytic research. See for instance Heritage
(1990/91).
2 Levinson (Chapter 11, this volume, n. 13) discusses just such a level of
(self-)consciousness about the practice of handshaking during greetings and
partings, in circumstances where interactants are from different cultural
backgrounds. But such moments as he describes of indecision and misinter-
pretation are equally familiar in exchanges between people who share a
common culture, but are not sure which of the candidate greetings routines
they should adopt.
3 More complex sequential patterns are perhaps less likely to be consciously
managed because knowledge about them may generally not be part of the
vernacular. (For some observations on this, see Jefferson (1988a:439).)
4 Sequences in which there is a measure of control over the turn-after-next (i.e.
the third turn) are more familiar in institutional settings; for instance in
classroom interaction, participants' orientation to the 'correctness' (or other-
wise) of answers generates sequences in which teachers respond to students'
answers to their questions by evaluating those answers, in a third turn.
5 These findings are for a study of English (American and English) speakers.
The anthropological literature, some of which is cited in Drew (1987),
suggests that teasing in other societies generates different patterns, and that
therefore such patterns are cultural forms. Esther Goody (personal communi-
cation) reports that in Gonja, teases have to be responded to in a joking fashion.
I want to note, however, that the same might be said of teasing in English/
American culture: that is, there is a normative orientation to 'humour' as the
proper response to teasing, even though on many or most occasions people
who are teased 'fail' to respond as they should. So, for example, the laughter
which regularly precedes the po-faced response, as here in Nancy's initial
response in extract 1, is certainly evidence that participants orient to norma-
tive expectations, in advance of flouting them. Having not studied the details
of the design of responses to teases in Gonja or in other cultures, I am not
qualified to say whether anything similar happens (e.g., that the Gonja display
an orientation to a proper joking response, whilst simultaneously defending
themselves in a 'serious' fashion. But I am yet to be convinced that teasing and
conventionalized normative responses to teasing are indeed a cultural form.
6 They are 'backward looking' in a less obvious sense, that of 'recasting' the
character of the prior turns: so that the tease recasts the prior turn as having
been exaggerated, and the po-faced response treats the 'joking' tease as having
138 Paul Drew

had a serious import. Hence there is a retroactive quality to the manner in

which turns in the sequence 're-interpret' prior turns, as discussed by Good in
his contribution to this volume (Chapter 6).
7 Perhaps the clearest and most elaborate example of this kind of 'non-strategic'
sequential pattern is the six-stage sequence which Jefferson describes for
troubles telling in conversation (Jefferson 1988b). She refers to this pattern as
a 'candidate sequence' because it is an artificial construction of a sequence
which empirically never runs off in the precise order of the template she
documents. The disorder is generated by the troubles-telling sequence being
'constantly encroached upon, and recurrently breached, by the pressure
towards business as usual, to which talk about trouble seems irrevocably
vulnerable, and to the concerns of which a "trouble" appears to be irremedi-
ably subordinate and accountable' (Jefferson 1988a: 440). Whilst progression
through the troubles-telling sequence is therefore susceptible to participants'
orientation to a quite general interactional principle (business as usual), that
principle is embodied in the design of a given turn without being strategically
mobilized with a view to (facilitating or avoiding) any actions later in the
sequence.
8 In this respect it may be no accident that in their chapters in this volume both
Streeck and Levinson also cite instances of pre-sequences in the context of the
projectability associated with sequences.
9 This is illustrated in the greetings which precede the pre-request for a ride in
extract 4 in a rather interesting way. The person who answers gives his name,
rather formally; the caller then gives his name - but not before re-addressing,
as it were, 'J.P* Blenkinsop' as 'Jim', for which an appropriate reciprocal
diminutive, 'It's Skip', is used.
10 1 am leaving aside for present an issue of considerable importance for AIP
generally, and specifically for the possibility of the consciousness of AIP
strategies; this is the matter of how some topics, especially sensitive ones, are
initiated or introduced into the conversation. The recipient's design of topical
initial utterances which include explicit reference to such matters as what has
happened since we last spoke, or accounts of why one is calling or introducing
this topic (on which, see especially Jefferson (1984)), might be analysed for
evidence of a speaker's conscious planning of how to introduce the topic.
Wootton (personal communication) suggests that offer sequences frequently
involve some manifest planning on the part of the one making the offer, in the
face of the delicacy of what is offered.
11 This is the man Nancy is describing in extract 1.
12 However, my guess is that the particular details of the design of 'What are you
doing?' enquiries which are disinterestedly interested are probably different
from the design of the 'same' enquiry when it's intended as a preliminary to
something; and that from these design details, recipients have little difficulty
infiguringout which sort of enquiry it is. Whether that's true of other devices
such as 'I can't hear you' and other repairs we do not yet know.
13 Elsewhere I discussed a similar phenomenon in cases where, instead of
inviting the recipient, the speaker uses a device, a form of 'reporting', which is
designed to elicit a self-invitation by the recipient: similarly, reporting a
difficulty may be designed to give the recipient the opportunity to offer
assistance. (See Drew 1984.)
 DAVID GOOD

6 Where does foresight end and hindsight

begin?

There is much to be said in favour of the idea that anticipation and the
preemption and deception which it permits play an important role in the
social life of both humans and a number of other primates. This idea has
been developed anew, particularly for the human case, in a number of
different disciplines for a variety of different reasons. Goody (this
volume), points to a number of instances, and the list could be extened
almost ad nauseam. It ranges from computer scientists proposing plan-
construction algorithms and user-modelling procedures for natural lan-
guage front-ends, to economists speculating on the actions of economic
agents planning their moves in the market-place with respect to the
actions of others. The scope and explanatory range of this idea in any of
these fields has always been questionable. In economics, for example, the
very different effects of ignorance, habit and culture on individual choice
have often been far greater than those of individual ratiocination.
Nevertheless, any area of human activity which involves actual or
potential interaction with other humans (and that might mean all of
human life) can be seen to have characteristics which reveal an anticipa-
tion of how others will view and respond to one's actions.
If we restrict our consideration of anticipation to human conversations,
many writers have proposed that they have a measure of foreseeability
built in. For example, conversation analysts have argued that there are
clear expectations as to what kinds of turn may follow other turns, as in
adjacency pairs, and participants must mark utterances, which do not
conform to the expected trajectory of the conversation in some way so that
their marked or dispreferred status is signalled (see, for example, the
discussion of their work in Levinson (1983), or the papers in Atkinson and
Heritage (1986)). Alternatively, if one follows the line of most speech-act
theorists, all human languages have a number of resources which enable a
speaker to stake a claim for a particular future interpretation of what is
being said at any one point. The essential difference between an utterance
in which there is an explicit marker of illocutionary force, and one which
might be characterized as an indirect speech act, is that, with the latter,
the precise nature of the speaker's intent is more open to negotiation. For

139
140 David Good

both traditions, and one does not need to be an adherent of either to

appreciate this point, the speaker's interactional and linguistic resources
are marshalled to make claims on the future and to achieve the kind of
control which, as Goody (this volume) points out, is vital if anticipatory
interactive planning (AIP, henceforth) is to be of value to the individual.
AIP clearly requires substantial intellectual resources, and it is not
surprising that social life has been offered as an important part of the made
environment which has been important in the evolution of primate
intelligence. Much interest has been generated by the similarities
between the human and non-human cases, but in considering the final
steps in this evolution, it is important to understand the qualitative
differences in social interaction which are made possible by the availabil-
ity of human language. These will be crucial to an understanding of the
role of culture in social life, the nature of human intelligence, and,
conceivably, to an analysis of the evolution of language.
There are many differences between the human and other cases as the
other contributors to this volume demonstrate, but I wish to focus on just
one which I believe is amongst the most important, and possibly even the
most important. This is the way in which human language transforms the
nature of time in human social life. I will argue that it achieves this
through the ways in which it permits description and redescription of past
actions, a capacity which is bound up with the fact that it also permits
actions which are open to many interpretations, and which can be revised
in the course of production. Thus, retrospective re-appraisal becomes as
important as prospective planning, foresight becomes focused on what
hindsight can do. The properties of AIP and the actions to which it would
lead are constrained by the prospect of this revision, and the cognitive
resources required for successful performance become different and
much greater. There are a number of reasons why I believe this to be so,
but in this brief sketch, I will restrict myself to those which I take to be
central to the argument, and briefly consider their implications.

The interactional arms race

Long range tactics

If the ability to understand enduring social relations, to predict with a fair
degree of success what another individual will do and, as a result, to
deceive, can provide a distinct reproductive advantage, then it seems
reasonable to propose that the next level of anticipatory and interactive
sophistication will lie in the ability to discover the deceiver or free-riding
exploiter of these social contracts. An advantageous next development for
the deceiver would then be to find some way of minimizing the negative
 Where does foresight end and hindsight begin? 141

consequences of a deceit being discovered. There is a greater potential for

this when the original act in question is ambiguous, and that ambiguity
can be addressed. As a result an individual could disclaim full responsi-
bility for a deception, by characterizing it as a misunderstanding. This
could minimize the harm which might follow from the truth being
discovered, and maintain the individual's potential for future actions in
that relationship.
The progression from simple referring expressions which stand for real
world entities, as would seem to be the case, for example, with vervet
monkey calls, to human language facilitates this process so that ambiguity
and tentativeness become possible, as does subsequent denial. Take the
case due to Seyfarth reported in Dennett's discussion of the 'intentional
stance' (Dennett 1988:189). One vervet monkey, when involved in a
skirmish between its own troop and another, issued the 'leopard call'
which led both bands to take to the trees thereby creating a 'cease-fire' and
apparently saving the day for its own troop which had been losing the
fight. If a similar deception were used in an intra-group setting in an
attempt to gain personal advantage, though, if discovered, the animal
would presumably suffer. If, however, it could utter its cry in a way which
indicated a possibility, not a certainty, of a leopard, that is, a measure of
doubt, then if subsequently discovered to have misled other individuals,
it could defend itself by elaborating an account of both its motives and the
basis of its perception; it then has a basis for avoiding retribution. Such an
indication of doubt might ultimately lead to the cry not being completely
effective in causing all vervet monkeys in the vicinity to run for the trees,
but the cost of scepticism on their part would be so high that it is unlikely
that they would be. Therefore, any vervet monkey that could indicate an
uncertainty about the truth-value of an assertion as to the presence of a
leopard, and could subsequently exploit this indication as an excuse,
would have made the next important step in the interactional arms race. It
would also have uttered thefirstproto-question, and been thefirstto offer
a pragmatic variation in meaning by linking the meaning of a sign to a
property of its user. Esther Goody (1978a,b) has discussed the singular
importance of questions in a way which is redolent of this speculation, and
I will briefly return to this point later.
In the human case, deception and the breaking of promises, be they
explicit or implicit, is much easier when the nature of the original
contract, either implicit or explicit, is less than clear, because then it can
be claimed that in fact no one is being deceived or dishonoured. A point
which Brown and Levinson (1987) make so well in their analysis of face
and politeness, and which Arundale (1993) has interestingly extended.
Despite our usual perception, that what is said in a conversation is
relatively clear, the basis of that clarity relies as much on an inarticulable
142 David Good

set of background assumptions as it does on what is said. It is these which

make it possible for ambiguity and uncertainty of meaning to be claimed
at some future point in that conversation, or in a future engagement, and
an alternative meaning attributed to what was said or done. The specific
way in which this might be exploited has been considered by Weiser
(1975), who described it as 'deliberate ambiguity', and is what Brown and
Levinson (1987) refer to as an 'off-the-record' strategy.
At other times, of course, speakers seem to be forced to extremes of
commitment to the words they have uttered. Sometimes, they are held to
the very letter of what they said where that 'letter' is interpreted in a very
specific way. Yet on other occasions they are allowed to disclaim responsi-
bility for what was said. In a Humpty-Dumpty-like fashion they can
proclaim themselves to be the masters of what their words meant, usually,
via an argument that what was said did not reflect their true intentions
because the context in which it was said was poorly understood, bizarre,
or not the one which currently holds. Machiavelli was well aware of this
when he observed that, 'A wise ruler cannot and should not keep his word
when such an observance of faith would be to his disadvantage and when
the reasons which made him promise are removed... Men are so simple-
minded and so controlled by their present needs that one who deceives
will always find another who will allow himself to be deceived.' (The
Prince, ch. 18.) Each of these extremes is, nevertheless, dependent on the
compliance of others, and it would be wrong to ignore the role of the
relative power and positions of the speakers and hearers in determining
which of these extremes is adopted. A point of which Machiavelli was
very aware.
The idea that such uncertainty and under determination exist has been
challenged by that part of the speech-act tradition which has followed
from Searle's (1969) interpretation of Austin's (1962) original work. As
most will know, Searle proposed that it was possible to specify the full set
of conditions under which a particular speech act was felicitous. If the
speaker knows that these conditions are satisfied, then the act is guaran-
teed to go through. The felicity conditions have been re-interpreted in a
variety of ways, most interestingly, given the current topic, by workers in
artificial intelligence (AI) who have seen them as akin to the steps in a
plan-construction algorithm which can be used as the basis for the
production and interpretation of utterances.
In attempting to fulfil Searle's ambition, workers in this tradition have
come to realize that Austin's original characterization was correct, and
that the task of specifying all that must hold for any particular speech act
to be felicitous is never-ending. Essentially, Austin focused on the ways
speech acts fail and are thus infelicitous, and recognized that these ways
form an open-ended set. This focus undoubtedly reflected the legal aspect
 Where does foresight end and hindsight begin? 143

of Austin's philosophical background. Law-makers must rely on the

judiciary to interpret their intention in making a statute, and then,
through the application of that statute in ever new and unforeseen
contexts, the courts provide an ever-growing corpus of cases which refine
and redefine the meaning of that intention. This process is never-ending,
as technological developments in areas to do with human life repeatedly
demonstrate. The speech-act theorist and the speaker are in the same
position as the law-maker and can never foresee all the contexts in which
an utterance might be interpreted. Thus a specific interpretation can
never be guaranteed, nor can the results which might follow from that
interpretation. To guarantee an interpretation would require the indivi-
dual to have a level of foresight which we do not seem to possess even at
the level of the collected minds of the law-makers and the judiciary.

Medium-range tactics
In human conversation, the meaning of an utterance is not only revisable
some time later, but is often revised at the time of production. As has often
been observed, the individual speaker is not the sole arbiter of the
meaning and significance of what he or she says. At the turn-by-turn level
of organization, many have argued that what each speaker says offers,
amongst many other things, a demonstration of his or her understanding
of what the previous speaker has said. In so doing, the current speaker
contributes to the definition of the meaning of what was said by that
previous speaker. Consider an example taken from Heritage (1984): if, in
response to 'Why don't you come and see me some time?' I reply 'I'd love
to', then I have offered a reading of it as an invitation. If I reply, 'I'm
sorry, but I just never get the time these days', I have offered a reading of
it as a complaint. Now, the original speaker might not accept one
interpretation if the other was intended, and he or she can seek to redefine
the original utterance in his or her next turn, but if that is what happens
the character of the exchange overall is necessarily altered. A misunder-
stood complaint which is explicitly repeated has a far greater interactional
salience. Also, the point that each has taken a role in the definition of the
meaning which is thereby negotiated is reinforced.

Short-range tactics
This process of negotiation which can occur in the production of a
sequence of utterances suggests that the meanings derived are effectively
a joint production. There are good reasons to believe that this process
operates at the intra-utterance level too.
Necessarily, the content of an utterance is revealed gradually. Goodwin
144 David Good

(1979) has shown that speakers are sensitive to the reactions of others
during their turn at talking, and that they may revise what is being said as
a consequence of these reactions. He proposes that utterances are
constructed interactively as joint accomplishments, and this view is also
supported by work on within-turn self-initiated, self-accomplished
repair (see Streeck, Chapter 4 and Drew, Chapter 5, this volume).1 Also,
it is not implausible to suggest that as an utterance is being produced, not
only can it lead to recognizable reactions on the part of the hearer, but also
a more complete recognition by the speaker of the significance of what is
being said for that hearer. It is a common intuition that as soon as an
utterance is said, it is much easier for the speaker to understand the
hearer's likely reaction even if that reaction cannot be seen or heard. Both
kinds of feedback could then lead the speaker to revise the plan for what is
to be said next in the same utterance.
It is unclear how one could discern what limits there are on how often
this revision process can happen, but if the work of Goodwin, Drew,
Streeck and others is to be believed it can be rapid, and at any point,
although as the work of G. Beattie (1983) illustrates, how often this will
happen depends on the other cognitive and interactional demands the
speaker is facing.
This potential for revision suggests that there is a progression through
the production of an utterance in what is being anticipated. It also
provides a basis for allowing the ambiguity, which is necessarily present
in an incomplete utterance, to provide a test for the acceptability of what
might be said. Thus, what is said towards the end of an utterance can
revise what was said earlier so that the whole may end up as rather
different to that which was initially projected by the speaker. A classic
example of this is sentence-final negation in Japanese which permits the
reversal of the truth value of an utterance with the final morpheme (R.
Miller 1967).

Social life, chess and language

To understand fully the implications of the observations in the previous
section, it is useful to consider the metaphor which underlies much of the
thinking about non-human interaction, and the crucial feature of that
metaphor which allows the division of time with respect to anticipation.
If it is being proposed that some agent is anticipating some future
event, then for an adequate characterization of that anticipatory skill, it is
necessary to define the period during which it is employed. This is a
logical and theoretical necessity. An account of anticipatory planning
must be based on what is known before the event, and this is temporally
defined. The most important limit to this period is the latest point at
 Where does foresight end and hindsight begin? 145

which any anticipatory planning can be done, which is to say, just before
the event or action in question occurs: a point in time which we might, for
convenience's sake, refer to as the 'anticipation limit' (AL, henceforth).
What, at one moment before the AL, is anticipation becomes reaction as
the anticipated event or action occurs, the AL is passed, and different
information becomes available. Subsequently, further (perhaps differ-
ent) anticipatory planning will occur, but by definition it cannot be
anticipating something which has already happened.
If we examine a game such as chess, the ALs are easy to define. They are
the points at which each move is made. Before these points, each player
speculates on what the likely intentions behind, and outcomes of his or
her opponent's moves, the likely outcome of his or her own moves, and
how his or her intentions might be perceived by his or her opponent.
There is a period of reflection and planning, followed by a discrete and
well-defined action, followed by a further period of reflection and
planning as the opponent's next move is awaited. The actions, which are
well-defined, chunk the interaction by specifying the beginning and end
of every turn, and thereby the ALs. The time it takes to move a chess piece
is seemingly irrelevant and can be ignored in the same way that the period
of time for which a force is applied in an impact is ignored in classical
physics.
Humphrey (1976) explicitly uses chess as the model for interaction, and
in many respects this seems quite appropriate for the cases he discusses. It
is tempting to think that the same structure is to be found in conversatons,
and that each turn at speaking is similar to the sequence of plan and move
in chess. Some psycholinguistic work has tried to exploit this metaphor,
but the results of the related empirical work suggest that the proposed
cognitive cycles are, at best, somewhat ephemeral (see G. Beattie (1983)
for the best attempt). Such conceptualizations have been notoriously
difficult to operationalize or support in any other way. Elsewhere (Good
1989), I have argued against the idea that conversations are made up of
well-defined units that are like chess moves in their character, and the
argument above that ambiguity, redescription, and joint production
pervade every level of conversation renders the idea that an AL can be
specified untenable.
As if this were not enough, the difficulty in defining an AL for a single
turn is exacerbated when we contemplate the fact that many ambitions or
goals in a conversation might be accomplished across a number of turns.
For example, Brown and Levinson (1987) in their analysis of politeness
propose that speakers take redressive action to ameliorate the face-threat
offered by any utterance within that utterance. While this might be true in
some instances, and is certainly a sensible simplifying move for their
analysis, it is common to find that some kinds of redressive action are
146 David Good

accomplished in more turns than the one which contains the specific face-
threatening act. For example, if one is adopting a positive politeness
strategy by using various markers of common group affiliation, such as a
Basque speaker palatalizing his or her consonants (Corum 1975), it is
singularly odd if the marker, in this case palatalization, is 'turned on' for
just the utterance which contains the face-threatening act.

Implications
In the preceding paragraphs I have raised various points which, if fully
supported, will require a different perspective on the nature of anticipa-
tion and forward planning when the individual is using a human lan-
guage. The thesis I would propose is that given our imperfect knowledge
of one another, the absence of a clear point at which anticipation may have
a clear articulation, and the potential ambiguity of our actions which
permits subsequent revision of their significance, the major concern in
anticipating the social future is that the past may be re-interpreted.
Therefore, the concern of the individual will be to constrain those future
interpretations since no specific guarantee can ever be offered for the
accomplishment of a specific reading. The best we can do is to take action
to eliminate certain interpretations, and hope that the set of possible
interpretations which remains provides sufficient scope for casting our-
selves in a decent light in the future. The crucial thing for foresight then
becomes the knowledge of what hindsight can do. This thesis has
implications for our view of the role of culture, and the nature of the
intelligence which we have as a result of our evolutionary history. I will
conclude by speculating on these very briefly.

Culture, Machiavelli and Hume

The interpretation of past acts and the role of various actors in them is the
stuff of courtroom disputes and deliberations. While barristers are adept
at offering radical recastings of events so that their clients' or opponents'
actions are seemingly transformed, there are constraints on what they can
do. The same is true for all other speakers too. It is not the case that any
behaviour or utterance can be taken to mean anything the barrister
chooses. There is an accepted scheme of things, as in, for example,
Burke's Grammar of Motives, within which the discussion of past action
operates. A straightforward appeal to demonic possession as an account of
a person's actions will not be accepted in the same way today as it was in
sixteenth-century Europe. There is a culture of the courtroom which is,
in many ways, a crystallization of the mores of the wider community, and
which both facilitates and limits these re-interpretations. Similarly, the
 Where does foresight end and hindsight begin? 147

wider culture operates in everyday negotiations as to the significance of

what is said and done. These constraints can also be seen as facilitators of
everyday reasoning, and in many respects ease the cognitive burden
which the individual in an uncertain ambiguous social world must
shoulder.
The burden which is carried will almost certainly depend on the
individual's ability to understand the interplay and coordination of one's
own and others' actions. As Levinson (Chapter 11, this volume), rightly
argues, Machiavellian competition is child's play compared to the
Humean task which coordination produces. Again, the general expec-
tations and representations of each other made available through our
culture ease this task, but do not eliminate it.

Memory and self

To operate in this rather free-form ambiguous and ever-changing inter-
actional environment obviously requires substantial cognitive skills, and
it is tempting to think that, given my suggestion as to the role of hindsight,
the nature and extent of the individual's memory will be crucial. There is
much debate in the memory literature as to the nature and role of the
different kinds of memory which have been proposed. Short-term stores
of one sort or another, long-term memory, recall and recognition
memory, and so forth, have all been examined and seen to have greater or
lesser roles in language use. At the turn-by-turn level of revision and
correction, a short-term store will be required, and this will arguably be
different to that which is used directly in the production process. The
speaker must orient to what he or she has said in the same way as the
listener has if the need for repair is to be recognized, and this would seem
to depend on the kind of memory which is tested in short-term memory
tasks. The major problem when engaging in self-repair is to ensure that
the target of the repair is recognized. The further away from the repair
that the target is, the harder it is to ensure that the repair is linked to the
repairable. Beyond a certain distance, this becomes a very chancy affair. It
is tempting, therefore, to speculate that at least some characteristics of
short-term memory have evolved because of the demands of coordination
in utterance production and the limits on successful reference in accom-
plishing self-repair.
At a different level, a memory of self and the description of events in
which one has been engaged, with self as the key reference point, becomes
crucial for the language user in line with the position of the Symbolic
Interactionists. The memory need not include specific detail, though,
beyond the level which permits the elimination of those interpretations of
self s performance which were not desired. The frame for this may well be
148 David Good

culturally specific and will be both constraining and facilitating in the

manner discussed in the previous section. It must include, however,
memory for rudiments of the memories of self that others possess. Given
the potential indeterminacy of any social action or utterance, the frame-
work offered by others' memory of self is something which they must use
to understand self's actions. The nature of their memories will be
revealed to a greater or lesser extent by the character of their responsive-
ness in an interaction. Thus, the memory of self is as much a distributed
model of self as a personal one, if we are concerned with how a memory of
self determines and constrains our social performances.

The interactional moment

The process of coordination seems to both require and produce an
extended sense of the conscious present. Since no single moment can be
taken as an AL, the competent social performer must be able to preserve
from decay thefleetingevidence of what self and other have done, in full,
in case it is transformed by what is done next. The immediate past must be
maintained as it if were happening just then, as there is inevitably a delay
between self's own production and the other's reaction. Only then can the
speaker fully exploit the cognitive resources of the other and the produc-
tion be changed in an appropriate way. An awareness of how this
interdependence might be further exploited by the other, and how the
exchange and relationship might oscillate backwards and forwards
between cooperation and competition, could be the basis of one account
of self-consciousness.

Pragmatics and the evolution of language

As many studies of language-learning animals, be they pigeons, grey
parrots, gorillas or pygmy chimpanzees, have shown, it is not hard to get
an animal to pair a movement such as pecking at a disc or moving the
hands with an object, or in response to another stimulus. Indeed, it would
be very odd if this were not possible. Various claims have been made for
certain primates as being able to use such pairings in a more sophisticated
way partly as a function of their more sophisticated social lives, and
consequently in a way which is more akin to human language. None of
these cases, however, involves the animal in exploiting the relation
between itself and the language being used to generate a variation in
meaning. As was mentioned above, such a use, and thereby the introduc-
tion of pragmatic meaning, would produce a clear benefit in the interac-
tional arms race. Discovered deceit would become less problematic for
both the group and the individual in that it opens the possibility of less
 Where does foresight end and hindsight begin? 149

divisive outcomes in such cases. The simplest way in which pragmatic

meaning could be introduced is by a query by self as to the truth of what
self is indicating or asserting.

Endpoint
If the first part of this essay was speculative, the last part is wildly so. I
hope, however, that this brief sketch is sufficient to move the role of
hindsight to the forefront of everyone's mind so that we can give it the
greater consideration which I believe it deserves in the anticipation of the
interactional future.

Acknowledgements
My many thanks to Esther Goody for her encouragement, criticism, and endless
patience. Thanks also to all the participants in the workshop from which this
volume grew, for their comments on the original version of this chapter.

Note
1 There are, of course, some instances in a conversation where an action is
dependent solely on the action of self like a movement in chess (see below). A
real shibboleth is one such instance, as Jephthah the Gileadite realized to the
cost of the Ephraimites. See Judges 12, 4-6.
PART III

Genres as tools that shape interaction

 PENELOPE BROWN

Politeness strategies and the attribution of

intentions: the case of Tzeltal irony

In this chapter I want to take up the idea that human thinking is

systematically biased in the direction of interactive thinking (E. Goody's
anticipatory interactive planning, or AIP), that humans are peculiarly
good at, and inordinately prone to, attributing intentions and goals to one
another (as well as to non-humans), and that they routinely orient to
presumptions about each other's intentions in what they say and do. I
want to explore the implications of that idea for an understanding of
politeness in interaction. I shall take as a starting point the Brown and
Levinson model of politeness, which assumes interactive thinking, a
notion implicit in the formulation of politeness as strategic orientation to
face.
Brown and Levinson (1978,1987) have proposed a model of politeness
wherein human actors are endowed with two essential attributes: face and
rationality. We claim that face consists of two specific kinds of wants:
positive face (i.e., the desire to be approved of, admired, liked, validated),
and negative face (the desire to be unimpeded in one's actions). The
second ingredient in our model is rationality, the ability to reason from
communicative goals to linguistic means that would achieve those goals.
From these two assumptions - face and rationality - and the assumption
that speakers universally mutually know that all speakers have these
attributes, we developed a model of how speakers construct polite
utterances in different contexts on the basis of assessments of three social
factors: the relative power (P) of speaker and addressee, their social
distance (D), and the intrinsic ranking (R) of the face-threateningness of
an imposition. P, D and R are seen as abstract social dimensions indexing
kinds of social relationship (P and D) and cultural values and definitions
of impositions or threats to face (R). The claim, then is this: however
culturally variable definitions of kinds of social relationship and kinds of
face threat might be, underlying them are pan-cultural social dimensions
(relative power, social distance, ranking of face-threateningness) which
universally go into the reckoning - and the interpretation - of strategic
language choice, and hence we derive the cross-cultural similarities in

153
154 Penelope Brown

choice of linguistic realizations of politeness strategies that empirically

seem to be in evidence. We then went on further, to claim that this model
of politeness universals could be applied in particular cultural circum-
stances as an ethnographic tool for analysing the quality of social
relationships.
In this Brown and Levinson framework, then, sociolinguistic variation
(different kinds of language use in different situations) is portrayed as
rationally tied - via demonstrable means-ends reasoning links - to the
kinds of things people are trying to do when they speak, and cross-cultural
parallels are viewed as attributable to the existence of similar rational
processes underlying human interaction.
This model (and the Gricean model of communication which underlies
it) makes very strong assumptions about humans' abilities to reason
reflexively about each others' desires and intentions. To operate accord-
ing to the model, speakers have to be able to modify the expression of their
communicative intentions so as to take account of what they see as their
interlocutor's views of what they might be taken to be wanting to
communicate, including what impositions to face might be on the table, as
well as his or her assessments of the speaker's and hearer's relative power
and social distance. Correspondingly, interlocutors must be able to read
such modifications as evidence that the speaker is inferring particular
things about their own views, desires, and intentions in this context. An
enormous amount of reasoning, therefore, would have to go into the
construction of any utterance, and one might well be forgiven for thinking
that human communication along these lines is simply impossible.
One way of finessing some of this reasoning might be by using
conventionalized politeness strategies, where politeness meaning is con-
ventionally attached to certain linguistic forms: for example, formulaic
expressions like 'please' and 'thank you', or conventionally indirect
speech acts like 'Could you pass the salt?' where the polite hesitancy of
'Could you?' cannot be read literally and therefore automatically conveys
polite restraint. This is, I think, how the 'person on the street' tends to
think about politeness, as inhering in particular linguistic forms.
And indeed, many interpreters of the Brown and Levinson model have
fallen into the trap of thinking of linguistic realizations of politeness
strategies as necessarily conveying the particular sort of politeness we
associate them with, as if all politeness realizations were conventional-
ized, as if any instance of the use of an in-group address form or an
exaggerated expression of interest were necessarily positively polite, for
example.
What I want to demonstrate here is that not only is this clearly untrue -
no necessary politeness associations attach to any linguistic form in
situated discourse, or rather, whatever associations attach to a form out of
context may be undone in its situated use. But further, conventionaliza-
 Politeness strategies - Tzeltal irony 155

tion does not provide the easy solution to intention-attribution that it first
appears to. Conventionalized politeness strategies still have to be inter-
preted in relation to presumed speaker intentions in context.
One of the ways this lesson was drummed home to me was in my
confrontation with a conventionalized form of expression in Tzeltal, a
Mayan language spoken in the peasant Indian community of Tenejapa, in
southern Mexico, where my ethnographic fieldwork has been concen-
trated.1 I have called the phenomenon 'Tzeltal irony', as it is basically a
matter of uttering propositions which are in the context taken to convey
the 'opposite' or the 'inverse' of what they literally appear to mean. Ironic
assertions in Tzeltal most frequently appear phrased as hedged negative
assertions usually preceded with yu\ 'because', such that uttering a
proposition (P) in the form yu} ma P-uk ('because it is not, possibly, the
case that P') conveys the emphatic assertion that 'P' is the case.2 This form
of expression is elaborated in Tzeltal to an extent that far exceeds the use
of ironic expressions in, say, English or the other languages which have
been used to elucidate the nature of irony. There is an enormous
philosophical and linguistic literature on this topic; I do not address it
here but rather try to disentangle the Tzeltal phenomenon as a thing sui
generis.3
In the context of this book, there are two main reasons for taking an
interest in this phenomenon:
First, because of the cognitive processes involved: the production and
understanding of ironic utterances seems to require an enormous depth of
reflexive reasoning, and a reliance on thefirmnessof mutual understand-
ing between interlocutors, who must assume that they share mutual
knowledge about what each other thinks each other thinks could/must be
true. Just how is this mental acrobatics achieved?
Secondly, because of the ethnographic puzzle: in Tzeltal, ironies are
used to construct social relationships of a particular kind. How and why is
this done through conventionalized irony?
To elaborate: ironic expressions, and their close relative rhetorical
questions, are a very pervasive and salient feature of Tzeltal speech in
interaction. They are an important ingredient in the expression of
positive politeness in this society, used for stressing agreement, sym-
pathy, understanding and commiseration; they are therefore especially in
evidence in women's speech. Both syntactically and functionally, Tzeltal
ironies are very closely related to rhetorical questions, understatement,
and negative assertions (presuming a positive response) - an intriguing
class of utterance types that gives a distinctive flavour to Tzeltal interac-
tion, all of which involve conveying the 'opposite' of the proposition
expressed plus an attitude to it, and carrying the interactional pressure for
the addressee to respond to this attitude.
They are therefore quintessential examples of 'inter-subjective
156 Penelope Brown

perspective-setting and -taking' (Luckmann, Chapter 8, this volume),

and merit our attention to the complex reasoning processes involved in
the achievement of this.
In passing, I might add that Tzeltal ironic expressions lend themselves
to a Sperber and Wilson (1982) type of treatment, which views irony (and
more loosely ironical utterance types) as echoic mention. All these sorts of
'figures of speech' are seen as essentially 'echoing' or more loosely evoking
a proposition which is placed in the context to be laughed at, scorned, or
whatever, in order for an attitude to be conveyed towards it. Ironic
utterances are like things in quotes, so for example if someone says 'Nice
day, eh', in a context of a walk in the pouring rain and sleet, the false
description evokes the image of the accurate description ('Rotten day, eh')
as well as an image of who might have, or actually did, utter the hope/
prediction/expectation that it would be a nice day, this person (or image)
then being the imagined 'victim' or target of the ironic utterance. The
Sperber and Wilson account doesn't claim that all this is the 'meaning' of
the utterance, but rather a set of reasoning processes to get from what's
said to what's intended to be conveyed.
In Tenejapa, ironies, rhetorical questions and understatements are an
important ingredient in all sorts of evaluative discourse - gossip, com-
plaint/commiseration sequences, verbal play (joking, mockery, banter) -
as well as in other breaches of the 'quality maxim': verbal deceit, social lies
(e.g., the token refusal of offers, pseudo agreement, and denying know-
ledge of answers to nosy questions). They add up to a peculiarly
Tenejapan (and especially Tenejapan women's) way of speaking,
conventionalized and elaborated in many domains, and tied to other
aspects of Tenejapan social life - to non-committalness, fatalism, values
about privacy, vulnerability to gossip, avoidance of open conflict, for
example.4
Let's look, then, at the culturally standardized use of ironic expressions
among Tzeltal-speaking Tenejapans.
Tzeltal ironic expression - some politeness strategies in use
An understanding of ironic expression in Tzeltal requires two things: (1)
an analytical approach to irony within a universal framework - an analysis
of the kinds of jobs irony can do in conversation, by virtue of its rationally
based construction oriented to Grice's 'maxim of quality'. In Brown and
Levinson (1987) we analyse irony as a politeness strategy which is an off
record strategy in origin, but in context can be positively polite.5 But
ironic expression in Tzeltal is conventionalized off record, for very often
no non-ironic reading is entertainable in the context, and therefore we
also need (2) an understanding of normatively stabilized usages in Tzeltal,
and of how such normative stabilizing affects what ironically phrased
utterances can be taken to mean in different contexts.
 Politeness strategies - Tzeltal irony 157

I choose to develop such a complex example here because with it I can

show how a given type of linguistic form can have multiple situated
meanings and therefore multiple politeness (and impoliteness) functions;
an ironically expressed utterance can, in context, convey any of the three
types of politeness superstrategies: it can be (1) genuinely off record,
ambiguous in the context as to whether an ironic reading is intended; it
can be (2) negatively polite by virtue of being conventionally indirect, and
(3) perhaps most characteristically, it can be positively polite, emphasiz-
ing common ground; it can also convey impoliteness. But, though the
interpretation changes with the context, the underlying strategic orien-
tation to face concerns shows through across contexts. So in this society,
in which I have described the women as prone to emphasizing positively
polite expression amongst themselves (Brown 1979,1980), I want to show
what happens to women's affiliative speech style, and specifically what
happens to the interpretation of potentially ironic expressions, when
hostility, anger or mere suspicion characterizes their interaction.
First I'll briefly characterize the Tzeltal ironic formulations, and show
how they work in ordinary cooperative speech. Then I'll illustrate their
use in a Tene japan courtroom confrontation, where they become sarcastic
agreement used to emphasize ^agreement. And finally I'll illustrate the
intricacies of ironic/rhetorical expressions in a more ambiguous case,
where ironic expression hovers between affiliative and dissociative func-
tions during the conversation.

Conventionalized Tzeltal ironic/rhetorical expression

There are three basic ways of constructing ironic/rhetorical utterances in
Tzeltal, as summarized by the following formulae: (P represents the
proposition as expressed in the surface structure; -» is to be read as
'conversationally implicates', that is 'in the context conveys'):
Type 1: a hedged or questioned proposition, optionally introduced by
'because' (yu'uri), optionally with the subjunctive -uk, and optionally with
dubitative or emphatic particles which may force an ironic reading:
(yu'un) + Q/hedge + P + (-uk) + (emphatic particle(s))-+'not P'
Type 2: a negative proposition, hedged or questioned, again optionally
with -uk and dubitative or emphatic particles:
(yu'uri) + NEG + P + Q/hedge + (-uk) (emphatic-particle(s))-»'P'
Type 3: a WH-questioned proposition (who/why/when/where/what/
how), again optionally hedged and/or emphasized:
WH-Q + P + (hedge) + (emphatic particle(s))-»'No one, no where, no way, for no
reason, P'
158 Penelope Brown

My reason for treating ironies and rhetorical questions as essentially

the same is that the basic mechanism of construction is the same for both,
the only difference being that in ironies, the proposition (phrased as the
'opposite' of what the speaker intends to convey) is hedged with dubita-
tive particles, while in rhetorical questions it is hedged with a question
particle. In both cases emphatic particles and/or prosodies may (option-
ally) disambiguate the intended meaning.

Examples: the basic phenomenon6

Example 1. Context: story about a girlfriend who hangs out with several
different men; S is here emphasizing her own attitude to the risks of
associating with her.
yu'un niwan ya j-k'an ya j-ta mul ya k-a'y
because perhaps ICP I-want ICP I-find crime ICP I-know/feel
Because perhaps I want to find trouble, I know.
((-•'Of course I don't want to get into trouble!'))
Example 2. Context: talk about a man who owes S money and won't pay
it.
yu' bal jo'on ay ba ya j-ta tak'in
because Q I EXIST where ICP I-find money
Because is there anywhere I'llfindmoney?
((-•'Of course not!' (soliciting H's sympathy.)))
Example 3. Context: S is commiserating about H's problems with her
chickens, and offering sympathy.
ma wan tey-uk nax ay ek
NEG perhaps there-SUBJ just EXIST too
It might perhaps not be the case that (the weasels) just are there too.
(-•'They certainly are there!'))
Example 4. Context: following a humiliating request when a man has
asked his younger sister to mend his son's schoolbook, thereby revealing a
breach with his wife whose job it is; his mother sympathizes with his need.
bi_yu'un nix ay x-a'-na' s-tz'isel ek a
why just EXIST ASP-you-know its-sewing too
Just why would you know how to sew it?
((-•'Of course you wouldn't!'))

It is important to note that utterances constructed according to these

formulae are not necessarily ironic (though particle combinations -
especially hedge + emphatic particle - may force an ironic reading, and so
may prosodies, facial expressions, etc.). But they are contextually disam-
biguated, via the assumption of mutual knowledge of each other's beliefs,
 Politeness strategies - Tzeltal irony 159

attitudes, values, and likely events. Further, when ironic they do not
necessarily convey a barbed negative evaluation of some specific target;
the proposition that is held up to scrutiny is not necessarily attributable to
any particular person.

Sequencing in conversation
The archetypal use for Tzeltal irony and rhetorical questions is as
positively polite emphasizing of held-in-common opinions, values, and
understandings. This is displayed in the ways in which they are res-
ponded to in interaction. Sometimes the response continues in the ironic
vein, sometimes it 'translates' the irony with a straight response.
Example 5. Context: story about husbands working on coffee flnca.
A: ... mak ma wan xch' ajubotik ta stamel in ch'i
... perhaps it's not possibly that we get tired from bending over to pick
(coffee, that has fallen to the ground) then! ((->• sympathetic
understanding: 'Of course, it's tiring!'))
B: mak bi yu'uni ma xch'ajub
Perhaps why don't (we) get tired. ((-•'We sure do!'))
A: yak mak
Yes, perhaps. ((Agrees with implicature.))
Example 6. Context: grown daughter just arrived back from several
days' visit in town has been telling her mother what happened there. Her
mother (A) reacts sympathetically to her tale of trials.
A: yu' ma sakubenuk a jul ch'i
Because not sort of pale, you arrived, to be sure! ((-•'Gosh have you ever
come home pale!' i.e., 'Poor you, what problems you've had!'))
B: asakub
Oh (yes), pale. ((Straight reply, agrees with the implicature.))
A: saktuntun nanix a kil
Really pale, I see. ((Agrees with agreement.))
B: ya'benjulel.
Pale and thin (I've) come home. ((Agrees with agreement.))
Equally, ironies may be used to elicit sympathy:
Example 7. Context: household visit; Smalltalk about crops.
P: jm. bi yilel? lek bal a laj ala ch'i yilel?
Oh. How does it look? Is it (your chili) growing well? ((Straight question.))
M: banti lek ya xch'i. mak yu' ma ja'uk kola' a ya'y xan, jich nix a taot
xan ja'al ini.
Where does it grow well. ((-• 'Nowhere; it's not growing well'.)) Perhaps
because it's not the case that it's rotting again, thus it has just
encountered this rain again. ((-•'It is rotting, from the rain.'))
Example 7 also illustrates the fact that ironic utterances are not solely used
to engender repeat cycles of agreement on mutually known facts; they are
160 Penelope Brown

often used to inform on matters about which the addressee has no personal
knowledge. However, in these cases the addressee has the option to
respond without agreement - indeed, with doubt or outright denial of the
conveyed proposition, thus conveying that the prior speaker's presump-
tion of shared knowledge/attitudes was in this case unwarranted.
But when ironies are used to proffer an emphatic opinion, the response
may continue in the same mode, so that the whole interchange is carried
on in a non-literal mode (whilst appearing to be revelling in their
agreement that 'P', they are really assuring each other on their agreement
that 'not-P'). For example, in this exchange:
Example 8. Context: girls gossiping about a newly married friend.
A: ya wan xjalaj xkal
She perhaps will stay a long while, I say. ((-• 'Of course she won't stay long
with her new husband.'))
B: yu'wan ma ya sta o'tanil.
Because perhaps she won't get tired of him. ((-•'Of course she will tire of
him!'))
Here, by rephrasing A's ironic comment rather than merely repeating it,
B demonstrates her understanding of A's ironic intention, and provides
even stronger accord and agreement. This can presumably only be pulled
off if they can with reason presume that they do in fact share similar
presumptions about the likelihood of (in this case) the friend sticking to
her new husband. (If in fact they didn't share the same views, the thing
would proceed in a different direction, as each pulled the inferential
structure in the direction of her own (incompatible) set of beliefs.)
Similarly, in the following joking exchange, we find ironic irony, again
requiring shared attitudes (in this case, to the undesirability of marrying a
Ladinoized Indian and the probability of marrying a real Tenejapan) as a
prerequisite to its successful achievement:
Example 9. Joking exchange in context of gossip session about a woman
who has been pursuing a married man; they've just agreed that it's better
to go after unmarried men.
ANT: ... ja' ya jk'an xbaon ek in ta lumi. ((laughing)) 1
It's that I want to go to town, too. (Indirect reference to one of her
unwanted suitors, who wears Ladino clothes, speaks Spanish and
lives in town.)
((-•'I want to go to town to see this lovely boyfriend, ha ha.'))
x: ((laughing)) solel ta lum ya'k'an ban. 2
Wow, to town you want to go!
ANT: ja'ya jk'an ba jmulan ek in lumi, ja'eki sawuli. 3
It's that I want to go enjoy town, it's this Sawul-fellow (who tempts me
there).
x: yu'un nix wan ma ja'uk leke, joyob kaxlan ye tz'in. 4
Because perhaps it wouldn't just be good, he's completely Ladinoized.
((-•'That would be great, since he's Ladinoized.'))
 Politeness strategies - Tzeltal irony 161

ANT: joyob ye tz'in. 5

Completely, to be sure.
x: yu' wan indijena to yich'at yael ya'wa'y a tz'in ch'i. 6
Because perhaps an Indian will marry you you know then.
((-•first-level implicature: 'Of course not; you'll marry a Ladino.'))
((-•second-level implicature: joking; 'Of course we really know you'll
marry an Indian.'))
ANT: eh kaxlan ya xk'oon ta lum ya'wa'y tz'in ch'i. 7
Oh, as a Ladino I'll arrive in town you know then.
((-•agreement with ironic reading; 'Not as an Indian's wife but as a
Ladino's wife will I appear in town.'))
x: k'unk'un ala bestido ya'wich'ix ala bechel ek a. 8
Gradually you'll wear Ladino clothes yourself too!
Here (lines 6-7) we have irony within irony: they both know perfectly well
that Ant has no intention of marrying a Ladino; this is a joking scenario in
ironic mode.7
Example 9 also illustrates the characteristic in-group Tenejapan res-
ponse to teasing, which, apparently in contrast to English (see Drew 1987,
and Chapter 5, this volume) rather than being 'po-faced' (serious)
actually upgrades the tease as an elaborate self-tease, sequentially build-
ing up over several turns an increasingly ludicrous scenario in which
teaser and teasee collaborate. The initiators of such teases are designed to
provoke the sequence that follows, and this forward-looking joking is a
common form of verbal play among young Tenejapan women.
To sum up: perhaps the archetypal uses for the ironic forms of
expression we have been examining are in women's positive politeness
'grooming', relishing and elaborating their mutually shared views on an
issue. How they can so blithely assume that their views are mutually
shared (and despite the fact that they occasionally mistakenly make this
assumption) is a mystery; my data come largely from interactions between
people who know each other very well, and whether ironic expression
occurs freely between relative strangers I cannot say. I suspect that it
does, for just those sorts of attitudes that can be generally assumed to be
shared (attitudes to health, economic problems, ritual responsibilities, for
example).
This form of expression seems to be a highly conventionalized form of
feminine positive politeness in this society (see Brown 1979, ch. 4, for
more details). It takes perhaps its most extreme form in women's joking
scenarios like the one illustrated above, where they carry on in the ironic
mode as an activity in itself: a false picture of the world is elaborately
developed by speakers both uttering nothing but ironic utterances and
agreements (ironic or straight) with them. Who exactly the target is of
these 'false scenarios' - who might be being set up as the voice which is
being echoed in order to be mocked, in Sperber and Wilson's terms - is
not entirely clear. In many cases it seems to be something like the
162 Penelope Brown

personified voice of public opinion, or of malicious gossip - whoever

might be likely to be making these patently false, and patently contra-
normal female-role-expectations claims about the subject of the ironic
assertions. Structurally what seems to be going on is that the girls play off
their own (real) attitudes and fears against a personified voice of public
opinion that is bad-mouthing them. This bad-mouthing can be straight,
or it can be ironic itself, as we saw in example 9 where the girls were
teasing one girl about marrying a Ladino. In that case, the apprehension
about public opinion is that it will suggest that they, true Indian girls, will
marry Ladinos and 'go over to the enemy' as it were. So representing that
voice in an ironic manner, they say 'Oh yes, sure you're going to marry an
Indian' conveying ironically the voice of public opinion's conviction 'I
bet you're going to marry a Ladino.' It is against this puppet voice of
public opinion that the jokes about sex role identity take their force.8
The same sort of process, the setting up of a 'voice of malicious gossip'
against which communicative intent is to be read, seems to be involved in
other characteristic forms of Tenejapan expression. It appears in Teneja-
pans' self-mockery, as when they say, in effect: 'I'm stupid, lazy, useless,
my clothes are full of holes, ...' This can be friendly joking; it can
however take a more pointed turn, for example in teasing one's close
household members about non-love and non-care (scenario: 'You (I) are
poor and homeless, no parents, no food, will die of the cold and starvation,
will be thrown to the dogs to eat, etc. etc.'). Such joking can carry heavy,
non-affiliative implications, and the same format can be used as an
indirect expression of anger.
Which leads us to our next set of examples: what happens to women's
affiliative positively polite style in situations of open conflict}

Angry irony
Tenejapan women, of course, don't invariably engage in positively polite
affiliative interactions. Conflict, overt anger and scolding, do occur,
however normatively disapproved of, and it is instructive to look at ironic
expressions in these sorts of contexts.
But first a few words about conflict in Tenejapa. Women in daily
interaction tend to suppress conflict; it is veiled, even in private, and
between non-intimate adult women openly angry confrontation rarely
occurs. When one is angry, interaction with the provoker of one's anger is
simply avoided, and gossip, mockery and backbiting against the object of
one's anger are expressed to sympathetic intimates. Anger between
women who are intimates is normally expressed through controlled
'leakage': silence, non-responsiveness or terse replies, and kinesic dis-
tancing, which in contrast with normal relaxed behaviour may suggest
anger.
 Politeness strategies - Tzeltal irony 163

There are however two types of contexts in which open conflict is not
only tolerated but expected from women. One is in courtship, which
traditionally is coloured by the girl's (apparent) outrage at being
approached; the other is in court cases or more informal grievance
hearings. In both of these contexts, ironic/rhetorical expressions abound.
These, in such contexts of hostility or conflict, can be used to emphasize
the attitude of hostility, as in this (reported) example of a girl's reaction to
her unwanted suitor:
Example 10.

ANT: ej yu' bal xbiket mach'a ya ya'yat mene' xkut.

'Oh, is it that the one who hears you is happy', I said.
((-•'No, you sure don't make anyone happy, you creep.'))

In the next example, taken from a Tenejapan court case which was
filmed and tape-recorded in 1980 by Stephen Levinson and myself, two
women angrily confront one another in a public forum.
Example 11. Context: P (plaintiff) is claiming that D (defendent) owes
her various debts; D is denigrating the value of the goods she allegedly
owes P for. Immediately preceding context: P has been listing her claims;
J (judge) has just said 'wait a minute' and begun to write them down, one
by one. (Underlined line numbers indicate ironic/rhetorical utterances)9
j : te:um (1.5) tzekel tz'i 234
(for the) skirt, then?
P: jm (.) ox-chejp tz'in tzekele (1.5) cha'chejp chujkilal 235
Hm, three hundred then for the skirt, two hundred for the belt. ((Making a
claim for what D owes her for these items.))
D: bi yu'un ma ja'uk tz'in mak yu' ma jo'winikuk (sti) = 236
Why wasn't it then, perhaps it wasn't one hundred or so.
((-•it was only worth 100 pesos (disputing P's claim in line 235).))
= mak bit'il ta' ya stoytik yu' mak ja' te sle bi xan ae 237
How is it that they overstate (the price of the belt) because perhaps it's that
she (P) is looking for something more (from me)!
(1) 238
( ) 239
[ ]
P: ya stak xa'leben (.) 240
You can look for (it - a belt) for me,
ya stak' xa'leben sjol teme jiche 241
you can look for a replacement for it for me if
that's how it is. ((->-'If you think that
was a cheap belt, get a better one!'))
[ ]
D: ja' yu' wan ja' tz'i batz'il stzotzil tz'i mak = 242
It's that, perhaps it's that it's real wool then perhaps!
((-•It wasn't of real wool; i.e., it was
cheap!))
164 Penelope Brown

= ma ja' (tay tz'in men) 243

perhaps it's not (that)
t ]
P: manchuk a tz'in mak yu' ma toyoluk k'uxel tz'i = 244
So what about that then, perhaps it's not that it was
expensive then
((-*•'It was expensive!'))
D: = toyol nanix stukel a tz'i bi mak 245
Really expensive, (it is) itself then, perhaps eh?
((-»• sarcastic agreement: 'It was really cheap!'))

Now, as I develop in detail elsewhere (Brown 1990), not only in terms

of how ironic expressions are used, but also in many other ways, the
interactional conduct of a Tzeltal court case - a formal arena for face-to-
face confrontation with the aim of settling disputes between people - is
the inverse of interactional conduct in 'ordinary conversation' in Teneja-
pan society. That is, ordinary conversational structures and interactional
norms are systematically violated in a public display of indignation and
anger. In this context, certain features - like irony - pervasive in women's
speech in amicable conversation, features which are there used to convey
positive affect, empathy, agreement, sympathetic understanding, are here
used to convey the opposite: negative affect, hostility, contradiction.
In amicable interactions women's positively polite ironic phraseology
assumes and stresses shared values and norms, mutual sympathy and
understanding. In the courtroom confrontation this stance is evoked, but
from a distance, ironically, in the sarcastic politeness of hostile pseudo-
agreement. Presumably the courtroom frame, the mutual knowledge of
licence for inverted conventions in this special context, as well as all the
additional information about hostility (conveyed by kinesics and eye
contact, intonation and gesture) makes it not only possible but inescap-
able that the ironic utterances in this context must be interpreted as the
opposite of what they would be interpreted as, were they to appear in a
cooperative context. Certainly in the courtroom, the use of irony does not
imply that the interactors know each other well, or share the same
attitudes.
This suggests that in order to understand how even conventionalized
ironic/rhetorical utterances are actually interpreted in speech, we need to
look very closely at the uses of such forms of expression in different
situations. One situation casts light on others, especially if one (like
confrontation) is defined in opposition to the other (courteous interac-
tion). The meaning-in-context of ironic expressions is clearly different in
the two cases.
 Politeness strategies - Tzeltal irony 165

An ambiguous case
It will be instructive, therefore, to look at one more interaction, where
what is going on is much more indeterminate. A woman (T) had come
visiting to request an injection from one of the women (A) in the
household who is authorized to give injections. A wasn't home, but T
talked to A's sister X. Throughout the ten-minute encounter, the
kinesics, gaze avoidance, smirking, general body orientation, and dis-
placement activity combined to indicate acute discomfort and embarrass-
ment on T's part, and awkwardness mixed with surreptitious enjoyment
by X. The placement of ritual high pitch10 and the sequencing of
utterances shows where the awkwardness is focused, and it is instructive
to look at the treatment of ironically phrased utterances in this context.11
Example 12. Context: T has been getting a series of injections from A, a
local 'nurse'; she had come last night to get one but A had refused to give
it. So T came again today, understanding that there might be reluctance
on A's part to give her an injection. It emerges during this encounter that
there is reluctance, due to gossip that A heard T said about her (A).
In this excerpt T is trying again; A isn't home but T has been invited by
X to wait for A. After T and X chat about where A is, when she'll be back,
health, weather and work, X mentions directly that A is angry. (Under-
lined line numbers indicate ironic/rhetorical utterances.)
x: ay ja' chikan ya xlijkix sk'ajk'al ya'wil tz'ine mak 122
It's apparent that she has begun to get angry, you see, perhaps.
[
T: eh yak ye in = 123
eh, yes so it is
x: =yak= 124
Yes.
T: = ma sk'an a xjulonix tal ye tz'in = 125
She doesn't want to give me the injection.
x: =ej: lijkemlaj sk'ajk'al 126
Oh, she's gotten angry, (someone) says
T: jai: bi laj yu'uni 127
What! Why (is she angry), does (someone) say?
x: baxa'wili 128
Who knows? ((Lit: Where do you see it? implies 'your guess is as good as
mine'.))
T: yu' mati ay ya'yojben ka'yeje 129
Perhaps she's heard something I said.
x: ay niwanix yu'un ay (.) k'anix ta yu' ay a ka'ye 130
Perhaps that's it - I think I heard that that's it.
At this point, despite her disclaimer ('Who knows?') and hedges ('I think',
'perhaps') it is clear that X has definite information about what someone
166 Penelope Brown

said T said about A; and T's belief that this is the case is made evident in
her replies and direct question to X in line 134:
T: = T binti laj xon ek tz'in 134
So, what did (someone) say I said?
(0.5) 135
x: t m a xkil (•) aY n a x 136
Who knows? ((Lit: I don't see)). It's just -
[
mach'a (.) mach'a xan ya'yej ek tz'in 137
Who - who said it then?
x: t ma xkil me tz'in ma ba jojk'oyebe te banti ya'yoj a'yeje 138
Who knows, I didn't ask her where she heard the gossip (about what you
said about her).
T: jai: 139
What?
x: ju'uj 140
No (I didn't ask her).
T: ah solel ay wa'y ta ba'ay wejtem bi ka'yej 141
Ah, just look where my words have become available!
((i.e., 'just look how what I said got spread around!'))
x: ayniwanayin 142
That's perhaps what happened.
X's diplomatic denial of knowledge of the source of the gossip-mongering
prompts T's defence in line 141 ('just look how my words have been
spread around!') to which X's superficially agreeing reply ('That's
perhaps what happened') could be interpreted either as sympathetic
agreement or as sarcastic pseudo-agreement. She follows it up by quoting
T's alleged talk about A:
T: mm: (1.0) mach'a me xkal 143
Hm, who could it be (who told on me)?
((Lit: Who (is it) I say?))
x: [
kaxel= 144
'Golly,
= ma'yuk ba'ay bak'en ya xba sjulon mene me xat (.) 145
that one (A) certainly doesn't want to give me an injection,' if (that's
what) you said,
mm bi'ora ay k'anbe spoxil mene bi'ora sjulon mene me xat, 146
Mm, 'When is that one (A) going to get my medicine, when is she
going to inject me?' if you said.
t bi laj bal ut'il 147
What's the use of it? ((->-No use, A implicated, in this reported
conversation with X.))

Here the direct quotes put the rhetorical question in the mouth of the
quoted speaker (A), and X thus remains uncommitted as to whether she
shares the attitude indicated (i.e., 'why bother to provide injections for
this woman who bad-mouths me to others?'). T responds with the
 Politeness strategies - Tzeltal irony 167

standard indication of non-hearing or non-comprehension, jai?, p r o m p t -

ing X to expand on her claim:

x: xi a ka'y ja' ya yali (.) peru ma ba la jojk'o mach'a yaloj 149

I heard her say, so they say. But I didn't ask who said it.
T: mach'a yaloj xkal = 150
Who reported that, I wonder? ((-•'No one knows.'))
x: ju'u 151
No. ((Replies to implicature: 'No, no one knows.'))
Having agreed (for current purposes, at least), that they don't know who
the gossip-monger is, T goes on to deny her alleged words:
T: ma'yuk to ba'ay ya xlok'on ta (.) a'yej a ka'y 153
I didn't go around saying (those things) at all, I say!
x: ma'yuk 154
(You) didn't at all.
T: ma'yuk, ma'yuk ba xlok'on ta julbal 155
Not at all, not at all did I go out visiting ((-•and gossiping)).
x: ay (0.5) ay niwan ja' jich ya xlijk yala k'ajk'al yu'un ek tz'in 156
It's, it's perhaps that that made her (A) get a bit angry then.
T: ah: mak bi yu'uni lijk sk'ajk'al a ka'y tz'in 157
Oh, why did she get angry then?
((-•'How unfair, how could she?!'))
x: bi yu'uni ya xlijk a'k'ajk'ali uta (tz'in) 158
'Why are you angry?' you should say to her (then).
T: yakalbeimak 159
Perhaps I'll say that to her.
They go on, shifting around the issue, T continuing to deny she said
anything and to wonder who could have gossiped about her, X continuing
to prick her, indirectly. The topic winds down, as follows:
x: ya'wa'y (.) ya'k'anbe nix xan wokol (.) 233
You see, you just ask her again to take the trouble (to inject you).
T: yaka'y 234
I see.
x: slajinbet ine yip xix jich a tz'i = 235
She'll just finish you up (i.e.,finishthe series of injections).
T: = y i p = 236
just (finish)
x: = bi ja' to tz'in te ay ba'ay (.) yax- k'ejel xa'wakix a'ba ta julel
yu'un a tz'ine 237
So how is it then, that there is somewhere else you can go to get injected
then ((-•'Nowhere else can you go to get injections.'))
T: [
ja' to xan = 238
There still is somewhere else
((-•'There sure isn't!'))
= sok banti ya jta me ineksione = 239
and where will I find an injection? ((—•Nowhere!))
= mak yu' ay jich (.) jtaoj ta jleel ek tz'in = 240
168 Penelope Brown

Perhaps that's how it is, I've found (another injection-giver) then

((-•'I haven't.'))
= te manchuk ja'uk tz'in te (wa'y) bi jpasoj ch'e 241
so what if that's how it were then, (you see) what I'd have done (if there
were an alternative to A as injection-giver).
((-> 'I couldn't do anything if A refused, as there's no alternative to her
services.'))
x: ja'ya'wa'y 242
That's so, you see
T: ja' 243
It is.
x: ja' 244
It is.
T: ja' 245
It is.
(1.0) 246
x: jich ya'wa'y tal (.) ya niwanix slajinbet tz'in, 247
So you'll see, she'll perhaps finish you then,
yu' bal kejchel ya xk'ot a'wu'un 248
because is it that you'll befinished(with your series of injections)
otherwise?
((-•'Of course not, there's no one else to do it.'))
T: baxa'wil 249
Where, you see?
((-•'Nowhere else (can I get injections!').))
(The embarrassment more or less resolved, they go on talking of other
things.)
In this example, then, the shared common ground on which ironic
expression normally relies is missing - X doesn't believe T didn't do this
gossiping about A. X's scepticism is made clear despite her denials (e.g.
lines 128,138) and hedges (line 130), and her agreements with T's denials
(line 154). For example, the contradiction between line 154, where X
(superficially) agrees with T's denial of culpability to the accusation of
gossiping, and line 156, where X says that it's T's gossiping that made A
get angry, clearly indicates her actual belief.
Their relationship in this context is asymmetric, as T is a suppliant for
A's services and X is a mediator between the allegedly angry non-present
A and the suppliant T. In this context, T's rhetorical questions and ironic
utterances are essentially aimed at eliciting X's sympathetic understand-
ing for her plight (needing A's injection-services, A reluctant to provide
them) (e.g. lines 125,157, and 238^0). X's are more multi-functional, for
example denying knowledge and responsibility with the conventionalized
Tzeltal verbal equivalent to a hands-in-air gesture of hopeless fatalism (Jba
xa'wili or ma xkil ('who knows', lines 128,136,138), followed by a patent
social lie ('I didn't ask her where she heard the gossip', line 138). Others
are potentially barbed, carrying a dual message as in lines 237 ('Is there
somewhere else you can go to get injections?' and 248 ('Will you finish
your injections otherwise?'), which can simultaneously be read as taunt-
 Politeness strategies - Tzeltal irony 169

ing or as sympathetic understanding for T's plight. And since T, in her

response to X's potentially dual message, in line 238 takes up the second
of those meanings, responding with an ironic agreement to the ironic
implicature in X's line 237 (saying, in effect, 'I agree, there certainly isn't
anywhere else I can go to get injected, oh poor me.'), she humbles herself
by throwing herself in effect on X's (and thereby A's) mercy. As she
solicits sympathy with piled-up ironies (lines 238-40), and accepts X's
rubbing it in with agreements (lines 242, 244) and re-phrasings of T's
dilemma (line 248), the uncomfortable ambiguity vanishes (T is hum-
bled, X triumphant) and they can carry on conversing amicably.
This example shows that topic-summing-up with ironically phrased
utterances does not necessarily display real common ground in the sense
of shared attitudes. The agreement that there is nowhere else to go to
finish her injections does not necessarily carry with it sincere humility on
T's part, nor sincere sympathy on X's; the surface agreement simply
provides a negotiated end to the topic.

Conclusion
The use of ironic expressions in Tzeltal is an elaborate and complex
phenomenon, but I have used it to make a very simple point. One cannot
mechanistically apply the Brown and Levinson model of politeness
strategies to discourse data; particular linguistic realizations are not ever
intrinsically positively or negatively polite, regardless of context. Polite-
ness inheres not in forms, but in the attribution of polite intentions, and
linguistic forms are only part of the evidence interlocutors use to assess
utterances and infer polite intentions. So however many pleases, thank
yous, bows and scrapes you may make, the polite attitude supposedly
conveyed by them can be undermined, inverted, or cancelled by their
interaction with other elements in the context, and interactors can't just
sit back and let conventionalized expressions do their interpretive job for
them. Rather, they must constantly work at inferring each other's
intentions, including whether or not politeness is intended. This is
especially obvious with irony, where they have to infer whether an
utterance is ironically intended, whether it is ironic irony, who, if anyone,
is the intended target or victim (obviously crucial to an interpretation of
its politeness value), and who is the intended audience (in the court case,
for example, the judge's role as arbitrator is crucial to the interpretation of
the litigants' confrontational performance).
Successful irony thus relies completely on interlocutors' (and
audiences') ability to decode mutual knowledge assumptions about what
each thinks the other thinks must be true. The Tzeltal examples demon-
strate that the reflexive reasoning involved must be at least four levels
170 Penelope Brown

Table 7.1. Levels of reflexive reasoning

Sincere assertion I assert P

I believe that you believe I believe P is true
I hope therefore to get you to believe P is true
True deception I assert not-P, though I believe P is true
I hope you believe I believe P is not-true
I therefore hope you'll believe P is not-true
Pseudo deception I assert P, though I believe not-P is true
(social lies) I believe you believe I believe P is not-true
I believe you know (and believe I know) the customs concerning
constraints on revealing certain kinds of information
I therefore don't expect you to believe P, but to desist from
pressing for this kind of information
Irony I 'mention' P, hoping to evoke Q (which is systematically related to
P)
I want you not to think I believe that P is the case
I believe you believe I believe that not-P is the case, and
I believe you believe I believe you believe that not-P is the case
I therefore hope you will respond to Q
Ironic irony I 'mention' P, in the frame of non-serious joking
I believe P is true, and I believe you believe P is true
I believe you know this is non-serious talk, and therefore hope you
will within this joking frame take not-P to be true for current
purposes

deep. An informal Gricean account of these levels might go along the lines
presented in Table 7.1.
There are of course many clues, in addition to Gricean flouts of
sincerity, to prod the inferential process in a particular direction, restrict-
ing the range of possible meanings a speaker is, in the context, likely to be
taken as intending, or hinting at his or her actual intentions. These
include 'background knowledge' of the interlocutors' previous inter-
actional history, their social roles and relationship, the nature of the
speech event, and the immediately preceding discourse, insofar as it
provides a record for current purposes of their respective beliefs. The
characteristics of the ironically phrased utterance itself may also provide
clues: non-verbal affective cues in the kinesics, facial expressions, and
gestures accompanying the utterance; intonational and paralinguistic
cues (timing, voice quality, stress).12 In Tzeltal, especially important is
the special set of emphatic and hedging particles which combined
together force an ironic reading: an utterance along the lines of 'Perhaps it
might be the case that P, to be sure!' can only be read as emphatically
asserting not-P.
Furthermore, in Tzeltal the fact that irony is a conventionalized form
for expressing sympathetic understanding affects the interpretive pro-
cess, providing an extra layer as it were of possible embedding in the form
of joking or of sarcastic irony-as-pseudo-agreement.
 Politeness strategies - Tzeltal irony 171

Nonetheless it is important that irony plays a role not only in well-

defined interactions where a speaker's intent can in general be presumed
to be either clearly sympathetic (as in positively polite 'grooming') or
clearly hostile (as in confrontations). Irony in Tzeltal is exploitable as a
resource for obscuring one's communicative intentions, muddying the
waters so that interactors with potentially different goals can pursue them
without flaunting their differences overtly. The indeterminacy of the
relation between a speaker's expressed attitude and his or her real attitude
may be put to the service of various devious interactional goals.
We might well ask: why in the world do Tenejapans bother to do all
this? Why elaborate a form of discourse where you have to be constantly
figuring out whether your interlocutor means what (s)he says, or the
opposite, or the opposite of the opposite? Part of the answer might be that
the very reliance of irony on mutually shared knowledge and values makes
it a good test of whether in fact interlocutors do share knowledge and
values, and thereby an excellent way of emphasizing and reinforcing
claims to common ground. It is thereby an essential element in construct-
ing the social relationship between the interlocutors, insofar as it succeeds
in constructing the 'phatic communion' that inheres in demonstrations of
mutual understanding. It is clear that in Tenejapa, for women at least,
such interactions are an integral part of their social relations, they must
have such conversations in order to consider themselves related to each
other. That is perhaps the motivation behind positively polite uses of
irony; then, given this resource, it is an obvious next step to its exploit-
ation in order to sneer or insult.
I would suggest that there is a further reason why Tenejapan women,
especially, make such elaborate verbal play with 'literally false' assertions,
especially in relation to public opinion, and why they have elaborated
their humour around the paradoxes which they pose. The ways in which
women joke by claiming false things about themselves and each other -
precisely those false things which in others' mouths would humiliate and
destroy them (about their sexual exploits and desires, their failures as
fulflllers of female roles, etc.) - is a way of neutralizing the destructive
power of such falsehoods, just as joking about illness and death ('Maybe
I'll be permanently lame from this sore on my leg; maybe I'll die from it.')
is a way of undermining the poignancy of fears about incapacity and death
that both men and women have in this society. Men, not nearly so
vulnerable to the effects of gossip about their sexual - and sex role -
misbehaviours, do not take the same lines in their solidarity-stressing
joking. I think it is significant that men, to stress solidarity with each
other, joke about what they do do, would do if they could, or have caught
someone else doing; whereas women joke about doing what they wouldn't
be caught dead doing and would be mortified and appalled if anyone
seriously suggested that they would.
172 Penelope Brown

It seems probable that irony is a typical communicative strategy for

certain kinds of social groups. Tenejapa, for example, is a 'gossip society',
a small-scale face-to-face society where people are obsessed with what
everyone is saying about others, and especially about themselves.13 This
may make it a good candidate for the sociolinguistic elaboration of ironic
expression. I would also suggest that irony is a common strategy for the
underdog, for persons in subordinate or vulnerable positions. Irony is a
way of tacitly complaining about one's lot in life - of some of the
consequences of one's underdog position - without actually upsetting the
status quo. The complaint is off record (very much so in these ironic
joking scenarios where interactors set up public opinion against them-
selves) and it essentially affirms the right of public opinion to maintain
such an eagle eye on their behaviour. Indians have for 500 years been the
underdogs in Mexican society, and women have been the underdogs
within the Tenejapan community; this objective status contrasts with
their strong sense of ethnic identity and community pride. As a technique
for making play with the ambiguities of this contrast, irony seems ideally
designed.
Finally, to return to the AIP model - Tzeltal irony provides a prime
example of one way in which humans' highly developed intellectual
machinery for inferring alter's intentions is put to the service of social
relationships. Language allows us to make propositions, with which we
can plan, discuss the future and non-present events, propositions which
can be used non-seriously to joke, which can lie, or pretend (play-lie), and
which can evoke mental models which are held up for affective comment
(e.g., ridicule or contempt). The process of doing this successfully is to a
large extent the social relationship of solidarity in Tenejapa, and this
institutionalized socio-cultural strategy, in the form of a particular 'way
of putting things', puts a characteristic stamp on Tenejapan social
interaction.

Acknowledgements
This chapter amalgamates some points described in more detail in Brown (1990)
with my analyses of Tzeltal irony first presented at seminars in Australia
(Australian National University, 1982) and London (Sociolinguistics Sym-
posium, 1989), as well as in Brown (1979: ch. 4). I am grateful to Stephen
Levinson, Esther Goody, and John Haviland, and to the other participants in
these seminars and the Workshop On the Implications of a Social Origin of
Human Intelligence, for their helpful comments.

Notes
1 Thisfieldwork- over the past twenty years - has focused primarily on social
interaction in Tenejapa, and a large corpus of tape-recorded and/or filmed
 Politeness strategies - Tzeltal irony 173

naturally occurring Tzeltal interaction has been built up, from which the data
analysed here has been drawn. Tenejapa is a municipio in the Chiapas
highlands, in a heavily populated rural area where there are many other
communities of Tzeltal or Tzotzil speakers, each of which maintains a strong
ethnic identity distinguishing it from the others and from the dominant
Ladino (Mexican national) culture.
2 That is, where P can be expressed by any well-formed Tzeltal sentence, the
position of the subjunctive suffix -uk (which conveys possibility-hedging)
mutatis mutandis. The notion of'opposite' is not quite accurate for the case, as
Stephen Levinson has pointed out. Tzeltal irony is not just a matter of
asserting P and conveying not-P, or vice versa, but rather something like the
following: an assertion is made to the effect that one end of a scale is (possibly)
the case, and in context this implicates the emphatic assertion that the other end
of the scale is the case. Irony thus forces descriptions to polarized ends of a
continuum of evaluation on some dimension (good/bad, desirable/undesir-
able, likely /unlikely, and so forth). I use 'opposite' as a shorthand for this flip
between conveying the ends of an evaluative continuum.
3 For some discussions of irony in relation to the Gricean maxims, and in
discourse contexts, see the references cited in Brown and Levinson (1987: 28).
4 Despite its omni-presence, ironical phrasing of utterances is not to my
knowledge an 'emic' Tzeltal category. While Tenejapans explicitly dis-
tinguish two basic categories of speech, poko-k'op (traditional, ritual speech
styles - in Stross's (1974) characterization 'elegant, stylized, serious, non-
malicious speech') vs. ach'k'op, ('recent speech' including all forms of non-
serious speech), and while they have lexical items denoting particular kinds of
non-serious speech (e.g., lotil, 'lies'; ixta-k'op, 'joking'; lo'il k'op, 'carnival-
style joking'; tajimalk'op, 'verbal games', labanel, 'mockery'), I don't know of
any label specifically designating conventionalized Tzeltal irony.
5 In Brown and Levinson (1987) we make the following distinctions: joking,
which as a positive politeness strategy stresses in-group relations and common
ground, and can include non-serious insults; mockery, which when of someone
other than the addressee can stress in-group solidarity; teasing, which can be
joking or mockery even with the addressee as target, but playful; and irony,
which is at root an off-record strategy but if on record in context can be
positively polite solidarity stressing.
6 S stands for 'speaker' H for 'addressee'. The Tzeltal practical orthography has
the following conventions (where they differ from the International Phonetic
Alphabet): j represents /h/, ch represents /c/, x represents l\l, tz represents /ts/
and' represents either glottalization of the preceding consonant or, following a
vowel, a glottal stop. Abbreviations for morpheme by morpheme glosses used
in the text are as follows: ASP stands for neutral aspect, EMPH for emphatic
particle, EXIST for the existential predicate, ICP incompletive aspect, NEG
negative particle, Q question particle, SUBJ subjunctive.
7 Note that in line 6 of this example the trigger for irony is not a patent falsehood
but a patent truth, which within the non-literal joking frame must then be
reinterpreted. The levels of inference required may be diagrammed as follows:
174 Penelope Brown

lit: Because perhaps an Indian will marry you, you know. jj-.
c.i.l: Of course no Indian will marry you; you'll marry a Ladino.^ I
But: Ha ha, we both know you'll certainly never marry a Ladino.«—'
Therefore c.i.2: You're joking; of course you'll marry an Indian! '

8 Tenejapans distinguish 'real' lies (lotil- lies intended not to be seen through)
from such joking lies, which are intended to reinforce solidarity. Akin to 'real'
lies, perhaps, are those social lies which are non-joking but, though intended
to conceal some fact, are so obviously false that there is no real deception. For
example, in response to a 'nosy' question that one does not wish to answer
(e.g., 'How much do you earn at that work?') a patently false answer is
customary (e.g., 'I don't know, I can't count' or 'Two pesos per month'); this
implies that the speaker has no intention of answering such uninvited
questions. The effect of real lies, then is interactionally distancing, in contrast
to the joking lies (or elaboration of non-truths) which reinforce mutual
knowledge, mutual values, and friendship.
9 For conventions used in the transcription of this and the following example,
see the 'Conventions used in transcripts' on p. x.
10 Very high or even falsetto pitch is used in Tzeltal to convey deference in ritual
speech, greeting and farewell exchanges, and wherever formality is being
emphasized. It tends to characterize the beginnings and endings of encounters
in general, and often marks socially sensitive or potentially threatening
utterances, as well as fatalistic resignation. It is marked in the transcription by
an t preceding the utterance.
11 This interaction took place in 1980 and was filmed by S. Levinson and myself.
The soundtrack was transcribed with the assistance of one of the participants,
X herself.
12 In his analysis of 'sarcasm as theater', Haiman (1990) has observed that a
number of these cues appear to be cross-linguistically applicable. He mentions
three kinds (1990:181): 'a. formal indices of direct quotation or repetition . . . ;
b. incongruity between segmental and suprasegmental texts (incongruous
suprasegmentals include the phonetic reflexes of sneers and laughter, deadpan
monotone, caricatured exaggeration of the appropriate melody, and stylized
or singsong intonation); c. hyper-formality (including both high register and
the substitution of linguistic signs [like "ha, ha"] for paralinguistic symptoms
[genuine laughter]).'
These 'stage separators' or mechanisms for indicating that what is acted out
behind them is 'not serious' are not generally involved in routine Tzeltal
ironies, though they (especially those under 'b') do mark as 'on stage' the
joking sequences illustrated by example 9. Tzeltal speakers do not seem to find
it necessary to heavily mark their non-straight utterances as non-straight; the
inferential trigger is more likely to be simply a conflict between the expressed
proposition and mutually assumed knowledge.
13 See for example Haviland's (1977) description of the neighbouring Tzotzil
community of Zinacantan.
 THOMAS LUCKMANN

8 Interaction planning and inter subjective

adjustment of perspectives by communicative
genres

The aim of this chapter is to show that communicative genres play a

distinct and rather important role in the reciprocal, inter subjective
setting of perspectives in human communicative social interaction. It will
be useful first to establish the socio-cultural context within which
communicative genres perform their function and to specify the levels of
theory on which propositions about communicative genres in social
interaction are to be placed.
It will eventually make sense to restrict the term social interaction to
behaviour with which an individual organism which is phylogenetically
equipped with consciousness in its 'biogram' associates a more or less
definite meaning (a motive, a projected goal).1 Here I first take the term to
refer to any behaviour which is directed by one individual organism at
other(s) of the same or a different species, and which has consequences for
such behaviour of the other individual(s) as, in turn, is directed at the
'first' individual.2
Social interaction in the human species may be viewed in a comparative
biological perspective and compared to social behaviour in species as
closely related to ours as other apes and primates and as distantly as
'lower' mammals. It may be compared with respect to structural and
functional analogies, and possible evolutionary connections may be
considered. If ethological comparison navigates successfully between the
Scylla of biological reductionism and the Charybdis of anthropomorph-
ism, it may show suggestive similarities in the structures and functions of
social interaction.3 General features of human social interaction are also
revealed by a philosophical approach. Phenomenological reduction strips
away, layer after layer, the concrete socio-cultural, historical components
of the processes of social interaction and shows their universal constituent
elements.4 Both approaches meet in the 'discovery' of the reciprocal
attunement of social actors and of 'anticipatory interactive planning'.5
Phenomenology can show this for the human, ethology for the human and
for related species.
In the broadest sense of the term, all social interaction is communica-
tive. Behaviour directed at others cannot help but convey something to

175
176 Thomas Luckmann

others if they are in a position to observe that behaviour.6 However, it

makes sense to restrict the concept of communication to behaviour which
expresses emotions, signals an incipient action or consists of'abbreviated'
gestures.7 Even with this restriction, communicative interaction is found
in many species, not only in those closely related to Homo sapiens?
Human communicative social interaction shows the most elementary
features of communicative interaction among apes. However, even in
comparison to our closest relatives, the chimpanzees, there is a qualitative
leap in several essential aspects of communicative interaction. This leap is
attributable to the development of language as a referential, time- and
space-transcending sign system.9 With the emergence of language(s), a
crucial step is taken which adds to the evolutionary level of causation
obtaining on a time-scale measured in hundreds and thousands of
generations, a historical, socio-cultural level of 'causation' measured in
several dozens of generations - and less.
Even if the general formal structure of social interaction and communi-
cative social interaction remains basically unchanged, or changes only
slowly, the social, interactive and, most significantly, communicative
construction of social reality (typical motives, typical goals, typical actors,
and typical courses of action, typical explanations and interpretations,
typical legitimations and delegitimations) provide historical patterns of
meaning for the concrete contents of social interaction and communica-
tion. It should be noted that the distinction between formal structures
and empirical (i.e., historical) contents is somewhat artificial. In human
affairs, a historical level of 'causation', an evolutionary emergent, may
interpenetrate with the phylogenic one.
It is evident that propositions about the function of communicative
genres in communicative social interaction only apply to the concrete
level of socio-historical processes and realities. On this level phylogeny,
ontogeny and social construction seem to have merged indistinguishably.
The results of phenomenological analysis, of the search for universals
in cross-cultural comparison, and of the simulation of 'social' processes
on the basis of minimal 'logical' assumptions show a remarkable conver-
gence regarding the elementary structure of human social and, more
specifically, communicative interaction. It can, therefore, be assumed
that anticipatory interactive orientations are one of the universal consti-
tuent elements of the latter. These anticipatory orientations sometimes
take the form of conscious planning of future courses of action. In these,
account is taken of potential, more or less likely courses of action and
reaction by those individuals to whom the original project of action is
directed. Such activities of consciousness will be involved when social
interaction, and communicative interaction in particular, is uncertain and
when it occurs in problematic situations. However, once plans with
 Interaction planning - communicative genres 177

respect to certain courses of action directed to certain types of individuals

prove reasonably satisfactory, they will routinely serve as plans for similar
purposes under similar circumstances. The construction of plans, stra-
tegic calculation of possible reactions of others, complicated intersubjec-
tive adjustment of perspectives, and similar conscious activities which are
characteristic of problematic situations are likely to recede into routines
in unproblematic situations. That is not to say that social interaction
becomes automatic. It may, however, become more or less routinized. Of
course this applies to communicative interaction as well.
It is evident that routinization of social interaction may also be brought
about inter subjectively, in overlapping stretches of social interaction.10
In the following I shall try to show that communicative genres perform
an important function in many kinds of communicative interaction by
serving as ready-made plans in at least partly routinized 'anticipatory
communicative interaction planning'.

'Freedom' and 'constraint' in communicative social interaction

Communicative genres operate on a level between the socially constructed
and transmitted codes of 'natural' languages and the reciprocal adjust-
ment of perspectives, which is a presupposition for human communica-
tive interaction.11 They are a universal12 formative element of human
communication.
The relations between communicative codes as elements of social
stocks of knowledge, social institutions as constitutive parts of social
structures, and communicative acts as embodied components of ongoing
human social life are complex, many-layered, and multi-directional.
Conditions of social life, social codes of communication in general (and
languages in particular) and concrete communicative acts influence and
'determine' each other. The lines of 'determination' criss-cross in social
interactional space and time.
Human communicative acts are predefined and thereby to a certain
extent predetermined by an existing social code of communication. This
holds both for the 'inner' core of that code, the phonological, morphologi-
cal, semantic and syntactic structure of the language, as well as for its
'external' stratification in styles, registers, sociolects, and dialects. In
addition, communicative acts are predefined and predetermined by
explicit and implicit rules and regulations of the use of language, e.g., by
forms of communicative etiquette. Furthermore, communicative acts are
a form of social interaction and are therefore predefined and predeter-
mined by non-communicative rules and regulations: by institutions, a set
of social relations, a system of production and reproduction, in short, by a
historical social structure.13
178 Thomas Luckmann

Languages originate, develop and change under varying social con-

ditions. Social conditions determine the circumstances under which
language is used in communicative acts. On the other hand, communica-
tive acts - and, more generally, social interactions - are instrumental in
changing social circumstances. They influence the development of the
'external' stratification of language; less directly and, ordinarily, also less
swiftly, they influence the maintenance and the change of the elements in
the 'inner' structure of language.
The social regulation of communication is an elementary prerequisite
for the day-to-day working of any society. Theflowof communication in
institutional settings is channelled according to the functional require-
ments of the institution. The frequency and direction of communicative
acts are subject to regulation. Special communicative networks may
become established and segregated in order to prevent outside interfer-
ence. There is, of course, considerable variation in the level of complexity
and degree of specialization that characterize different institutions in the
same society and the same kind of institutions in different societies.14 The
processes of sychronic and diachronic cross-influence between languages,
social structures and communicative acts continue in history and result in
new 'syntheses' in the real lives of real people.
Looking at the relation between society and language from the perspec-
tive of 'real people', we see that both the individual's initial access to the
various means of communication and his subsequent actual use of them
are socially determined. The child's chance of access to the repertoire of
the means of social communication depends upon its specific location in a
historical social structure. Socialization, by definition a communicative
process, is determined by a given historical system of social inequality. In
addition to the unequal distribution of goods, the structure of inequality
consists of an uneven distribution of the social stock of knowledge and of
unequal access to the means of social communication. In addition to its
being a highly significant factor in socialization, the social structure
regulates the actual use of the means of communication in concrete social
interaction.15 These means consist of language(s) as well as of the
(generally) more loosely structured (and mostly also less strongly
conventionalized) mimetic, gestural, postural, etc. expressive forms.
Social interactions in general, and communicative acts in particular, are
only in part a matter of situation-bound intersubjective reciprocal attune-
ment and interactive planning; a planning which originates in subjective
projects of action and their step-by-step adjustment to the actions of
others. But in part they are just that. The degree of interactional
'freedom' inherent in a situationally negotiated adjustment of perspec-
tives varies historically from society to society, and it varies within a
 Interaction planning - communicative genres 179

society in any given epoch from one social domain to another. Intersub-
jective adjustment of perspectives, anticipatory communicative interac-
tion planning and socially constructed 'plans' in the form of communi-
cative genres converge at the intersection of 'freedom' and 'constraint' in
communicative interaction.

Interactive and socially pre-constructed solutions to

communicative problems
The linguistic and institutional determinants of social communication
operate simultaneously on the level of concrete communicative interac-
tion. It is on this level that 'reciprocal perspective setting' and 'perspec-
tive taking'16 must be accomplished by the persons engaging in social
interaction. This intersubjective process involves situational adjustments
of the basic principle of human social life, the principle of the reciprocity
of perspectives.17
The adjustments may present practical problems. In clearly predefined
situations, when the actors perform institutionally determined social
roles, the adjustments may be accomplished quasi-automatically. All that
is needed is the routine application of previously established, socially
constructed and transmitted knowledge about typical settings, situations,
social roles, and courses of action. At the other extreme one may imagine a
situation in which neither background nor specific knowledge could be
applied successfully. In such a situation almost everything that goes
beyond the basic principle of the reciprocity of perspectives must be
'locally' negotiated after a trial-and-error exploration of the perspectives
(previous knowledge, interests, etc.) which are brought into the situation
by the participants. In fact, there will be only approximations of either
extreme. Most cases will be characterized by a mixture of successful
application of routine knowledge and situational exploration and
negotiation.
Some, if not most of these adjustments concern specifically communi-
cative aspects of interaction. They often become problematic. Specifi-
cally, communicative problems are normally solved in communicative
'repair procedures' for faulty 'recipient designs'.18 All social communi-
cation is, of course, addressed to other persons, but there is a broad range
of variation in the degree to which it is subject to intersubjective
^redefinition. The 'normal' mutual assumption of the participants - valid
until evidence to the contrary appears - is that the social models for
particular types of communicative acts available to one of them are also
available to the other one. When such modelling is minimal, we may speak
of 'spontaneous' communicative acts. On the other hand, when such
180 Thomas Luckmann

modelling reaches a high degree of structural 'closure', and when it is

associated with intersubjectively recognizable formal constraints, we are
in the presence of communicative genres.
In 'spontaneous' communicative acts an individual does most of the
selecting and constructing on his own, taking from the communicative
code(s) available to him those elements which seem to fit his momentary
communicative project, in calculated anticipation of typical interpre-
tations by the addressee. When he follows a clearly defined communicat-
ive project he may do so with a high degree of awareness, selecting - and
rejecting - minor and major components of his communicative act in
turn-by-turn responses to the acts of the addressee. It is more likely,
however, that at least some of the steps are routinized, that some of the
parts are pre-assembled habitually.19 Sentences are formed by taking
those words and phrases from the semantic inventory of the speaker
which seem appropriate to the purpose at hand, if they conform to the
pragmatic principles of 'recipient design'. As utterances are built up, the
internalized syntactic schemata are employed more or less automatically.
In addition, the speaker will use stylistic devices and rhetorical strata-
gems, and he will obey - or break - the prevailing rules of communicative
etiquette in relation to particular social types of addressees.
'Spontaneous' communicative acts are 'produced' by an individual
from a mixture of explicit intention and habit in alignment with his
communicative project and, occasionally, by following a superordinated
interactional plan formulated in advance and geared into the intersubjec-
tively constructed sequences of conversation. However, the parts of the
communicative project are not assembled according to an intersubjec-
tively recognizable overall model.20
Nevertheless, in many types of communicative situations individuals
may draw upon pre-cut communicative patterns which seem to fit their
communicative intentions, and which they use in order to carry out their
communicative project - if they have grounds for assuming that the
pattern is also known to the addressee. Such an assumption is based,
generally, on the principle of the reciprocity of perspectives and, specifi-
cally, on the individual's knowledge of the general outlines of the social
distribution of the elements contained in the social stock of knowledge. If
one may say that the individual 'chooses' a genre according to his
communicative project and the perceived requirements of the social
situation, one may say correspondingly that once a genre is 'chosen', it is
the model which 'chooses' the parts in executing the communicative
project. Of course, the communicative realization of a genre may or may
not 'fill' an entire communicative project. Whether it does depends on the
span of the project and on the 'size9 of the genre involved.21
Although all communicative situations involve problems for the parti-
 Interaction planning - communicative genres 181

cipants, some of these may be important, some less important, and others
trivial. Some problems are new but many are familiar. Finding solutions
to communicative problems may require much thought and effort, and,
then again, solutions may merely involve the application of habitual
procedures. The participants may solve their communicative problems
'spontaneously' or by using various minor or major communicative
genres, and they may shift from one to the other in interactive turn-by-
turn responses.22
Communicative genres may be 'functionally' defined as socially con-
structed models for the solution of communicative problems. It is
plausible to assume that such solutions are constructed for recurrent
problems and that, on the whole, they will be provided for relatively
important problems of social communication.23
Communicative genres are part of the social stock of knowledge. Genre
repertoires, along with all other components of social stocks of know-
ledge, differ from society to society; however, some repertoire of genres
will be found in all societies. The social distribution of genre-related
knowledge may be just as equal or unequal as that of any other part of the
social stock of knowledge. An essential element of genre-related know-
ledge is knowledge about its appropriate use, including knowledge about
alternative options and the degree of constraint for the employment of a
particular genre in a particular situation. Evidently, this kind of know-
ledge goes far beyond the formal mastery of communicative codes, or for
that matter, of competence with respect to the internal structure of genres.
In other words, the use of genres is normally linked to clearly defined
types of social situation. A given genre may never appear in one type of
communicative situation, rarely in another, frequently in still another,
and always in some. There may be situations in which an individual is
forced to use a particular communicative genre, others in which he is
merely likely to do so, and still others in which he will rigorously avoid its
use. This is evidently of considerable significance for the role which
genres play as ready-made 'plans' in relatively unproblematic communi-
cative situations.

Social institutions and communicative genres

If social institutions are routinized, more or less obligatory solutions for
elementary problems of social life, regulating functionally clearly defin-
able kinds of social interaction (such as production, reproduction, the
organization of power, etc.), communicative genres may be said to offer
solutions for specifically communicative problems.
It hardly needs to be pointed out that it is often difficult to draw an exact
line between the two kinds of problems. They are closely interwoven in
182 Thomas Luckmann

human life. The elementary social problems are always also a matter/or,
and often even a matter o/, communication. But these matters are first and
last something other than communication: they are things to be done
rather than things to be talked about.
There are many instances where social institutions and communicative
genres intersect. This is the case wherever talking is a constitutive part of
the resolution of elementary problems of social life. Of course, social
communication itself is an elementary problem in human life - but it is a
problem which in human societies underlies and overlies all other
problems. It is not solvable by a specific set of institutions.
The elementary function of communicative genres in social life is to
organize, routinize, and render (more or less) obligatory the solutions to
recurrent communicative problems. The communicative problems for
which such solutions are established and deposited in the social stock of
knowledge tend to be those which touch upon the communicative aspects
of those kinds of social interaction which are important for the mainten-
ance of a given social order. Of course what is important differs from one
kind of society to another, and different societies therefore do not have the
same repertoire of communicative genres. The communicative genres of
one epoch may dissolve into more 'spontaneous' communicative pro-
cesses, while heretofore unbound communicative processes congeal into
new genres within the history of one society. Nonetheless, because of the
essential similarity of the human condition beneath the widest variety of
ecological, socio-economic, technological, and cultural circumstances,
cross-cultural and historical comparison shows that communicative
genres are a universally important organizational principle of social
communication and reveals, beyond that, certain similarities in their
specific historical forms. Some genres, in somewhat variable forms, may
be universal.
At any particular time in any particular society the repertoire of
communicative genres constitutes the hard core of the communicative
dimensions of social life. The patterning - with varying degrees of
freedom and constraint - defines the situations in which a specific genre is
to be employed, the 'external' (gender-, age-, kinship-, class-, etc.) status
of those who are to be participants in the communicative process, and
their 'internal' status (as speaker, listener, etc.). It preselects the linguistic
(lexical, prosodic, etc.), paralinguistic, mimetic, gestural, etc. repertoires,
co-determines the selection of topics and styles ('formal', 'informal',
ironic,24 etc.), and affects the management of turn-taking. Whenever
genres are employed, communicative 'production' and 'reception', in
their dialogical relationship, are thus not only constrained by the com-
municative codes and the general etiquette of communication but are also
 Interaction planning - communicative genres 183

additionally pre-patterned by the genre model. The expectations of the

participants are moulded by the knowledge that communicative processes
with certain functions in certain situations will proceed in a certain
manner. Genres are 'real' inasmuch as they are 'real' for the participants -
that means, when knowledge pertaining to them is sedimented in the
social stock of knowledge and has been transmitted to the actors on the
social scene.25 Knowledge of a genre need not be explicit 'theoretical'
knowledge of rules; it suffices that one is acquainted with their practical
operation.

Conclusion: genres as communicative plans

A recurrent, in fact, a universal communicative problem of varied severity
which, in communicative processes, logically and temporally precedes all
other potential problems is the 'reciprocal adjustment of perspectives'. It
is the prototypical problem of 'anticipatory interactive planning'. It is a
problem which does not normally become acute if clear indications of the
perspectives and of the knowledge and interests brought into the situation
are reliably available to the participants, or if the participants - or types of
participants - are well known to one another as individuals or as social
types from the same - or the same type of - communicative situation,
where the perspectives have been adjusted beforehand, the 'planning'
having been done in advance. Whatever else it may do for the solution of
specifically communicative problems26 - the use of genres in communi-
cative interaction, combined with knowledge about the social distribution
of genre-related knowledge, serves a double purpose with regard to the
'reciprocal adjustment of perspectives'. It provides easily recognizable
indications for the initial mutual adoption of a communicative perspec-
tive for a definable kind of communicative interaction. And it helps to
maintain the interlocking perspectives in its subsequent production and
reception.
Communicative social interaction originates in action projects which
are based on anticipations of the future actions of others - which, of
course, are assumed to be based on corresponding typical anticipations.
In everyday social practice, the infinite regress of these 'loops' is resolved
by a variety of means, among which shared typifications (derived from the
social stock of knowledge or built up in past social interactions27),
institutions, and institutionally defined social roles are the most import-
ant ones. With respect to the communicative dimension of social interac-
tion it is the communicative genres which play a particularly important
role by providing 'communicative plans' for 'anticipatory communicative
interaction planning'.
184 Thomas Luckmann

Notes
1 Whenever human social interaction is not systematically considered in a
comparative biological frame but in a comparative historical, i.e., socio-
cultural frame this is a useful restriction. It distinguishes motivated, goal-
oriented action from 'mere' human behaviour and includes both overt action
gearing into the world, which thus becomes observable by others, as well as
'thinking' as a covert form of action. For the sources of these distinctions, and
for further analyses of the constitution of the meaning of action as motive and
project, see Weber (1922: 1-16 [1968: 1-31]), Schutz (1932) and Schutz and
Luckman (1989). For another approach, see Miller et al. 1960.
2 Analytically, social interaction should be distinguished from social action. The
latter is one-sided, the former reciprocal.
3 See the contribution to this volume by Richard Byrne (Chapter 1). See also
Byrne and Whiten (1991).
4 See Schutz and Luckmann (1989).
5 For a discussion of the concept see the Introduction to this volume by Esther
Goody.
6 Presumably this is the meaning of the well-known sentence 'You cannot not
communicate' (Beavin et al. 1969).
7 A term introduced by Mead (1967: 42-51 and 253-60).
8 Useful, although somewhat dated references to the relevant literature may be
found in the subsections of Part I ('Phylogenetic and cultural ritualization')
and Part III ('Ontogeny of primate behaviour') of von Cranach et al. (1979).
9 It seems reasonable to assume that - whatever the specific constellation of
'causes' and whichever time-scale is considered to have been obtained - the
phylogeny of language presupposes a fairly complex intersubjective (i.e.,
involving a certain level of conscious, reciprocal orientation in the production
and reception of communicative behaviour) level of communicative social
interaction. At least in this instance there is a partial ontogenetic parallel: the
emergence of'dialogue' from the 'action dialogue' between mother and child,
to borrow the terms from Bruner (1978). For a parallel phenomenological
analysis of the constitution of language, see Luckmann 1983. For an attempt to
simulate the generation of a 'stored system of signs' (with the assumption that
an analogous procedure would yield an elementary grammar) from a mini-
mum set of assumptions concerning the production and 'interpretation' of-
potential - signs, see the contribution to this volume by Edwin Hutchins and
Brian Hazlehurst (Chapter 2).
10 For a general discussion of routinization, institutionalization and the forma-
tion of social stocks of knowledge, see Berger and Luckman (1966:47-91). Cf.
also Luckmann (1982). An account on the formation of kinship terminologies
in the social stock of knowledge of gatherer-hunter societies, and some
thoughts about their role in 'early anticipatory social interaction planning' is
offered by Nurit Bird-David in this volume (Chapter 3).
11 Major parts of this and the next two sections are adapted from Luckmann
(1992).
 Interaction planning - communicative genres 185

12 Universal in human societies - but not in every process of human communica-

tion. On this point I disagree with Bakhtin who maintains that all human
speech is genre-bound: 'We speak only in definite speech genres, that is, all our
utterances have definite and relatively stable typical forms of construction of the
whole' (Bakhtin 1986: 78).
13 Social structure and communicative codes are neither god-given nor nature-
given nor 'autopoetic' realities. Although they are objective components of
human life, they originate in human activities. On the inter subjective consti-
tution and social construction of language see Luckmann (1983).
14 The literature dealing with these matters is enormous and rapidly growing.
My discussion of it in the 'Soziologie der Sprache' (Luckmann 1979) is
already outdated.
15 The original impetus for, and the main contributions to, systematic research
into 'language in use' have, of course, come from the ethnography of
communication and related fields. For an early collection see Gumperz and
Hymes (1964); see also Bauman and Sherzer (1974).
16 For a discussion of these concepts see Rommetveit (1990, 1992).
17 This was a key concept in Alfred Schutz's phenomenology of the social world.
For a detailed presentation of the general thesis of the reciprocity of perspec-
tive see Schutz and Luckmann (1973, esp. pp. 59 ff.).
18 To expand the meaning of these terms beyond the very useful technical sense
which they have in conversational analysis. For that, see Sacks et al. (1974),
Sacks and Schegloff(1979), Schegloffe* al. (1977).
19 For a detailed analysis of the structure of action, of projects, routinization,
social interaction etc., see Schutz (1962); Schutz and Luckmann (1973: ch. 4)
and Schutz and Luckmann (1989: ch. 5).
20 Different fields, among them especially the psychology of speech planning,
discourse analysis, the 'new' rhetoric, and the ethnography of communication,
have produced substantial bodies of literature on different aspects of these
issues. In some fields there is a tendency to limit analysis to individual
psychological perspectives. In others the problems are formulated in relation
to social interaction. Occasionally the concept of 'genre' is used to good
purpose. By way of illustration I mention a few examples: on rhetoric,
Atkinson (1984); on communicative etiquette, Henn-Schmolders (1975)
Laver (1981), and Brown and Levinson (1978); on speech planning, G. Beattie
(1979) and the systematic analysis of Keenan (1977); on stylistic and rhetoric
choice, Enkvist (1984); on individual 'assembly' despite the use of'formulas',
Coulmas (1981b).
21 With few exceptions the discussion of genre 'size' concerns literary genres.
See, especially Jolles (1972), Jauss (1972), Nies (1978).
22 For a typology related to 'anticipatory interactive planning' in the context of
turn construction see the contribution to this volume by Paul Drew (Chapter
5).
23 I have treated the function of genres more extensively in earlier papers
(Luckman 1986, 1989).
24 On irony in social interaction see the contribution to this volume by Penelope
Brown (Chapter 7). On formality, see Irvine (1974).
186 Thomas Luckmann

25 To paraphrase W.I. Thomas: 'When men define their situations as real they
are real in their consequences' (Thomas 1960).
26 This will depend on the specific functions of the communicative interaction.
An adequate identification of such functions presents serious difficulties. See,
for example, Bascom (1954).
Theoretical speculation was rarely supported by detailed systematic
investigation. In recent years, a close look was taken at the 'reconstructive'
function in a research project (sponsored by the German Science Foundation
in the Department of Sociology at the University of Constance) directed by
Jorg R. Bergmann and myself. First results were published in a series of
papers by Jorg Bergmann, Angela Keppler, Hubert Knoblauch, Bernd
Ulmer, and myself. The joint publication of Reconstructive Genres will be
published by de Gruyter, first in German (Berlin 1994), then in English (New
York, probably 1994). For a full-fledged study of gossip see Bergmann (1993).
27 For a general consideration of the link between sedimentation of past
experience and projects of action see Schutz (1962). Cf. also the contribution
to this volume by David Good (Chapter 6).
PART IV

Expressions of a social bias in

intelligence
 DAVID ZEITLYN

9 Divination as dialogue: negotiation of

meaning with random responses

In her introduction Goody proposes AIP as a set of fundamental

assumptions pervasive in human society. In this chapter I shall examine
some of the implications of AIP for the subject of divination.1 Divination
is an extremely widespread phenomenon which illustrates the 'dialogic
template'. (It has also been used to illustrate communication without
intention (Du Bois 1987).) The AIP template underlies the adjacency pair
organization which structures conversation. To speak is to assume the
possibility of a response. In particular, to question is to assume the
possibility of an answer. Goody (Chapter 10, this volume) discusses
prayers which may receive non-verbal replies, but from the point of view
of the believer the prayers are answered.
Divination provides a means of asking questions. The questions are
usually those which cannot be answered by other means available.
Examples are questions such as 'Will my child recover from this illness?',
'Will there be an accident on the road if I travel tomorrow?', 'What is the
cause of this misadventure?' and so on.
It is possible to do divination as a game, as a procedure without any
cognitive or emotional load being carried. However, such cases (which
occur both in Europe and elsewhere) are aberrant. They point to the usual
perceived purpose of divination: to find answers to questions. Generally,
divination is used as a means of resolving problems. Some corollaries of
this will now be examined. If we view divination as the pursuit of answers
to questions we can apply some of the insights of conversational analysis
to understand the details of a divination session.
Divination must be recognized as being a means of obtaining true
answers. Only if it is held to be a truth-telling exercise is it worth
performing. Not that this rules out the possibility of cynical manipula-
tion, but even that eventuality assumes the belief in the procedures of
those being duped.
Most classifications of divination contrast divination by possession
with technical divination. In divination by possession a spirit or other
agency uses a human to communicate directly with people. Questions

189
190 David Zeitlyn

may be put and answers received. Sometimes speech may be uttered

unprompted which must then be interpreted to find the answers to the
problems at issue. On the other hand, technical divination is not, in
general, concerned with the state of the diviner. True answers result from
the correct performance of a set of procedures. Diviners are technicians.
In my earlier work I have stressed that technical divination differs
importantly from divination by possession since it alone allows for the
possibility of'faulty' or 'incorrect' practice. However, the presumption of
AIP means that clients and practitioners view technical divination in the
same light as possession, despite this difference.
The detailed case which follows considers contradiction management
in Mambila 'spider' divination. Mambila ensure (by their own lights) the
truth of their divination by regular inductive testing, and by requiring
consistent answers to repeated questions.
Moreover, communicative constraints on the praxis of divination are
found in the management of contradiction. If we apply the 'dialogic
template' to the interaction between diviner2 and divination (whatever
type of divination is being used) then contradiction becomes an interest-
ing test case. Contradiction may be a threat to the continuing dialogue. If
it is perceived as such then Grice's cooperative maxim is no longer in play
and the dialogue breaks down.
McHugh (1968) gives examples of experiments in which students were
being counselled via an intercom, and receiving only yes/no answers to
polar questions. When the students perceived (accurately) that the
answers were in random sequence and were not 'generated as answers to
their questions' they stopped in disgust. Garfinkel (1984) has discussed
the same experiments and has analysed some of the reasons why the
participants were so reluctant to recognize the randomness of the replies.
The situation is directly analogous to Mambila divination as I show
below. The point becomes critical when a contradiction occurs. When the
same question first receives a 'yes' and then a 'no', a problem arises for the
diviner as well as for us as analysts. Justification for this claim is that it is
just where we as analysts perceive a problem that the line of questioning
changes tack.
Although Goody has talked about a 'dialogue template', this should not
be taken to imply a fixation on dyadic interaction.3 Diviners divine to
satisfy their peer group of fellow diviners as well as their clients. They
must be judged to have really been possessed or to have performed the
techniques correctly, or to have correctly interpreted the signs. Interpre-
tation occurs both between divination and diviner as well as between
diviner and client. My earlier discussion of classifications of divination
sought to draw attention to the way in which the division of interpretative
labour between these two affects the manner of divinatory practice
(Zeitlyn 1987, 1990a: ch. 3).
 Divination as dialogue 191

Garfinkel's examination of 'Commonsense knowledge of social

structures'
Garfinkel's paper presents an account of an experiment and an analysis of
it, followed by his reflections on this analysis. The experiment involved
American university students being counselled through an intercom. The
students were told to pose questions in an either/or format. They received
yes/no answers. The questions, and the students' 'reflections upon the
answers they received', were tape-recorded. What the students were not
told was that the yes/no responses were supplied in a predetermined
random sequence. Each student received the same sequence of answers.
Garfinkel quotes two sessions in full and then summarizes the results in a
set of aphoristic comments.
I quote a short passage from the first case, in which a male Jewish
student debates whether he should continue to date a non-Jew in the face
of his family's displeasure.
SUBJECT: My question is, do you feel under the present
circumstances [which he has just explained] that
I should continue or stop dating the girl? Let me
put that in a positive way. Do you feel that I
should continue dating this girl?
EXPERIMENTER: My answer is no.
SUBJECT: NO. Well that is kind of interesting. I kinda feel
that there is really no great animosity between
Dad and I but, well, perhaps he feels that greater
dislike will grow out of this. I suppose or maybe
it is easier for an outsider to see certain things
that I am blind to at this moment. I would like to
ask my second question now.
EXPERIMENTER: Okay.
SUBJECT: DO you feel that I should have a further discussion
with Dad about this situation or not? Should I
have further discussion with Dad over this
subject about dating the Gentile girl?
EXPERIMENTER: My answer is yes.
This brief extract suffices to show how the subjects constructed for
themselves a meaningful dialogue. The subject took the random res-
ponses of the experimenter as answers to the questions put and imputed
an argument to the experimenter. Subjects behave as if they were
negotiating meaning in the manner found in ordinary conversation.4
However, in this experiment, just as in divination, the responses are not
intentionally produced answers but are 'mere events'. It is this similarity
which makes Garfinkel's discussion relevant to the analysis of divination.
My concern is with the process of constructing an argument, or of
imputing meanings.
192 David Zeitlyn

Although few anthropologists would wish to use such cold-blooded

'experimental' methods of research, Garfinkel's techniques can help the
analysis of the praxis of divination.5 Eleven of his conclusions which are
applicable to the analysis of divination are listed below (some are
quotations, others have been rewritten in an attempt to render them more
readily comprehensible).6 The 'answerer' to whom I refer is not Garfin-
kel's counsellor but the divination.

After Garfinkel: eleven principles of divinatory interaction

(From Garfinkel 1984: 90-1, sections D and E)
1. Questions may be retrospectively redefined in the light of later
answers.
2. 'The identical utterance is capable of answering several different
questions simultaneously and of constituting an answer to a
compound question that in terms of the strict logic of proposi-
tions does not permit either a yes or no or a single yes or no.'
3. One response can be understood to be answering several previous
questions simultaneously.
4. 'Present answers provide answers to further questions that will
never be asked.'
5. 'Where answers are unsatisfying or incomplete, the questioners
are willing to wait for later answers in order to decide the sense of
previous ones.'
6. Incompleteness is attributed by questioners to deficiencies in the
method (of yes/no answers), or to incomplete comprehension on
the part of the answerer (i.e. the divination) which can in turn be
attributed to poor question-setting.
7. Reasons are assumed to exist for 'inappropriate' answers. Those
reasons explain the answer which is given, and determine its
sense.
8. 'When answers are incongruous or contradictory, subjects are
able to continue by finding that the 'adviser' has learned more in
the meantime, or that he has decided to change his mind, or that
perhaps he is not sufficiently acquainted with the intricacies of
the problem, or the fault was in the question so that another
phrasing is necessary.' In other words, explanations or excuses
can always be constructed ad hoc.
9. 'Incongruous answers are resolved by imputing knowledge and
intent to the adviser.'
10. Contradictions force reinterpretation of the questions: further
meanings to the questions are imputed which explain the
answers, thus removing the contradiction.
 Divination as dialogue 193

11. Contradictory answers lead to a 'review of the possible intent of

the answer so as to rid the answer of contradiction or mean-
inglessness, and to rid the answerer of untrustworthiness'.
Points 7-11 are particularly relevant to the analysis of contradiction in
divination which follows.
Garfinkel also provides some maxims concerning the questioners'
suspicions of the system.
12. The possibility of random answers may be considered by the
subjects but is not tested. Suspicions are allayed if the answers
'make good sense'.
13. Suspicion turns the 'answers' into 'mere events' and there is then
no point in continuing. Therefore,
14. Those who became suspicious are unwilling to continue.
During my fleldwork with Mambila in Cameroon, about which more
below, I never encountered such express doubts about divination. I
suggest that this is due to the intellectual protection given to the basic
assumptions, a suggestion in accord with the argument presented by
Evans-Pritchard (1937). These points are therefore complementary to
Evans-Pritchard's twenty-two reasons why the Zande do not perceive the
futility of their magic (1937:475-8). These are a catalogue of ad hoc
hypotheses used to protect the validity of the central tenets, in this case
that of divination itself, whereas the results of any individual divinatory
session remain open to question. The creation and interpretation of a
sequence of questions and answers is independent of any actions
employed to protect divination, such as those listed by Evans-Pritchard.
There is neither methodological nor logical contradiction between
Evans-Pritchard's analysis, and that of Garflnkel.

Contradiction and re-interpretation

The second case used by Garflnkel to illustrate his argument includes an
example of re-interpretation; remarkably, he does not comment upon this.
The 'experimenter' (Garfinkel's term) should give yes/no answers to the
questions put. A student of physics is considering whether to leave college
or to change his subject. His first question is whether he should change
subjects (answer: 'No'). Eight questions later he asks: 'Will I get a
degree?' Answer: 'No'. I quote the questioner's response to this in full
(Garfinkel 1984: 87):
STUDENT: According to that I won't get a degree.
What should I do?
Are you still there?
EXPERIMENTER: Yes, I am.
194 David Zeitlyn

Consider the likely reactions of the subject if the experimenter had

replied 'My answer is no'. It is clear that the experimenter has answered
only the second of those two questions, and that he has stepped out of his
'experimental' framework. It is important to understand why the crisis of
faith which occasioned this exchange occurred when it did. When the
question 'Will I get a degree?' was answered in the negative the subject
perceived the experimenter to be contradicting himself. By his negative
answer, the experimenter was understood to be committing himself to the
proposition
A: YOU will not get a degree in physics.
This was perceived to contradict an earlier answer wherein the experi-
menter was understood to commit himself to the proposition
B: YOU should not change your subject.
If one assumes that
c: If you will not get a degree in physics then you should change
your subject,
then B entails
D: YOU will get a degree in physics [if you persevere].
But from above,
A: YOU will not get a degree in physics.
By asserting B, the subject is allowed (granted C) to infer D. But A and D
are contradictories.
To invoke a Gricean relevance principle: if the questioner is not going
to get a degree then it is irrelevant what subject he studies. The 'relevant'
answer to the question 'Should I change my subject?' would be: 'You will
not get a degree'. This would constitute a rejection of the question.
However, the schema of the dialogue explicitly permitted only yes/no
answers. In order to make sense of the answers received, the subject
interpreted contradiction as a rejection of the question. This is a common
conversational ploy. For example:
Shall we go to the pub?
Have you looked in your diary?
A question is rejected by answering it by another question. In order to do
this the subject had to disregard the rules which had been presented. The
fact that only yes/no answers to polar questions were possible was
ignored, or perhaps subtly re-interpreted. The subject is now in the
position of saying 'When I get answers like this they don't really mean yes
and no, they mean something altogether more complicated ...'
 Divination as dialogue 195

Conversational analysis encourages the detailed examination of the

negotiation of meaning between speakers. Such negotiation is particu-
larly apparent where misunderstanding or disagreement occurs. We can
draw on the literature on 'repairs' for examples of the negotiation of
meaning. A 'repair' occurs after normal turn-taking rules of dialogue
breakdown. This is often caused by the speakers having different assess-
ments of what is being said. Achieving sufficient consensus to continue
the dialogue necessitates negotiation of meaning (see Levinson 1983: 16-
19, 39-47; Schegloff et at. 1977). Conversational analysis can reveal
instances in divinatory discourse where, in order to make sense of an
utterance, context is strongly implicated.7 It is clearly necessary to
consider this context in order to understand the utterance. The perspec-
tive of the analysis is perforce widened to include not only divination but
also the circumstances of the participants and the social structure within
which the divination is practised.
An ethnomethodological focus on the negotiation of meaning between
speakers is of more assistance than formal logic in understanding the
manoeuvres adopted, for example, in response to contradiction. Formal
logic identifies a contradiction but allows no other solution than the
rejection of a premise. It cannot, however, suggest which premise is at
fault. Ethnomethodology, on the other hand, identifies the redefinition of
a premise as a constructive solution to the problems caused by contradic-
tion. The empirical techniques of conversational analysis allow for the
identification of the premise at issue, and for the study of the process of its
redefinition. This seems more consistent with the questioners' responses
as evidenced in the data.
In the cases of both divination and Garfinkel's experiment, analysis of
the negotiation of meaning is facilitated since in both, one of the parties to
the perceived conversation is not in fact negotiating. Paradoxically,
divination can be seen to say so much precisely because it is mute.

Studying divination
Many reasons may be given to warrant the study of divination. I shall
briefly present two of these, and go on to distinguish between two major
classes of divination. A consideration of the limitations of classical
sociological analysis of divinatory practice will serve as a prelude to my
presentation of the Mambila material.
Evans-Pritchard sought to convince European readers that the Azande
were rational to persist in their beliefs, and that their actions were
therefore subject to rational explanation. In particular, he expressly
addressed himself to 'political officers, doctors and missionaries in
Zandeland, and later to Azande themselves'. Divination provided an
196 David Zeitlyn

excellent subject with which to challenge colonial prejudices. More

recently, divination has featured as a leitmotif in the 'rationality debate'.
Of its contributors however, only J. Beattie (1964, 1966 and 1967) and
Horton (1967 and elsewhere) have published works about divination per
se. Divination has figured so importantly because it is perceived as a
paradigm for 'rationality in irrationality': that is, belief in divination is
held to be irrational, but its practice is extremely rational according to the
ethnographies. It therefore serves as an amenable synecdoche of religious
belief and practice.
A second reason for studying divination is that it reveals the actors'
understanding of their social structure. The process of posing the
questions, by their phrasing, and by the range of possible solutions
proposed, provides evidence about indigenous models of the world.
Divination thus ranks alongside disputes as a social activity whose study
can provide information about, and understanding of, much broader
matters than the stated topic of analysis. Thus, although divination may
occupy relatively little of the attention of a group, as with the Azande, it
may provide a rewarding starting-point for a wider analysis.

Classes of divination
Many different classifications of different types of divination have been
proposed. Some typologies cover more comprehensively than others the
wide range of activities which may be glossed as divination. Whether they
are sociologically revealing is another matter. Without entering into the
arguments here I shall follow Cicero in making a distinction between
'artificial' and 'natural' divination (De Div. 1 .vi. 12). Later authors prefer
the terms 'mechanical' and 'emotive' (see Devisch 1985; Vernant 1974;
and Zeitlyn 1987 for further discussions.8
Natural, or emotive divination depends on the recognition of a direct
relationship between the operator and some occult force or spirit, such
that truth is achieved through contact with spirits or by exercise of
'intuition'. It typically involves some sort of 'possession' (this is further
discussed in I.M. Lewis 1971).
By contrast, artificial divination aims to reveal truth through the
performance of a variety of technical operations, all of which are mechan-
ical in nature. The divination practices used byMambila are exclusively of
this kind, and are the subject of my discussion here.9 Unlike emotive
divination, technical divination appears to involve much clear ratiocina-
tion, and its results are open to question in quite different ways. Although
practitioners of any type of divination can be accused of deceit and fraud,
only technical divination can be performed 'incorrectly', thus allowing
the possibility of mistaken practice. For in the case of emotive divination,
 Divination as dialogue 197

the truth of the divinatory results is guaranteed by the possessed state of

the diviner. Since possession is an unequivocal state, mistaken practice is
impossible.10 Any divinatory techniques associated with possession are
employed simply as preliminaries necessary to attain this state; they are
not means by which the results are obtained.
The factors which I examine in this chapter do not obtain in emotive
forms of divination, although these clearly provide an alternative means
of seeking truth. While focusing on ratiocination I do not intend to imply
that technical divination is the sole means to the end of truth-seeking.11 It
should also be stressed that many kinds of 'technical' divination do not
pose polar questions, and have answers in other than yes/no forms.

The incompleteness of sociological analysis

Sociological analyses of divination have treated divination as a procedure
either for legitimating decisions (Park 1963) or for providing therapeutic
benefit to the consultants (J. Beattie 1964). However their focus is on the
social consequences of the use of divination rather than on the divination
per se. They do not, therefore, consider the possibility, admitted only in
technical divination, of mistaken practice. While such analyses may
reveal important aspects of a divinatory system, the theoretical stand-
point adopted allows for no detailed analysis of the praxis of consultation.
Neither the interaction between diviner and client, nor the interaction
between diviner and divination can be understood from this perspective.
Conversational analysis, however, provides techniques to understand
these interactions.
An analysis of this kind neither precludes nor invites functionalist
arguments, rather it precedes and anticipates them. This is possible since
the results of the analysis of divinatory practice may themselves be further
analysed by conventional anthropological theories which have little or
nothing to say about the details of divinatory practice. The results of our
analysis of these details could be subjected to further interpretation in
terms of any theory of society, for example, in a functionalist, a Marxist or
a dramaturgical approach, for the analysis is independent of these types of
theory. Thus, I regard as a neutral boon that which Gellner criticizes as
sociological myopia in ethnomethodology (Gellner 1975).

Why study divination as situated dialogue?

Ethnomethodology provides productive techniques with which to study
divination. Both the ratiocination involved in producing an answer and
the contextualization of question and answer may be examined in the
detail they deserve. Moreover, by analysing the process of divination we
198 David Zeitlyn

can avoid the reliance on abstract accounts which has previously limited
descriptions, for example, of the Ifa divination practised by Nigerian
Yoruba and neighbouring groups.12 Garfinkel lists some of the ways by
which (objectively random) utterances are endowed with meaning by
listeners so as to construct a sensible dialogue. The study of divination
reveals how a similar process occurs when the participants pose questions
and receive answers. Such a procedure poses problems for conventional
discussions of rationality. AIP strategies work for people. Diviners claim
to have tested similar strategies with their divination systems. What
grounds are there for analysts to call this irrational?

Ethnomethodology and Mambila divination

Divination systems allowing only yes/no answers involve similar pro-
cesses of interpretation, particularly when the divination apparently
contradicts itself. Following a summary introduction to Mambila society
I shall illustrate this point with an account of Mambila spider divination
(further details may be found in Zeitlyn 1993).
The Mambila lie on either side of the Nigeria/Cameroon border, the
bulk of them living on the Mambila Plateau in Nigeria. A smaller number
(c. 12,000) are to be found in Cameroon, especially at the foot of Mambila
Plateau escarpment, on the Tikar Plain. My fieldwork was restricted to
these latter groups, and in particular to the village of Somie. Somie had a
population of approximately one thousand (based on the official 1986 tax
census) at the time of fieldwork. Self-sufficient in food, the villagers have
grown coffee as a cash crop since the early 1960s.
Cameroonian Mambila on the Tikar Plain have adopted the Tikar
institution of the chiefship, yet their social structure otherwise closely
resembles that described for the Nigerian village of Warwar given by
Rehfisch (1972) based on fieldwork in 1953. Nigerian Mambila did not
have the same type of institutionalized chiefship as is found in Cameroon.
In Nigeria, villages were organized on gerontocratic principles, and
largely lacked political offices. Most people in the villages are members of
either the Catholic or Protestant church. However both men's and
women's masquerades are still performed, and cases heard at the Chief's
palace are regularly concluded with a ritual swa-oath.13
Most married men know how to divine, but have varying degrees of
confidence in their own skills. Hence if a problem is serious, it is likely to
be taken to one of the acknowledged experts. In the case considered below
a man, named Wong, in his late thirties went, by arrangement, to divine
with Bi, the head of Njerup hamlet, and an important elder. He is also well
known as an accomplished diviner. In order to be confident of the results
of divination Wong came to Bi in order to have Bi's sanction. Bi could
 Divination as dialogue 199

correct any mistakes of interpretation, and thereby help ensure the

accuracy of the results.
Mambila spider divination entails the posing of questions couched as
binary alternatives, often requiring yes/no answers.
A hole in the ground inhabited by a spider is covered by an enclosure,
usually an inverted pot. A stick and a stone are placed within this
enclosure, near the entrance to the spider's hole. A set of marked leaves is
placed over the entrance to the burrow. When questions are posed the pot
is tapped; in response to the knocking the spider emerges from its hole. In
doing so it disturbs the leaves. The resulting pattern of the leaves in
relation to the stick and to the stone is interpreted as an answer to a
question. Questions allow one of two responses, one is explicitly asso-
ciated with the stick and the other with the stone.
Several different spiders may be consulted simultaneously. This
enables a faster rate of questioning since some twenty minutes elapse
before the diviner can check whether the spider has responded to a
question. It also allows a consistency check to be made by asking the same
question of different spiders. Diviners admit that ambiguous or unintel-
ligible answers are possible, but few such instances were observed.
Table 9.1 shows the response to contradiction which arose during a six-
hour divination session recorded in January 1987. The divination con-
cerned a child (Wong's daughter) suffering from malaria. The main
points at issue were whether the illness had been caused by witchcraft,
and whether the taking of a sua-ozth would be sufficient to protect the
child from further attack. The table shows the questions addressed to two
different spiders, and the answers received. In each case the alternative
said by the diviners to have been chosen has been marked with asterisks.
Forty-two questions were posed during the session; their order is
indicated by the question numbers.

Responses to contradiction
Table 9.1 contains answers which directly contradict one another. The
acceptance of direct contradiction is, according to the canons of tra-
ditional logic, a symptom of 'illogicality'.14 Further comment is
warranted.
The sequence starts with Question 33, which was addressed to Pot 1:
will sua end the problem or not? A straightforward yes/no response was
sought. Another pot (Pot 2) was asked a similar question (Q 34) before the
response was obtained from Q 33. The response to Q 34 was taken to
advocate the use of sua, as opposed to other sorts of treatments. This was
immediately followed by Q 35 which repeated Q 33. The response to Q 35
was that sua would not end the problem. This contradicts Q 34. The
200 David Zeitlyn

Table 9.1. Divination questions

Potl Pot 2

Q33:
sua will end it Q34:
vs. *s«<z will end it*
*sua will not end it* vs.
divine further/cut kare1
Q35:
sua will end it
vs.
*sua will not end it*
Q36:
male witch
vs.
*female witch*
Q37:
Q38: *sua will end it*
Something buried2 vs.
vs. witchcraft continues
*sua will end it*

Notes:
1
The kare rite is sometimes referred to as a variety of
sua. It is a domestic version of the sua-oath.
2
'Something buried' refers to some witchcraft treat-
ment, which unless detected and removed would
continue to act although its perpetrator might be
caught by sua.

problem was then assumed to be one of witchcraft. In response to this,

Q 36 sought to identify the sex of the witch. The divination was taken to
have identified a female witch. The response to Q 36 was understood to be
identical with an earlier diagnosis of 'problems among the women in
Wong's house' (Q 26 and Q 29). There had in fact been a long-standing
quarrel between Wong's wife and his classificatory sister about the
usufruct of afield.Disputes over land tenure are typical of cases in which
swa-oaths are taken. The parties to the dispute swear that, although they
may be quarrelling, they bear no malice and will not seek to win their case
through witchcraft, for example by causing illness among the children of
the other litigant. The identification of the female witches as being those
women embroiled in that dispute was assumed, and therefore not tested
further by divination. This resolved the dilemma posed by the contradic-
tion. Once the diviners are assured that the witchcraft referred to is only
that connected with this quarrel, then sua becomes an appropriate and
sufficient course of action. When that reassurance has been given they can
return to the previous line of questioning. The earlier question was then
 Divination as dialogue 201

repeated in a modified form: will sua end it, or is there other witchcraft to
be dealt with, for example, in the form of buried treatments which remain
active until discovered and destroyed (Q 37)? After putting this question,
the response to Q 33 was sought by inspecting the pot. The answer found
was 'sua will not solve the problem'. This was immediately pursued in the
light of the question which had just been put (Q 37), as to whether any
witchcraft was to be dealt with. The diviners understood this to indicate
that there might be further witchcraft. Hence Q 38 draws the distinction
between buried witchcraft substances and the ending of the affair by sua.
The responses to both Q 37 and Q 38 indicated that performance of the
sua rite was the appropriate action to be taken. Thus a believable, because
consistent, result was obtained.
The contradictory results preceding this were thenceforth ignored.
They had, however, forced the diviners to examine the possibilities of
more complicated problems. Once these possibilities had been eliminated
the diviners could return to the main strand of the enquiry as if no
contradiction had occurred.
We must take seriously the diviners' assumption that the sequence of
questions is a dialogue between divination and diviner. Mambila diviners
talk of asking (tie) questions of divination (nggam) as if it were a single
entity. Looking at a result they say 'Divination says ...' {Nggam je ...). I
was always given inductive and empirical justifications for the veracity of
divination. Even formal initiation into the technique of spider divination
contained no information about the origin of divinatory knowledge, nor
was any account given of what divination was, or how it worked. The
belief in the efficacy of the technique is held to be warranted by the success
of the diviners.
If the process of divination is, in part, to be regarded as a dialogue
between diviner and divination, then contradiction may be regarded as a
rhetorical device used by the divination to make the diviner cast the net of
his questions more widely. In the example illustrated, the divination has
forced the diviners to consider the possibility, previously not addressed
by them, that buried witchcraft substances may be responsible for the
child's illness. Garfinkel's methods may be used to reveal the way in
which diviners construct the dialogue. In essence: contradictions were
understood as question-rejecting moves. They give pause for thought,
and lead to changes in tack. If we return to Garfinkel's maxims listed
above, numbers 7 to 11 may be summarized in a single maxim: 'The
problem of contradiction may be defused by treating it as a rejection of the
question.' Evidence for the validity of this position comes from our
success at reconstructing the observed dialogue, which otherwise remains
obscure.
It may be objected that the contradictions are only there from our
202 David Zeitlyn

analytical point of view. However, 'crises of faith' or 'changes in tack', or

other breaks in the flow of dialogue, occur at or following the points where
we identify a contradiction. This demonstrates that they are more than
etic constructions. From an emic point of view no contradiction may be
perceived, but it is clear that the actors have been given 'pause for
thought'. I would suggest that since conversation assumes the absence of
contradiction, speakers then strive to preserve the conversation by
removing the contradictions.
A cynical account of this divination would be that since the perfor-
mance of sua is the standard response to many problems, it is only to be
expected that a swa-oath will be taken when there is concern about an
illness. The process of divination would then be seen as an empty
validating act whose outcome is known in advance. But in divination all
the participants know the background and the likely results. Although I
am sure that the participants would have admitted that sua was a likely
result, I nonetheless reject such an approach. It allows no room for
analysis of the actions and, most particularly, the ratiocination of the
diviners is not considered. A similar objection applies to those analyses
which see divination as a means of increasing psychological comfort by
reducing stress (Park 1963). That the actors believe in what they are doing
is clear from the attitudes expressed, and from the manner in which
divination is practised. The analyst has a responsibility to be faithful to
their beliefs.
Contradictions and inference merit consideration which is not possible
with conventional sociological analysis. Both chains of reasoning and the
consideration of hypothetical possibilities are involved, and these are
capable of reconstruction, as I have attempted to show above. That some
outcomes are highly probable may be regarded as a measure of the
predictability of the world. The fact that time-worn techniques are
repeated does not mean that they are not chosen with care and delibe-
ration each time they are adopted.
Ethnomethodology and the techniques which have grown out of it,
such as discourse or conversational analysis, provide means by which the
care and deliberation exercised in making choices may be brought to light
and analysed. The propositions by Shaw (1985), Parkin (1979) and
Werbner (1989) that divination is best analysed as dialogue, can thus be
realized.

Conclusion
I have been arguing that AIP forms part of the background to divinatory
practice. It is found in the basic assumption that divination is a means to
finding answers to questions. The management of contradiction exempli-
fies this. Contradictions are transformed out of court, and are taken to be
 Divination as dialogue 203

allusions to unsuspected complexity. Diviners assume that divination

produces contradiction in order to give subtle answers. The premise of
dialogue is never threatened. A repair is performed by the diviner which
makes the response (retrospectively) rational and informative. In divi-
nation (as elsewhere) we may find that we were asking the wrong
questions. This chapter has been examining some of the means by which
we come to such conclusions. Creative repairwork, redefining our own
understanding of both the other party and ourselves is an essential part of
AIP.

Acknowledgements
The initial research on which this study is based was funded by the ESRC (grant
no. A00428424416) and by a scholarship from Trinity College, Cambridge. My
research in Cameroon could not have been conducted without the research
permits granted by His Excellency the Minister for Higher Education and
Scientific Research (R.P. 13/85 and 62/86), and the help provided by his staff.
This chapter was begun during the tenure of a Junior Research Fellowship at
Wolfson College, Oxford and completed while holding a British Academy
Research Fellowship. During this time I have been the holder of a short Junior
Research Fellowship at Wadham College, Oxford followed by a Research
Fellowship at Wolfson College, Oxford. I have also received assistance from the
ESRC (grant no R000233311). I am very grateful for all the help I have received
from these different sources.
Substantial portions of this chapter have already been published in D.
Zeitlyn (1990) Professor Garfinkel visits the soothsayers. Ethnomethod-
ology and Mambila divination. Man (n.s.) 25(4): 654-66. Permission to
reprint these portions has been kindly given by the Royal Anthropologi-
cal Institute.

Notes
Much of the data considered in this paper has also been published as Zeitlyn
(1990b). Further consideration of related questions will be found in Zeitlyn
(1993).
1 AIP may also explain why gods behave like people. Horton has argued that
gods, spirits, deities of all kinds have been ascribed human-like attributes so
that we may interact with them. He argues (1982: 227-38, esp. 237-8) that in
'simple' technological societies human interaction is reliable and predictable
in a way in which the 'physical world' is not. One then becomes a model for the
other. In other words if AIP strategies work with gods then these gods must be
like us in certain respects. Human-like gods may be approached to influence
the world they control. The argument is safely circular: gods are like humans
so they use AlP-based strategies. AlP-based strategies work for the gods so
they must be like humans ... Horton's argument, however, rests on 'analogy'
rather than AIP.
204 David Zeitlyn

2 It applies in a non-controversial way to the subsequent discussion of dialogue

between diviner and client.
3 As was recognized in the Garden of Eden myth the dyad is the minimum sized
unit from which society may develop!
4 There is a telling similarity between this and the manner in which gamblers
interact with croupiers and understand roulette as a game of skill rather than a
game of chance (Oldman 1974).
5 This is quite apart from the suggestion that psychiatry and other therapies are
the major type of divination used in America.
6 It would be interesting to attempt a sociological explanation of the opacity of
the prose of those working in ethnomethodology. Sadly, this seems to have
contributed to the sidelining of the approach, at least within British social
anthropology. Nor has the prose style improved with the years. A recent paper
by Garflnkel and some colleagues is a masterpiece of recondite impenetra-
bility, a fact to which the comments made by Holton, who had the unenviable
job of commenting on the paper when it was presented, mutely attest
(Garfinkel et al. 1981; Holton 1981).
7 The process of such implicature was first outlined by Grice: superficially
uncooperative utterances are taken to be relevant, meaningful contributions to
a continuing conversation by making extra inferences. These presume that the
basic principles of conversation (called Grice's maxims in the literature) hold
at a deep level. There remains uncertainty as to the detailed workings of such
'implicatures' (see the discussion in Levinson 1983: ch. 3). Sperber and
Wilson (1986) have produced a general theory of relevance to account for such
implicatures. This remains controversial (see Sanders 1988; Levinson 1989).
8 Confusion may be caused by inconsistencies of usage between different
authors. Evans-Pritchard (1937: 10-11) contrasts oracles with divination in
order to express the distinction made here between mechanical (or technical)
divination and emotive divination.
9 An anti-witchcraft cult called 'Makka' existed around the time of the Second
World War. Some practitioners of 'Makka5 appear to have been possessed
when detecting witches, although others relied on the administration of
ordeals.
10 A Sudanese counter-example has been pointed out to me in which possession
at the wrong time of day leads to wrong predictions which can be attributed to
a mistake about the time. Nevertheless, I believe that the broad contrast drawn
here is helpful in the analysis and comparison of different divinatory systems.
11 While focusing on ratiocination I do not intend to imply that mechanical
divination is the sole means to the end of truth-seeking (Esther Goody,
personal communication). I should also stress that there are many kinds of
mechanical divination which do not pose polar questions, and which have
answers in other than yes/no forms.
12 1 know of no work on Ifa or on Ifa-type divinations which devotes much
attention to the details of praxis. Maupoil (1943) concentrates on the math-
ematics, Bascom (1969) on the verses. Only de Surgy (1981) includes some
case material, albeit principally as an introduction to the sacrificial sequences
which follow. It is surely the case that sufficient is now known about both the
 Divination as dialogue 205

ese verses and the cult of Ifa. But what actually happens, the way in which the
documented theory is put into practice, remains unstudied.
13 Sua names a variety of ritual oaths, often accompanied by the ritual killing of a
chicken as well as some masquerades. It is the nexus of Mambila religion, and
has been extensively analysed in Zeitlyn (1990a). It is sufficient here to note
that it is both an oath which binds the oath-takers not to cause illness, and a
death threat to any other persons seeking to do evil.
14 Possible responses are discussed in Zeitlyn (1990b: 665).
 ESTHER GOODY

10 Social intelligence and prayer as dialogue

Homo sapiens is the clever hominid: primate social intelligence

plus language
This chapter is part of a long-term project for the teasing out of
implications of a social origin of human intelligence - for of course our
species is distinguished as Homo sapiens sapiens - the intelligent hominid.
Recent ethological research argues convincingly that primates are on
many measures highly intelligent, and that this can best be explained as a
product of social interdependence. Higher primate species appear to have
a progressively greater capacity to cognitively model responses of others.
For convenience this modelling of alternative contingent responses to
others' actions can be termed anticipatory interactive planning (AIP). It is
also argued in the Introduction that human intelligence was intimately
linked to the emergence of language, which fundamentally changed the
ways in which primates managed social interdependence.
Spoken language offers ways of influencing others' behaviour. It
permits more elaborate negotiation of social interactions, the negotiation
ofjoint AIP strategies. It becomes possible to seek cooperation with one's
own goals through the exchange of information ('there are more nuts over
there under the big tree'); and to manipulate relationships themselves
through asking, pleading, begging, pretending, threatening, insisting,
promising. Spoken language became, as it continues to be, central to our
human AIP modelling. As speech presumes an exchange of messages, this
makes dialogue, literally 'speaking alternately', a fundamental form in
hominid sociality. The importance of an exchange of messages for the
emergence of a lexicon, i.e. shared meanings, is modelled by Hut chins
and Hazlehurst in Chapter 2.
Particularly significant here is Vygotsky's argument that conscious
thought is internalized speech (1962). He showed experimentally that
young children speak aloud to themselves the reasoning accompanying
complex actions; the more difficulty encountered in the action, the more
elaborated the spoken accompaniment. This speaking-out of action plans

206
 Social intelligence and prayer as dialogue 207

appears to be a regular preliminary to the internal conscious planning of

older children and adults. Vygotsky's account of learning emphasizes that
the roles of both novice and expert are internalized; in the case of speech it
would be the dialogue template that is internalized by each child.
The symbolic power of language multiplied the levels of complexity in
the mental representations necessary for AIP strategies, particularly
through the capacity for naming, leading to classification. To this is linked
the ability to represent, to ourselves and to others, accounts of how the
world is - of what is happening, has happened, will happen, should
happen - leading to cultural accounts of how to understand, how to
behave and what to teach. Such accounts create a new kind of reality for
past and future time, for intention, obligation, and causality, and for
representations of dreams, the dead and spiritual forces. They come to
have a separate identity as beliefs, myths and religions (see Rappaport
1988; Goodenough 1990).
The ethnomethodological term 'accountability' refers to the fact that
orderly social relations demand that we are accountable to one another for
our actions, that is, where actions differ from what we expect (often what
we feel is our right), they have to be capable of explanation, of repair, and
of negotiation (see especially Heritage 1984). This is possible in an
entirely new way through the medium of language, and several chapters
in this volume explore in detail ways in which accountability occurs in the
course of conversation. Accounts of 'how the world is' set the conditions
for the accountability of our actions. In a sense prayer, the subject of this
chapter, can be seen as an attempt to engage powerful non-human forces
in accountability.

Prayer as dialogue
Some of the puzzles about prayer can be illuminated by seeing it as the use
of social intelligence. For social intelligence seeks to reach goals and solve
problems by modelling the ways in which our actions are contingent on
others' responses. This creates a premise that problems will be resolved
by modelling and managing others' reactions to our own actions. One
might call this a dyadic premise. Social intelligence means that problem-
solving schemata have a slot for modelling the responses of a social Other.
The premise itself is invisible in ordinary social life where problem-
solving concerns real people. The social Other slot is occupied by
grandma, or the milkman, or 'the government'. Occasionally the premise
of a social Other is revealed where an individual is relating to the non-
social world.1 The dyadic premise is so deeply embedded in human
schemata that wefindit difficult to tackle problems such as illness without
positing an Other.
208 Esther Goody

On another level, major problems - illness, death, famine, earthquakes

- force individuals, and communities, to recognize our powerlessness;
just as important, they make us afraid. At the same time these experiences
lead people to wonder and feel awe. But as socially intelligent creatures,
with the capacity to construct concepts around experience, and to
conceptualize cause and effect, we try to construct an account of what is
happening, and to explain why it is happening to us. Who/what is this
power? Why this injury to me} I suggest that it is because of the dyadic
premise of social intelligence that we find explanations based on chance
difficult to accept. We cannot easily incorporate chance occurrences into
an AIP model. Instead, as socially intelligent creatures possessing,
through language, the capacity to name feelings and ideas, when we try to
model strategies for solving grave problems we construct ancestral spirits,
ghosts, and gods to fill the 'social Other' slot.
The particularly human form of pursuing AIP strategies (though
obviously not the only one) is through talk. Thus 'dialogue' with the
social Other slot becomes prayer when we believe that the Other is a
powerful spiritual being.

The problem of evidence

A major problem for the project of looking at the implications of a social
intelligence concerns the possible nature of evidence. This chapter offers
one approach by proposing that prayer may be seen as an expression of the
dyadic premise, and trying to understand the dynamic of prayer in a few,
particularly well-documented societies. What follows is not an exhaustive
study of prayer; certainly not from a theological perspective. Even
treating a few ethnographies in adequate detail would result in a book
rather than a chapter, so the examples used here are necessarily
abbreviated.
It seems that there are two stages, two processes, involved in the
emergence of prayer as an institutional form in a given religious
community.
First, and necessarily prior, is the process in individual social cognition
of seeking to reach goals by negotiation with social Others. The evidence
that prayer involves social cognition would lie in its use in negotiating
goals with objectifications of power, misfortune and benevolence.
The second process is socio-cultural; it concerns the construction of
accounts that explain to us how the world is as it is. When things happen
to us, we explain them in terms required by our AIP contingency models.
Religious beliefs can be seen as one of the forms taken by the stories we
construct to explain things we don't understand - about the universe,
about morality, about misfortune. These stories are also a way of helping
ourselves predict the actions of others, which we must do for AIP
 Social intelligence and prayer as dialogue 209

modelling. Evidence for the influence of social cognition in religious

beliefs and institutions would lie in their incorporation of objectified
Others and in the elaboration of cultural processes both for negotiation
and for the control of negotiation.
The nature of the Other to whom prayer is addressed varies widely
from one society to another. But in every society we examine there is some
form of prayer, of using spoken language to 'make an earnest request or
petition; to make entreaty or supplication; to ask earnestly, humbly, or
supplicatingly, to beseech' a powerful force or being. The 'solemn and
humble request to God, or an object of worship', is 'usually expressed in
words' (as defined by the Oxford English Dictionary). In this sense,
prayer embodies the dialogue mode of negotiating with another; itfitsthe
dialogue template.
As anthropologists our 'laboratory' is ethnographic descriptions of
prayer set in their normal daily context, and these can contribute to an
understanding of the process involved when individuals use prayer to
solve problems. If these processes are at least congruent with an AIP
model of social cognition in the ways outlined above, then they provide
one sort of evidence for intelligence as social.

Prayer as dialogue in preliterate societies

Ancestors in Ashanti
The clearest case for prayer as dialogue can probably be made for ancestor
worship since known dead individuals may be addressed. This can be
particularly obvious at the time of a funeral, and in turn makes evident the
extension to those longer dead, the ancestors. Among the Ashanti of West
Africa, just before the body was carried out for burial, one of the family
members came forward and addressed it:
Today you go
We have fired guns
We have killed sheep
We have brought cloths
We have made a fine funeral
Do not let anyone fall ill
Let us get money to pay for the expenses we have made
Let all the mourners have strength
Life to the chief
Let him beget children
Let all be fertile.
Then en route to the graveyard the head of the matrilineage touched the
coffin with branches, saying:
Kwame [name] I separate your soul from us. (Rattray 1927: 159-60.)
210 Esther Goody

When an elder of the matrilineage dies, his or her personal stool is

consecrated with sacrificial blood and wine and placed in the shrine room
of the lineage. Every forty days there is a lineage ceremony at which
libations are poured to these ancestral stools; at this time the ancestors are
told of lineage affairs and asked to bring peace, health and fertility to
lineage members.
Here we can see the continuity between conversation with a close
kinsman, and the attempt to conduct a dialogue with him after his death.
The dead person is asked not to harm those left behind, and to help them -
to be fertile, not to become ill, and to get money. The elders whose stools
are prayed to in the shrine house are those who controlled the affairs of the
lineage when alive. All important affairs are still brought to them as
ancestors at the forty-day adua ceremonies. In this way they remain in
intimate contact with their descendants. And indeed dreams of the dead
are taken as evidence of their continuing involvement (see Humphrey
1986).
We could stop here, claiming that prayers are effective because for
believers they are dialogues with 'real' superior beings who actually have
the power to aid the supplicant. Ashanti prayers are entirely consistent
with the stories Ashanti tell themselves about the role of ancestors in the
causes and remedies of blessing and misfortune.

Prayer: dialogue in performative ritual

Ancestor worship offers the strongest possible case for the dialogue of
prayer as a straight extension of relations with the living to the realm of
the supernatural. What of other forms of religion? The problem of the
efficacy of prayer is similar to the problem of the efficacy of ritual.
Anthropologists have often asked 'Why are rituals so central to religious
practice when "objectively" they cannot have the effects they are sup-
posed to produce?' It seems obvious that most 'magical spells' cannot cure
illness, ensure rich harvests or success in war. Yet their use has probably
been universal in preliterate societies (and might well prove to be more
widespread in our own than we think). Tambiah (1970, 1985) has
approached this problem through the idea of 'performative force' as used
by Austin and his followers in sociolinguistics (see also Finnegan 1969).
Ritual has its efficacy in and through its performance. One way that this
works is that words in ritual are often used with performative force in
Austin's sense. Speaking them is held to cause the action they refer to. For
instance, cultural and ritual conventions ensure that verbs like 'swell,
grow, be fertile' are understood as commands that bring about their
object. There are much more complicated recursive patterns, like the
rubbing of leaves of a prolific species onto seeds while commanding them
 Social intelligence and prayer as dialogue 211

MOUNTAIN WITH CROSSED YUCCA

V ^VV
ONE-SUNG-OVER |_PJ LIGHTNING SNAKE SANDPAINTING

SINGER [ § ] — . - . . - . -

Figure 10.1 Diagram of hoop ceremonial process. {Source: Gill 1981.)

to germinate. There are direct parallels to this metaphoric use of

performatives in prayer.
Tambiah also holds that the performance of rituals has itself a special
sort of efficacy; the activity of ritual can be seen as directly parallel to the
performative force of words. Verbal and ritual performative force have an
interwoven dynamic.

Navajo prayers as both dialogue and performative ritual

Navajo prayers are not to ancestors but to the Holy People, those who
emerged from the lower worlds and created humans. Why do these
prayers 'work'? Navajo spoken prayers are part of what Gill (1981) calls
'prayer acts' - rituals which summon one of the ancestral Holy People into
the presence of the person or community being prayed for, and which
symbolically act out the assistance required. The pragmatic force of these
prayers comes from the myths of Navajo cosmology familiar to everyone.
Praying calls up Holy People from their other worlds and tells them what
is needed, and through prayers, the desired changes are 'created' - much
as the Holy People once created the Navajo culture and people. All these
worlds continue to exist and interact.
In the Ch'eehoyateeh prayer act the hoop ceremonial takes the singer on
a journey to find the life form of an ill person which has been stolen by
Holy People; the singer returns and restores it to the one-sung-over. The
ceremony is a complex composition of ritual acts, ritual objects, and ritual
speech acts (prayers). A Ch'eehoyateeh prayer is intoned inside the
ceremonial hogan each day at the conclusion of the rite. The singer sits in
front of the one-sung-over and recites the prayer; the one-sung-over
repeats each verse in turn. 'The prayer describes the journey of Talking
God over the obstacles which guard the home of the ugly one [one of the
Holy People] as he moves to rescue the life form of the one suffering from
212 Esther Goody

its captivity there. The return journey is directed to the home of the one
suffering. Its conclusion will result in a return to health and blessing' (Gill
1981: 147).
The prayer journey is parallel to the journey performed by the singer,
who inside the ceremonial hogan moves over feathers, through hoops,
over 'mountains' and finally over the snakes to reach the one-sung-over.
'The hoops ... may represent the levels of the lower worlds through
which the prayer "talks" the one praying. This interpretation... suggests
that the transformation enacted in a hoop prayer ceremony is modeled
upon the emergence paradigm of the transformations which were exper-
ienced in sacred history as the way was made to the earth surface' {Ibid.:
150; see Figure 10.1).
'The other primary ritual object [beside the hoops] is the talking
prayerstick bundle. The talking prayersticks provide the power of
movement in the prayers bridging the gap between the earth surface and
other worlds. Without the talking prayersticks the singer could not pass
through the hoops to rescue the one-sung-over, nor could the one-sung-
over pass back through the hoops to be retransformed' (ibid.: 150-1).
The dialogue form of Navajo prayer is particularly clear in the
ha'ayateeh Liberation prayer. The singer describes the journey of Mon-
ster Slayer, 'armed with his dark staff and lightnings and carrying his rock
crystal and talking prayersticks, as he searches through mountains,
clouds, mists, mosses, and waters for the means of health which had been
abducted and is being detained in the home of the ghosts in the lower
worlds' (ibid.: 143). When he returns to the earth's surface Talking God
reaches the home of the lost one; 'Upon reaching the most interior, and
thus the safest, place in the home, Talking God reunites the lost one with
the things with which he had been formerly associated.' Then Talking
God bestows blessings on the restored person through identification with
him. The prayer concludes by announcing a state of blessing (ibid.:
143-4).
The prayer chants of the ceremonies can be seen as successive
dialogues:
1. Singer recites prayer addressed to one of Holy People
= dialogue between singer and Holy Person.
2. Singer places supplicant in role of addressee
= singer is praying on behalf of supplicant, to Holy Person.
The dialogue form is particularly clear in the Blessing prayer act for an
expectant mother. First the Singer speaks on behalf of the Holy Person to
supplicant:
From the heart of Earth, by means of yellow pollen blessing is extended.
From the heart of Sky, by means of blue pollen blessing is extended.
 Social intelligence and prayer as dialogue 213

To which the expectant mother (echoing the singer) replies:

On top of pollen floor may I there in blessing give birth!
On top of afloorof fabrics may I there in blessing give birth!

May I give birth to Pollen Boy, may I give birth to Corn Beetle Boy, may
I give birth to Long Life Boy, may I give birth to Happiness Boy!
With long life happiness surrounding me may I in blessing give birth!
May I quickly give birth! [etc. etc.]
{Ibid.: 69-70.)
This sequence sets up a complex discourse with the prayer containing
'parts' for the supplicant and then the Holy Person. Both 'parts' are
recited first by the singer and then, verbatim after him, by the supplicant.
First the singer takes both parts, then the supplicant himself/herself takes
both parts. In this process there is a clear identification between the two
main participants, the supplicant, and the Holy Person. The singer never
establishes an independent identity, but acts to blur the boundaries
between the two.
While Navajo prayer acts elegantly illustrate the processes of performa-
tive ritual, they clearly take their meaning - and thus their performative
force - from prayer in dialogue form.

Dinka prayers: priest and congregation

Prayers accompanying sacrificial ritual among the pastoral Dinka of the
Sudan are described by Lienhardt (1961) as 'essentially public forms'.
They are spoken by men of certain agnatic groups who, by virtue of
inherited powers, are known as Masters of the Fishing Spears. The
prayers are more or less efficacious depending on their ability to speak
with 'bite' - forcefully, efficaciously, truly. They speak the truth in two
senses; they do not speak falsehood, and that which they say comes true.
Lienhardt uses the term proleptic: 'the representation or taking some-
thing future as already done'. The prayer of a powerful Master is a
performative speech act in that it causes to be true what is spoken.2
Lienhardt says that in the prayers the subjective experience of sufferers
(from illness, sterility, guilt) is objectified - represented as external. The
Master then addresses this objectified power in prayer and, by speech-
which-becomes-truth, forces it to leave the human victim and to go away,
with the life-breath of the sacrificial animal, to 'the land of those with
lower incisors [i.e. the land of the dead]' (see especially Lienhardt 1961:
ch. 6).
The prayer speech act is given greater force by the symbolic nature of
the sacrificial ritual, and by the echoing of the key phrases by the audience
as a choral response. It is given force in a further sense when deities show
214 Esther Goody

their presence and willingness to help by possessing the men of the

Master's lineage. Lienhardt's discussion of who is being addressed in
such prayers and invocations is subtle and persuasive. Briefly, the Master
addresses the Power, a deity or perhaps a witch, who represents the
affliction the sacrifice is intended to treat; but he addresses also the
sacrificial ox, affirming an identification between sacrificial victim and
sufferer and between victim and Power causing the affliction. Many
things are happening in these rituals. The prayers objectify the source of
the suffering, and separate this from the sufferer himself; they then
describe what will happen to the Power causing the suffering - and as the
words are spoken by the Master whose words become/contain truth, the
repairing events will indeed happen. Through many prayers and invo-
cations directed at the sacrificial animal, and echoed by the congregation,
the suffering is transformed into an emblem of the sufferer and his agnatic
group, and control and departure of the threatening Power is proclaimed
and enacted. The congregation participates with increasing intensity in
the whole process, verbal, rhythmic and constraining, and thus comes to
share in the experience of transformation achieved by the prayers and
sacrifice.
The Dinka provide a critical instance of an unwritten religion in which
prayers are initiated and led by a priest on behalf of a supplicant, and
reinforced by their repetition by the congregation. This pattern is
strikingly similar to that found in the 'religions of the book', where the
priesthood assumes a central mediating role between the individual and
his god. Lienhardt describes the Masters as holding a role at once
religious and political, and stresses their importance in warfare. Much the
same could be said for the archbishops and popes of medieval Catholi-
cism, the elders of the early Protestant churches, the founders of Islam, as
well as contemporary ayatollahs.
The nature of the powers addressed in the prayers of Ashanti, Navajo
and Dinka varies, but in each case they are believed to 'hear' the prayers,
and to respond with actions which affect living people. In the prayer
ceremonies of the Navajo the words and actions of the ritual enact not
only the seeking of assistance from the Holy People, but also their
intervention. In Dinka sacrificial ceremonies there is an added element of
constraint. The power of the Master of the Fishing Spear to 'speak truth'
forces the compliance of the Powers. The constraint arises through the
power of the prayers of the Master. But the power is also in the words as
part of a ritual speech act. Lienhardt notes the similarity between such
prayers and the exhorting of men about to go into battle tofightbravely, as
well as to the cursing of an enemy.
 Social intelligence and prayer as dialogue 215

Monotheistic religions - does dialogue become monologue?

The several monotheistic religions originating in the Middle East shared
a striking difference from these religions of multiple powers; they all
insisted that prayer was the daily duty of every individual. This is basic in
early Zoroastrian religion, in Judaism, in Christianity and in Islam. Of
course priests become central for religious institutional structures (often
linked to states), and to the welfare of believers. But the priest cannot
fulfil the duty of daily prayer which falls on each individual.
With this individual duty to pray, a reflexive element appears in the
relationship of the individual to the High God: God demands not only
certain words and actions from the devotee, but he makes demands
concerning his mental attitude as well. The prophet Zoroaster required
that his followers love the god of light and the right - Ahura Mazda - and
hate and renounce the god of darkness. Good and evil were here cast as
perpetual opponents, battling in the world of the gods, but also in men's
actions and in their hearts. The individual could determine his own
ultimate fate by following the ways of good and renouncing evil (Boyce
1979; Moulton 1913). The Jew is warned not to pray unless he can attain
kavanah, - a sense of standing in the presence of God, and the intention to
fulfill one of His commandments (Donin 1980). Christians are enjoined to
love Jesus, just as they are promised that Jesus and God love them; and
they are exhorted to do the will of God - 'Thy will be done'. Immanuel
Kant3 wrote that 'Praying, thought of as an inner formal service of God...
is a superstitious illusion ... for it is no more than a stated wish directed to
a Being who needs no such information... hence God is not really served.
A heart-felt wish to be well-pleasing to God in our every act and
abstention ... is the spirit of prayer' (1934: 182-3, italics in original). The
core of the Quaker faith is the conviction that God speaks to us in our
hearts, but He can only do this if we have the correct attitude. The quiet of
the Quaker meeting is necessary so that God's words may be heard in each
heart. Both the impulse to speak out in meeting, and the very words
themselves are given by God (Bauman 1983: 12Iff.).
One of the classic debates in western theology concerns whether prayer
is a dialogue or a monologue. Is prayer simply the individual talking to
himself? Clearly for the Ashanti their ancestors are real - have they not
known them in life? But what about prayer in monotheistic religions?
Alhonsaari (1973) says, of Christian prayer, that this is not a monologue
because the one who prays believes he is really addressing the Lord. But
for those who do not share this belief, prayer is a 'false' dialogue; the
person thinks he is addressing a superior bring, but there is no one there.
A better stance would be that this is an intended dialogue; but surely still a
dialogue.
216 Esther Goody

Kant on the other hand argues that we are really confronting our own
consciences when we pray.4 It is a monologue - or at best a dialogue with
ourselves. This corresponds to Lienhardt's framework in which the
powers addressed by the Master of the Fishing Spears are images,
representations, of our own feelings. But the monotheistic religions have
required individuals themselves to address these (culturally provided)
representations. And, further, they have laid great stress on the sincerity
with which this is done. So, by an ironic twist, the daily individual prayers
of monotheism, instead of externalizing passions and conflict, have again
brought the attempt to deal with them inside our own heads.5 It is true
that these sufferings may be labelled as external acts - sins; however a new
source of suffering is added, that of wrong belief, wrong attitude, lack of
faith. At its most extreme the goal of prayer also lies in the realm of
representation of subjective feeling. St Theresa of the Little Way
resolved her conflicts about a Christ who allowed her to suffer by no
longer seeking relief, only the assurance of His love. Her way of achieving
this love was through prayer which allowed her to construct an intensely
personal relationship with Jesus. Thus even where prayer is an internal
'monologue' about subjective experience, it may be explicitly directed to
external powers.

Buddhism: prayer without a god?

Buddhism is identified with meditation, 'the exercise of the mind in
religious contemplation' rather than prayer. Indeed Anesaki (1980
[1918]: 166) claims that since Buddhism teaches that there is no personal
creator or ruler of the world, and since perfection of religious and moral
ideals rests solely with one's own self-perfection, 'in the Buddhist religion
there is no room for prayer, in the sense of a petition or solicitation
addressed to a god'. The Buddha carefully guarded against the use of
prayer (mantra) addressed to a god for the purpose of securing a certain
benefit through his special favour. However as Buddhism later deve-
loped, particularly in the Mahayana form, prayer became important. The
novice took solemn vows (pranidhdna) to one of the manifestations of
Buddha to commit himself to the search for enlightenment. These vows
were believed to be answered by the Buddha with an assurance (vydkar-
ana) that ultimately his efforts would succeed: 'this ... vydkarana is the
necessary counterpart of the vow' (ibid.: 167). Anesaki considers that
'many of these vows were in reality prayers, addressed to the Buddha, as
well as to the universal truths revealed by him' (ibid.: 166). This pattern of
vow and assurance clearly fits the model of a dialogue between believer
and superior being. In one account of Buddha's early training his own
prayer is rendered as:
 Social intelligence and prayer as dialogue 217

Indeed let it be so, that I could be born as one who,

having overcome the world,
would work in the world for the benefit
of the world and should live for the weal
of this world ... [etc.]
Dipankara reassures him:
Thou shalt at a certain future time become a Buddha,
being born as a son to the Sakya clan, and work
for the benefit of men and gods.
(Mahdvastu i:3; ibid.: 167.)
In Mahayana Buddhism all Buddhists are considered to be earnestly
seeking enlightenment, and thus to be bodhisattvas. 'Thus every prayer
addressed to Buddha is at the same time a vow ... vows are taken by a
Buddhist, and assurance given by a Buddha.' One of the most frequently
recited prayers is a verse from the Lotus of Truth:
Let these merits (now performed) universally pervade all,
And let us, together with them, soon realise the life of
Buddhahood.
{Lotus of the True Law (Saddharmapundarika); ibid.: 168.)
The most extreme realization of prayer in Mahayana Buddhism may be
seen in the transformation of the Lotus of Truth into a schematic visual
representation which itself became the object of prayer for the followers
of Nichiren (AD 1222-82).
Prayer is of central importance in Thai Tharavada Buddhism, both in
the life of the monasteries, and for the community. In his detailed study of
the Tharavada practice in northeast Thailand, Tambiah found that
through their withdrawal from secular life and their attempt to realize the
mastery of those things which chain them to earthly existence, the
Buddhist monks are believed to have great sacred power. They use this to
recite the words of the scriptures to do religious 'work' on behalf of the
villagers - to bury the dead, ensure the round of the seasons, and create
merit for future rebirth. The compact with the villagers is 'We will feed
you, but you must pray for us'. The monks' prayers are on each occasion
formally 'invited' by the lay leader of the village congregation, and after
monks receive food, there is a formal response; that is, there is a dialogue
form between the congregation and monks' prayers, as well as between
the monks and Buddha (Tambiah 1970: 133-4).

Prayer and meditation

But in Buddhism it is meditation, not prayer, that is central to practice
and theology. The difference between the two would seem to be that
218 Esther Goody

between 'contemplation of religious ideas' and 'addressing a superior

being, usually through speech'. In meditation the focus is on emptying
the mind of all but the chosen thought in its most abstract sense. There is
no Other, and no dialogue. Anesaki (writing well before the medical use of
tranquillizers) says that meditation is intended to tranquillize the mind:
That disciple should concentrate (panidahitabbam) his mind on a certain thing
(attha) as the condition of tranquilizing; when the mind is concentrated upon that
tranquilizing condition, cheer arises and from cheer, joy arises.
(Samyutta xvii (10); Anesaki 1980 [1918]: 166.)
This 'tranquillizing function' of meditation is important in the rela-
tionship between prayer and anticipatory interactive planning, discussed
below. But despite the theological interdiction on prayer, at the everyday
level Buddhism has often 're-invented' prayer - presumably because the
functions it fulfils for the individual are important ones.

Animism: an early false belief, or a basic paradigm?

Writing in 1871, Tylor (1958) proposed the term animism for the 'deep
lying doctrine of spiritual beings' which he saw as the minimum defini-
tion of religion, and which he attributed to the transformation of images
of the dead in dreams. Marett's animatism, 'a thing, situation, or state of
affairs that is enlivened or animated, but not in any individual, soul-like
manner' partly corresponds to notions of the Navajo kind (1909). Using a
Darwinian evolutionary model, these writers considered primitive reli-
gion to be an early stage, later superseded by more 'developed' religions
until the last 'high' form of monotheism was finally attained. But
theologians as recent as Martin Buber (1970) have seen that even in this
'final' stage we seek to have 'I-Thou' relations with inanimate objects in
nature. And Elaine Scarry (personal communication) finds that even in
twentieth-century litigation there is a tendency to treat objects - for
instance a ship - as morally guilty. Sophisticated western humans still
tend to attribute intentional actions to things with which we are interde-
pendent. We are no more ready to leave all to chance than earlier peoples.

Conclusions
Let me return to the suggestion that there are two processes involved in
the emergence of prayer as a religious form, first in individual social
cognition, and then in the socio-cultural construction of accounts of the
world.
It has been argued that Homo sapiens' use of spoken language funda-
mentally altered the modelling of actions through anticipatory interactive
planning. If this has resulted in the dialogue template as a constant feature
 Social intelligence and prayer as dialogue 219

of individual social cognition, then we would expect tofindit expressed in

all human societies; prayer may be such an expression. But as human
societies vary hugely in their social structures and cultural forms, so we
would expect to find they have constructed different cultural accounts,
different stories, to explain to themselves this universally human exper-
ience of seeking assistance in crisis by 'speaking' to a powerful Other. We
can, and should, ask under what conditions different sorts of accounts
tend to appear. But in looking for the factors leading to particular kinds of
accounts we need not assume any particular evolutionary sequence, only
that certain forms of society in their historical contexts tend to form
certain kinds of representations of non-human powers.
Here it is useful to recall the Buddhist emphasis on meditation as a way
of achieving tranquillity of mind through the focusing of attention. For
Buddhists this may serve as an experiential model of the desired state of
'nothingness' in which the individual merges for ever with the universe.
In terms of a human capacity for modelling problem solutions as
contingent responses to others' actions, deep meditation on 'the absolute'
may finally succeed in eliminating the Other. While a central tenet of
Buddhist thought holds that all external experience is ultimately mean-
ingless, it is also insisted that this cannot be truly understood intellec-
tually, but is realized through mental exercises such as meditation. In our
own terms the mental modelling of AIP strategies is a necessary prerequi-
site to action. Once represented mentally, these strategies must be
negotiated with the social Others they represent. But in meditation,
mental modelling is directed not to action, but to its absence. The goal
ceases to be effective management of relations with others, and the
significance, indeed the validity, of interdependence with others is
denied. For a species in which thought entails plans for contingent action,
this is a profound change. Perhaps the peace which rewards truly effective
meditation comes from disengaging from the constant challenge of social
intelligence.
Prayer, on the other hand, can be seen as an attempt to negotiate goals
with powerful beings in the Other slot. If this is part of what is happening
in prayer, then it suggests that the model of individual cognitive processes
as fundamentally social can help to account for this very pervasive
dynamic in individual behaviour, a dynamic so powerful that it has
shaped, and in the case of Buddhism, reshaped, the major world religions.

Acknowledgements
This paper was written while the author was a Fellow at the Wissenschaftskolleg
zu Berlin in 1989-90. Without freedom from teaching through the generous
Fellowship it would not have been possible to consider, even in this partial way,
the vast amount of material on this subject.
220 Esther Goody

Notes
1 Johnson-Laird and his colleagues studied mental models where even university
students find difficulty with abstract syllogisms although they can solve exactly
comparable problems phrased in terms of actual people and situations (John-
son-Laird and Wason 1977b; Johnson-Laird 1983). Elaine Scarry comments
that there have been legal cases in which a ship is treated as though it had acted
intentionally like a person (personal communication).
2 It is unlikely that Lienhardt could have read Austin at the time this study was
written. Had he done so the appropriateness of Austin's idea of performative
force for Dinka prayers would surely have struck him.
3 The full title of Kant's book is Religion Within the Limits of Reason Alone*\
4 The psychoanalytic view would insist that the conscience with which we
wrestle in prayer is an internalization of morality based on our parents'
authority. If indeed we are conducting a perpetual dialogue as though still
children with all-powerful parents, this is perhaps closest to the Ashanti
continuing dialogue with ancestors.
5 The several monotheistic religions lay different stress on guilt and sin. And of
course individuals vary greatly in their concern with both. But anxious concern
with the sincerity of feelings and intentions towards a deity would seem to be
associated with monotheism.
 STEPHEN C. LEVINSON

11 Interactional biases in human thinking

The human mind is something of an embarassment to certain disciplines, notably

economics, decision theory and others that have found the model of the rational
consumer to be a powerful one. (Schelling 1988: 353.)

Background
This chapter sets out to weave an improbable web through such topics as
animism, common tendencies in the purchase of soap powder, extra-
terrestrial lifeforms, the phrase 'the whatdoyoucallit', and the theory of
communication. The thread, if it doesn't break, is the theme of a
systematic bias in human thinking, in the direction of interactive thinking
(E. Goody's anticipatory interactive planning or AIP). Because the
argument is somewhat indirect, let me state the thesis right here in the
beginning in semi-syllogistic form:
1. Communication is logically impossible
2. Nevertheless we humans can communicate
3. Therefore, we must use non-logical heuristics and a special form
of reasoning to bridge the gap
4. For communication to work routinely, these heuristics must be
dominant in our thinking all the time
5. Therefore, these heuristics spill over to bias our thinking in non-
communication domains.
As in the famous conclusion to Wittgenstein's Tractatus, where we are
advised not to think that which cannot be thought, so there is a certain
paradox in thinking about biases in human thinking. (You can climb
outside human thought, Wittgenstein hinted, just so long as you throw
the ladder away and climb quickly back in.) We can only do so with real
confidence, perhaps, where we can discern an indubitably correct way of
thinking, guaranteed by the laws of mathematics or logic, from which
human thinking tends to deviate. One such area is human judgement
about uncertain events, and it is here that there has grown, largely

221
222 Stephen C. Levinson

through the efforts of Tversky and Kahneman, a rich literature on biases

in human thinking. I am a complete novice in this field, but I can't help
considering that it might offer rich pickings for those who discern an
underlying human preoccupation with social interaction as the evolution-
ary source of human intelligence. This then is an entirely exploratory
foray out of the theory of communication into a neighbouring field. I am
not optimistic that it will be well received in that neighbouring field, but
interdisciplinary activity has always been a risky enterprise.

I Interactional intelligence, coordination and communication

Interactional intelligence1
In an engaging book (Frames of Mind, 1985), Howard Gardner argued
forcefully for the diversity of kinds of human intelligence, using a range of
evidence from psychological theory, neurology, case studies of cultural,
personal hyperachievements, so-called 'idiot savants', and so on.
Amongst the specialized, compartmentalized intelligences, he listed
linguistic, musical, logico-mathematical, spatial, kinaesthetic and
personal intelligences. Within the latter, he includes what I would choose
to isolate as interactional intelligence, and he lists as evidence for such a
specialized skill the special role of the frontal lobes of the human brain.
Persons with frontal lobe damage of various kinds exhibit different but
related inabilities: 'No longer does the individual express his earlier sense
of purpose, motivation, goals, and desire for contact with others; the
individual's reaction to others has been profoundly altered, and his own
sense of self seems to have been suspended' (1985: 262). Conversely,
patients with massive brain damage to other areas who retain fully
functional frontal lobes - like Luria's 'man with the shattered world' -
retain the capacity to plan actions and to relate to others to a surprising
degree. Similarly, one can point to autistic (and perhaps schizophrenic)
patients, who often show signs of unimpaired reasoning ability, but who
cannot relate to others; and conversely to reasoning-impaired individuals
(like those with Down's syndrome or Alzheimer's) who seem to retain
great interpersonal skills.
Making due allowance for the lay misuse of neurological and patholo-
gical data, there is here the same kind of range of suggestive evidence for a
specific interactional intelligence as there is for other specialized human
skills. One should add to this further evidence from the cross-cultural
study of interaction: although still in its infancy, and still largely unpub-
lished, such work would seem to establish that there are striking univer-
sals in interactional organization, facts compatible with a theory of the
biological basis for interactional skills. Studies with infants strongly
 Interactional biases in human thinking 223

suggest such an innate basis under rapid maturation: newborn infants are
subtly adaptive to the caretaker's presence and handling, and by two
months the child already displays 'a rich repertoire of expressive behav-
iours ... combined with ready orientation of the gaze to or away from the
mother's face and immediate response to her signs of interest and her
talking' (Trevarthen 1979b: 541). That biological basis for interactive
skills is further attested to by a wide range of facts about human
perception, for example our hearing is acute precisely in the range of
wavelengths where speech is broadcast (rather than being specialized like,
say, the owl's auditory system, to the noises of prey).2 Similarly, there is
considerable evidence for a specifically human neurological specialization
for face recognition, implying the fundamental importance of human
face-to-face interaction in human phylogeny.3 And of course all the
physiological, neurological and ontogenetic foundations for language
point in the same direction.
The theory of multiple intelligences should not, though, be equated
with the modular theory of mind a la Fodor (1983); the latter is a
particular theory about how specific specialized skills or 'modules' fit
together with general thought processes to form a computational whole.
The Fodorean requirements for modularity seem altogether too strong to
be correct even for linguistic ability taken as a whole (although they
plausibly hold for specialist subsystems, like segment recognition),
because language understanding necessarily involves general thought
processes. In the same way, interactional intelligence (for reasons that
will become clear) would have to involve central processing and could not
therefore be remotely 'modular' in the Fodorean sense. Nevertheless the
skills that jointly make up interactional intelligence seem to be connected
intimately enough to make up a package of abilities that can suffer
simultaneous neurological impairment.
In this chapter, I shall assume that there is such a form of intellect as an
interactional intelligence, and my central concern is whether we can
detect a systematic bias in human thinking in other domains which might
be attributed to the centrality of interactional intelligence in our intellec-
tual makeup. In order to explore this bias (if such it is), we will need to
have some characterization of the central properties of interactional
intelligence, which I will attempt to provide.4

Anecdotal evidence in favour of an interactional bias in human

thinking
Those who would like to replace Homo ludens. Homo hierarchicus and
other such creatures and caricatures with Homo interactans (not a possible
Latin formation unfortunately,5 but much more plausible) canfindmuch
224 Stephen C. Levinson

anecdotal evidence for such a chap. One of the things that struck Victorian
observers (like Fraser and Levy-Bruhl) of 'primitive' peoples was that
their world is apparently pervaded by mystic forces in para-human form.
Natural causes are mere means subtly utilized by witches, sorcerers,
spirits, gods and demons. It is as if the perceptible natural world were a
stage set, manipulated by supernatural agents always in interaction with
man. Although later ethnographic research (as with Evans-Pritchard's
(1937) classic work on Azande witchcraft) has shown us how systems of
witchcraft and sorcery have an irrefutable internal reason, make sense in a
world imbued with the primacy of social relationships, and so on, it has
not thereby made the central problem of such intellectual genera disap-
pear - namely, why we as a species seem predisposed to such intellectual
systems, even when they are not socially reinforced or are contrary to our
own ideas about real knowledge (as with the astrological systems of early
modern astronomers like Tycho Brahe or even Newton6). That natural
science and magical systems have not only coexisted but often mutually
reinforced one another is now a commonplace of the history of science
(see, e.g., Lloyd 1979). Scientists often operate (like Watson and Crick) as
if nature were a book to be read, a message to be decoded, a syntax to be
parsed, a mode of thought that harmlessly enough might be held to
presuppose a writer, a coder, a puzzle-setter rigging things behind the
perceptible veil.
If scientists are sometimes covert magicians and animists, so of course
are children. Piaget (1929) found that children imbue some inanimate
objects with intentions, feelings and knowledge, and although later work
by Trevarthen and others has shown that very young infants distinguish
interactional partners from other kinds of objects for purposes of com-
munication, yet there seems to be a residual blurring of the distinction in
the belief world of the child (Gelman and Spelke 1981: 56). One is
reminded too of Vygotsky's views about 'inner speech', and indeed the
role that an imaginary interlocutor plays in adult thinking.7
Other areas where animistic and interactional thinking abound are not
hard to find. Consider for example Kahneman and Tversky's finding that
experimental subjects treat random processes as if the processes them-
selves are acting to achieve their own randomness: 'Idioms such as
"errors cancel themselves out" reflect the image of an active self-
correcting process. Some familiar processes in nature obey such laws ...
The laws of chance, in contrast, do not work that way' (Kahneman et at.
1982:24). Economists are often puzzled by the odd purchasing behaviour
of consumers - why do they often just buy the most expensive soap-
powder? My strategy is to buy the cheapest; my wife's to buy the most
expensive. I operate with a vision of some mean, cheating fellow filling
different cartons all with the same rotten stuff; she operates with the
 Interactional biases in human thinking 225

vision of the old-fashioned but always reliable and trustworthy owner of

the corner-store, whose goods are always more expensive but worth it.
The moral is that it's hard to dehumanize even soap-powder. In 1959
astronomical observers started picking up patterned radio signals from
outer space. Someone had the idea that extraterrestrial intelligence was
trying to contact us: suddenly the signals were being scrutinized in a quite
different way, no longer as 'natural signs' of distant physical events, but as
'signals' coded in such a way that any intelligent receiver should be able to
decode them.
Presuming an interactor in the inanimate world is one kind of striking
conceptual 'error' in human thought; another, less obvious, is the
tendency to think of social agglomerations as human actors: we talk
happily of what Russia intends in the Baltic, how it will react to NATO, or
respond to Islamic fundamentalism. Diplomatic protocol is based on the
same principles as interactional politeness (Brown and Levinson 1987);
game theory is applied equally to the moves of military or economic
conglomerates as to the moves of individual players of parlour games;
historians talk in terms of the will of nations. In fact, of course, human
agglomerations are propelled through history largely by forces beyond
their control: the Russian dismantling of the Iron Curtain may have been
no more intentional than an earthquake. Such animistic thinking can have
pernicious consequences: we may detect threats where none exists,
interpret delayed responses as reluctant or hesitant in character, and find
strategic intentions attributed to our non-intentional collective 'actions'.
And so on. There is room enough in the natural history of human belief
systems for much speculation about a bias towards the assumption of a
world constructed out of human interaction with human and super-
human agents. But we seek for some less Victorian level of speculation.

Properties of an interactional intelligence

Human interaction is clearly characterized by an inordinate concern with
the implications that an actor's actions have for other actors' expectations,
emotions, self-esteem, social status - in short it takes place within a highly
structured and often restrictive set of social relationships which permeate
the most intricate details of interactional patterning. Nevertheless I want
to abstract away from that social matrix, to ask about the underlying
conceptual abilities that make social interaction possible.8
The properties exhibited by human interactants are (from an ethologi-
cal perspective at least) extraordinary in a number of ways - exactitude of
timing, complexity of response, layeredness of meaning, and so on.9 It is
obvious enough that interactional capacity relies on a number of core
abilities: the ability to make models of the other, to 'read' the intentions
226 Stephen C. Levinson

behind actions, to make rapid interactional moves in an ongoing sequence

of actions structured at many levels. But what I think perhaps has not
been appreciated is the computational complexity, indeed intractability, of
some of these processes, which is what I want to highlight here.
The computational feat is well illustrated by the ability of humans to
defy the laws of chance: to coordinate mutual actions even when unable to
communicate with one another.
Schelling, in the Strategy of Conflict, reports on some informal experi-
ments that showed that, roughly nine times out of ten, subjects can
coordinate plans without communication (1960: 54fT.). Subjects were
given a joint goal, but then had to independently work out which means
the others would use to solve it, and to choose just that same means. The
kinds of problems solved were (a) to think of the same number, the higher
the number the higher the reward; or (b) go somewhere determinate in a
city to which the other party will also go simply by knowing that each is
trying to select the same location and the same time for a meeting (ibid.).
Subjects coordinated on the number one million, or on the information
booth in Grand Central Station. As Schelling remarked, 'the chances [of
a successful coordinated solution] are ever so much better than the bare
logic of abstract random possibilities would suggest' (p. 57). The joint
goal can require different actions from each party, as when during a
telephone conversation the line is cut, and each party must independently
but jointly decide who will put the receiver down to enable a reconnec-
tion. How coordination is achieved so reliably against such overwhelming
odds remains I think a mystery; but both Schelling and later commenta-
tors have pointed out that it must have something to do with (a) the
provision by the situation of a unique determinate clue, around which
coordination can be achieved; and (b) some powerful property of the
reflexive reasoning that inevitably comes into play: each must do what the
other thinks that the other is likely to do. The two factors together, mutual
salience and mutual computation of mutual salience, seem to be sufficient
to turn a mere lottery into a near certainty.
Schelling (and later commentators like Schiffer 1972) was keen to point
out that the computational problem posed by coordination is really very
different from the formal properties exhibited by agonistic interaction, as
explored in the mathematical theory of games (Luce and Raiffa 1957). In a
zero-sum game (game of pure conflict), you can lay out the action-
reaction sequences 'in extensive form' as a game tree or directed graph
and calculate the relative utilities in advance of play. In contrast, in a
cooperative game of pure coordination, each option of each player yields
zero payoff unless it is matched by the coordinating option of the other
player.10 Both win if and only if each does what the other expects each to
do; otherwise, both lose. In a zero-sum game, one's own preferences are
 Interactional biases in human thinking 227

clear in advance; in a coordination game, it doesn't much matter which

action is taken as long as it matches the other's expectation. Zero-sum
games can be reduced to relatively simple mathematics; but nobody
knows if a mathematics for coordination games could even be formulated.
Thus, calculating optimal behaviour in agonistic interaction is a far
simpler computational problem than calculating coordination: strictly
speaking, Machiavellian intelligence is child's play, a lower-order compu-
tational ability; Humeian intelligence (coordination through implicit con-
tract) is the adult stuff.11
Curiously, though more than thirty years have elapsed since Schell-
ing's work, there has been little empirical exploration of this striking kind
of human ability to coordinate action through apparent 'mind-reading'.
About the same time that Schelling was exploring these problems, Grice
(1957) was devising his theory of (so-called) meaning, which is in fact a
theory of communication which relies on intention-attribution. Although
the theory has been around for thirty years, was subjected to thorough
philosophical scrutiny twenty years ago, and continues to play an
important role in theoretical work on communication, its relevance to
empirical work has not generally been appreciated: it has appeared too
complex, too intentional, too armchair philosophical to form a theoretical
base for practical work. Recently, though, there has been a swing towards
exploring its practical consequences in subjects as diverse as ethology
(see, e.g., suggestions in Dennett 1988) and artificial intelligence (Per-
rault 1987; Cohen et al. 1990), not to mention linguistic pragmatics
(Levinson 1983; Sperber and Wilson 1986) and the psychology of
language (Clark 1992). But above all, it has stood the test of time, and
remains a theory without a systematic rival of any consequence (see
Avramides 1989 for recent commentary). The central idea is that com-
munication is achieved when a recipient recognizes the special kind of
intention with which a communicative act is produced. In one of many
formulations it runs as follows:

Grice's theory of 'meaning':

S means p by x if:
S utters x
(a) intending to get H to think p
(b) intending H to recognize (a)
(c) intending (b) to be the reason for (a).

The point of the theory is that a communicative action is distinguished

by its associated complex intention, which specifies that the 'signal' is a
chunk of behaviour emitted solely (or at least largely) with the intention
228 Stephen C. Levinson

of having its background intention recognized. (In contrast, many of

our behaviours have an instrumental intentional background, where
intention-recognition plays no part, as when I reach out to grasp a glass of
water.)
Grice's theory of communication needs to be placed in the context of a
general 'intentionalism', the view that any kind of interaction involves an
attribution of meaning or intention to the other; it is discerning a chunk of
behaviour as an action, that is, as a bundle of linked intention and
behaviour, that is the prerequisite to response. The response in turn relies
on the other's ability to read the intention or meaning from the behaviour.
Of course there is usually a variety of ancillary information available to aid
and abet this intention-attribution - preeminent sources being perhaps
social roles which act to stereotype intentions (as E. Goody (1978a and
this volume) suggests) and sequential patterns in interaction (see below).
But producing behaviour in such a way that its intentional background is
perspicuous requires a model of the other's ability to so recognize a
behaviour 'x' as an expression of intention 'p'. Thus we enter the peculiar
mirror world of reflexive intentions, now happily occupied by philoso-
phers, computational logicians, theoretical psychologists and others.
What distinguishes a Gricean reflexive intention from other kinds of
complex reflexive intention is that the communicator's goal or intention is
achieved simply by being perceived: recognition exhausts or realizes the
intention.
There have been various attempts to marry Schelling's observations
with Grice's, mainly with the aim of giving a philosophical account of how
linguistic conventions may arise (and thus provocatively raising the
possibility of reducing the concept of meaning entirely to psychological
concepts (see, e.g., D. Lewis 1969; Schiffer 1972; Avramides 1989)). But
there is a more direct and interesting application of Schelling's ideas to
Grice's (Levinson 1985). For the obvious problem raised by Grice's
theory is: how on earth are communicative intentions recognized? The
traditional answer is: by means of a linguistic code (see Ziff 1971 on
Grice). But this turns out to be no explanation even for linguistic
communication, as is explained in the next section, and certainly not for
non-conventional, non-linguistic communication. The fact of the matter
is that we can communicate with 'nonce' signals (Clark 1992: ch. 10). An
alternative answer suggested by Schelling's problems is that we can
choose a behavioural token that is mutually computable as having been
issued with a specific communicative intention, using the same tech-
niques that allow us to coordinate on a unique solution to one of his
coordination problems.
Assuming temporarily that some such picture is correct, let us take
stock of the computational consequences so far. We are already in deep
water:
 Interactional biases in human thinking 229

1. Propositional attitudes
Obviously, computations about other's intentions presuppose compu-
tations over propositions embedded under propositional attitudes. As is
well known, these are 'opaque contexts', contexts where Leibniz's law of
the substitution of referring terms salva veritate, fails: we cannot assume
from the assertion that Esther believes the Chancellor of Cambridge
University should be sacked that she also believes that the Duke of
Edinburgh should be sacked (she may think him competent, believe his
title to be inalienable and certainly not realize he's the Chancellor). This
well-known little conundrum is of course just the logical consequence of
computations over other people's belief worlds (see, e.g., Fauconnier
1985), but it obviously raises difficult computational problems. There are
a number of persistent logical paradoxes, like the Cretan Liar, which also
plausibly have their roots here.12

2. Mutual belief and infinite regress

As mentioned, the Schelling problems seem to require that A and B, in
order to coordinate, come to believe that some salient action is mutually
believed to be the coordination point. But the notion of mutual belief
seems to offer infinite regress: A must believe that B believes that A
believes... ad infinitum... that p. This has attracted much attention, with
philosophers (D. Lewis 1969; Schiffer 1972; Harman 1974; Avramides
1989), psychologists (Clark and Marshall 1981; Clark and Carlson 1982)
and artificial intelligence (AI) workers (see, e.g., Allen 1983: 149)
competing with different accounts each purporting to show how the
regression can be circumvented. Nevertheless the threat of infinite
regress has not endeared the idea of mutual knowledge to those interested
in plausible models of psychological process. But the point of the
Schelling experiments is that they demonstrate that people can indeed
handle just this kind of reflexive reasoning.

3. Gricean reflexive intentions and infinite regress

As was early pointed out by Strawson, Grice's analysis alone might prove
insufficient: one can produce counter-examples which satisfy the con-
ditions but which intuitively are not cases of communication. These are
cases where there is some higher intention of communicator S, not
available to recipient H, and there is thus a discrepancy between the
intention H is meant to discern and that which S actually has. This then
threatens an infinite regress of conditions, with S intending that H should
recognize that S intends that H should recognize . . . and so on.
Various proposals have been made to overcome this potential infinite
regress. One is to ensure that all intentions are out in the open, as it were, if
the behaviour is to count as genuinely communicative - Schiffer propos-
ing for example that there be a condition of mutual knowledge that S has
230 Stephen C. Levinson

uttered x with all the necessary intentions. But the notion of mutual
knowledge must itself be cashed out as an infinite regress of the form 'S
knows that H knows that S knows . . . ' , as we've just seen. Other solutions
involve self-reflexive intentions (Harman 1974; see discussion in Avra-
mides 1989: 58-9), or default inference rules relating communicating p to
believing that p (Perrault 1988: 13).13

4. Mutual salience
We need not only a notion of salience, but a notion of 'natural salience',
such that I can be sure, for an indefinite range of phenomena or scenarios,
that what is salient for me is salient for you. This turns out to be crucial in
ways I shall make clear (see also SchifTer 1972; for a variant suggestion see
Sperber and Wilson 1986).

5. Logic of action
Clearly we need to compute the intentions that lie behind behaviours, if
any kind of coordination is going to be achieved. That would seem to
presuppose an understanding of the derivation of action from intention in
our own planning and acting: one has to choose the means that will most
effectively achieve the desired ends, while balancing incommensurable
goals. As Aristotle argued, the logic of action is a distinct species of non-
monotonic (defeasible) reasoning, a practical reasoning (PR) as it has been
dubbed by philosophers. Von Wright (1971), Ross, Casteneda, Rescher
and others have explored such systems, but there is still much to
recommend the outlines of an Aristotelian system provided by Kenny
(1966), which was developed into a formal system by Atlas and Levinson
(1973) which we may call 'Kenny Logic' (see introduction in Brown and
Levinson 1978: 69-70,92-6). Kenny Logic has many interesting proper-
ties, like the fact that the deductive fallacy of'affirming the consequent' is
valid in this system, or the fact that if'p' deductively implies 'q', then if an
agent desires 'q' he'll desire 'p' (i.e., the logic of practical reasoning looks
like 'backwards' logical implication, a fact that shows up in AI planning
programs like Allen's - see the 'nested planning rule' in Allen (1983:
124)). But the relevant properties here are that Kenny Logic inferences
are both ampliative and defeasible. They are ampliative because one may
reason from a goal to a means that is more specific than is required to
achieve the goal ('I'm thirsty and would rather not be so; here's a Coke; if I
drink this Coke, I won't be thirsty; ergo I'll drink this Coke.'). They are
defeasible because any valid inference from goal Gl to the desirability of
action Al will be abandoned if there is a conflicting goal G2 ('Coke is bad
for my diet.'), from which the desirability of the negation of Al can be
derived. Such a logic of action must also explain how goals can be ranked,
and means of achieving them differentially weighted, in such a way that
 Interactional biases in human thinking 231

the action performed may depend on the 'cost' of the means of achieving
it.14
A logic of action is going to be a complex thing. However, all this turns
out to be the least of the computational problems. Despite all the
philosophical, logical and artificial intelligence work that lies behind all
these ideas, there as been a fatal neglect of one problem. The Schelling-
cum-Grice model of coordination and communication relies on the
recognition of intentions', that is, the need to compute not only from
intention to action (as in a logic of action or planning) but also in reverse as
it were, from behaviour to the intention that lies behind it. It may seem
that if we already have an account, in terms of a logic of practical
reasoning, linking utterances or other actions with the goals that lie
behind them, then all we now need to do is run the reasoning backwards,
from the utterances or actions to the goals. Even logicians who should
know a lot better talk as if intention-recognition is merely a matter of
practical inference 'turned upside down' (as Von Wright (1971:96) puts it
in an uncharacteristic moment of incautiousness).
However, there is an overwhelming problem in equating understand-
ing with 'upside down' practical inference, namely the very great differ-
ence between an actor-based account of actions (in terms of plans, goals
and intentions) and an interpreter-based account (in terms of heuristics of
various kinds). For the nature of logical inference in general, and practical
reasoning in particular, is that there can be no determinate way of inferring
premises from conclusions. Inferences are asymmetrical things. If I con-
clude from 'p and q' that 'p', you cannot, given the conclusion alone,
know whether the premise was 'p and q', 'p and r', 'p and r and s', or 'q and
(q p)', etc.: there are literally an infinite number of premises that would
yield the appropriate conclusion. Simple though the point is, it estab-
lishes a fundamental asymmetry between actor-based accounts and
interpreter-based accounts, between acting and understanding others'
actions. There simply cannot be any computational solution to this
problem, as so far described. The problem is intractable!15 Because the
point is important let me put it in a more concrete way. Suppose I see you
raise your arm outstretched in front of you: your doing this might be
compatible, let us say, given the environmental possibilities, with the
following intentions - waving off a fly, reaching for a glass, greeting an
acquaintance, stretching, etc.16 Even this set of descriptions is to 'cook the
books': instead of 'reaching for a glass' why not go down a stage in
specificity to 'extending an arm' or upwards to 'having a drink'? What we
take to be a natural level of action description is anything but given (as
philosophers from Anscombe to Davidson have been keen to point out).
But then how do we decide what the hell you are doing, and what we
should do in response (raise our hands too, do 'civil inattention', or
232 Stephen C. Levinson

whatever is appropriate), all in the twinkling of an eye (say, 100 milli-

seconds)?17 Going 'backwards' from the behaviour to the intention, at
some appropriate level of specificity, is an absolute inferential miracle.18

Language, communication and interactional intelligence

It is easy to imagine that the main role of language in the evolution of
interactional intelligence is as an independent channel of information
about others' plans and desires, which then makes coordinated interac-
tion possible. That threatens to miss the point - language didn't make
interactional intelligence possible, it is interactional intelligence that made
language possible as a means of communication.19
Non-linguists may require a word of explanation. The model we used
to have, both lay and expert (from Saussure to Shannon and Weaver), of
the way that language works in communication goes something like this:
we have a thought, we encode it in an expression, emit the encoded signal,
the recipient decodes it at the other end, and thus recovers the identical
thought. A moment's reflection will reveal that this picture is absurd.
Consider the 'thing-a-me-jigg' phenomenon:
A: Where the hell's the whatdjacallit?
B: Behind the desk.
Just as in a crossword puzzle, the filled blank, the whatdjacallit,
advertises itself to the recipient as a puzzle the recipient can solve. This
works. In fact it works all the time - we don't say exactly what we mean.
We don't have to and anyway we couldn't. For example, consider the
relation at in
'The car is at the door.'
'The man is at the door.'
'He's at his desk/at University/at work/at lunch/at the telephone.'
There is no unified concept 'at\ except in some highly abstract way: we
figure out the relation by thinking about the objects related and their
stereotypical dispositions. Everything is amplified and specified through
a complex mode of interactional reasoning.
The consequences of this kind of observation are rather far-reaching.
Linguistic competence is not sui generis (at least not this part of it); it is not
'encapsulated' in Fodor's (1983) sense of a specialized, closed-off, module
of mental processing. Semantic representations, or at least interpreted
semantic representations, can't be the 'language of thought' - we think
specifically, we talk generally.20 I can't say what I mean in some absolute
sense: I have to take into account what you will think I mean by it. One
can't encode a proposition; all one can do is sketch the outlines, hoping the
 Interactional biases in human thinking 233

recipient will know how to turn the sketch into something more precise (if
something more precise was intended). The slow realization of all this
(Atlas 1989; Clark 1992; Levinson, in press; Sperber and Wilson 1986)
portends a sea-change in the theory of language: linguistic mechanisms
are deeply interpenetrated by interactive thinking.21
But if we can't say what we mean, how do we understand one another?
When I say 'The coffee is in the cup', I don't have the same kind of IN-
ness in mind as when I say 'The pencil is in the cup'. And when I said 'The
coffee is in the cup', how come you didn't wonder: 'Does he mean the cup
is full of beans, or granulated coffee, or the liquid stuff essential to
academic life?' Nor for one moment, upon hearing 'The pencil is in the
cup', are you likely to think of granulated pencils. Nor are you likely to
worry that the pencil is more than half out of the cup, although on just
those grounds we might expect a quarrel about the truth of 'The arrow is
in the bull's eye'.
It is trying to understand mutual comprehension, given the paucity and
generality of coded linguistic content that now preoccupies theoretical
linguistic pragmatics (cf. Atlas 1989; Horn 1989; Sperber and Wilson
1986; Levinson 1989). We have made some progress in the last twenty
years or so, by identifying heuristics that guide the reasoning process. I
believe that the two cardinal achievements have been to identify two
rather different kinds of heuristics. The one kind is a set of heuristics
based on utterance-type, that is to say that the 'way of putting things'
suggests a specific direction of interpretation. The other kind is provided
by the intricate sequential expectations that are triggered by utterance
and response in conversation.
To take these briefly in turn, the first kind of heuristic, which has been
developed from seminal ideas of Paul Grice, in turn has a number of sub-
types. These play off each other. For example, there seems to be an
utterance-type heuristic that runs: 'normal expression indicates stereo-
typical relation'.22 Consider expressions of the form X is at Y: when we
say 'There's a man at the door', we have in mind a relation of proximity
such that the man can reach the door-bell, say, and is facing it in
expectation. But when we say 'Your taxi is at the door' it may be twenty
feet away and its front not oriented to the front of the door. If your taxi
was to nose its way in, the non-stereotypical event would warrant a non-
normal description; while if the man waited twenty feet in front of the
door, we might prefer another description, say, 'The man is standing
some distance in front of the door'. That seems to be based on another
heuristic: 'abnormal relation warrants abnormal/marked description'.
The two heuristics together explain why 'It's possible to climb that
mountain' and 'It's not impossible to climb that mountain' don't mean
the same thing. A third heuristic runs: 'If an informationally richer
234 Stephen C. Levinson

description applies, use it'. It's this that is responsible for the inference
from 'Some of the Fellows of the College are lazy' to 'Not all of the
Fellows of the College are lazy' - if you meant the stronger statement ('All
of them are lazy') you should have used it.
In what follows I shall rely heavily on the importance of the inference to
the stereotype. It's this that is responsible for such inferences as: 'The
pencil is in the cup', suggesting 'The standard-type pencil (as opposed to,
e.g., a propelling pencil or one with red lead) is projecting out of, but is
supported by, the inside walls of the cup'. As we saw, we come to rather
different conclusions from, 'The coffee is in the cup' (liquid rather than
beans, fully within rather than projecting, etc.), or from, 'The key is in the
lock' (projecting horizontally, not vertically). Some linguists will protest
that these inferences are not pragmatic in nature but rather attributable to
so-called prototype semantics. This I believe to be a rampant conceptual
error, but regardless of that, it really makes little gross difference to the
dimensions of the inferential problem: the particular relation intended by
in for example still has to be inferred by reference to the things related.
The combination of these preferred interpretations of utterance-types
can yield far-reaching enrichments of coded information. From 'Some of
the nurses are not incompetent', one may infer that all the nurses are
female (inference to stereotype), that not all of them are competent
(informative strength), and that the remainder do not fully deserve the
attribution of competence (marked description - the use of double
negation). Or from 'If you wash the dishes, I'll give you 10 Deutsch
marks' one may infer that if you don't, I won't (inference similar to
inference to stereotype), that in any case I won't give you more than 10
DM (informational strength), etc. But I refer the reader to Horn (1989),
Levinson (1983: ch. 3,1987a, b) and Atlas (1989) for details. The point to
grasp here is that without such inferential enrichments, what we say would
tend towards the vacuous: not only do we talk generally, tautologically
and elliptically (as in 'I'll be there in a while', 'If you manage, you
manage', 'Could you please...?'), but also, as illustrated with the example
of the relation at, even when we try to be precise we necessarily trade on
suppositions our interlocutors must make.
The second kind of inferential enrichment that seems to me critical in
language understanding is based upon the fact that, in the conversational
mode that is the prototypical form of all languages use, speakers alternate,
handing over to another party for response at the end of relatively short
turns at speaking. And there's an expectation that responses are generally
tied in close ways to what has gone before. As Sacks and Schegloff pointed
out twenty years ago,23 this makes it likely that if B responds to A in such a
way that it is clear that B misconstrued what A said, there's a good
opportunity provided in the third turn for A to correct, clarify or
 Interactional biases in human thinking 235

elaborate. Thus recipients can be nudged along into what at least passes
for understanding.
Take for example the following:

1. From Terasaki (1976: 45)

M: ... Do you know who's going to that meeting?
K: Who?
M: I don't know!
K: o::h prob'ly Mr Murphy and Dad said prob'ly Mrs Timpte ...
Here M asks a question of K; but K responds with a question ('Who?').
It is clear that K takes M's first utterance to be a prelude to an
announcement, as in the canonical example that follows:

2. From Terasaki (1976: 53)

D: Y'wanna know who I got stoned with a few w(hh)weeks ago?
R: Who.
D: Mary Carter and her boy(hh) frie(hhh)nd.
There are systematic reasons why M's utterance in example 1 might be
heard as the same kind of prelude or 'pre-announcement' as D's first
utterance in example 2. But in any case, K got it wrong: M's utterance was
not an offer to tell, conditional on K's not knowing the facts, but just a
question as made clear in M's second, corrective, turn.
The power of such a system of feedback is well illustrated by the game
of twenty questions: it's generally possible in just twenty question-
answer pairs to guess what the other is secretly thinking of despite the fact
that it might be anything under the sun.24 In addition to such general
corrective potentialities, we should add a large number of very detailed
expectations about how particular sequences may run (like: question
followed by answer, request followed by compliance followed by thanks,
and so on). For a two-turn sequence A-B (like question-answer or offer-
acceptance), each turn usually has rather restrictive specifications on form
and content: thefirstturn because otherwise it will fail to be recognized as
kicking off such a sequence, and the second because the first has been
designed specifically to elicit it.
These sequential clues and constraints help to explain the rather
astounding guesses that can be found in recorded conversations. In
example 2 above, we saw an example of a sequential pattern that runs:25
A: Pre-announcement (request, offer, etc.)
((a turn that pre-figures what will come in third turn, conditional
on B's signal to proceed))
236 Stephen C. Levinson

B: Go-ahead
A: Announcement (request, offer, etc.)
B: Appreciation (acceptance, declination, etc.)
Mutual orientation to such patterns then helps to explain how a
recipient can guess not only that something else is coming up, but that
what will come is of a particular sort:

3. Tape 170
E: Hello I was wondering whether you were intending to go to
Swanson's talk this afternoon
M: Not today I'm afraid I can't really make this one
E: Oh okay
M: YOU wanted someone to record it didn't you heh
E: Yeah heheh
M: Heheh no I'm sorry about that...

4. From Terasaki (1976: 29)

D: I-I-I- had something terrible t' tell you
souh
[
R: HOW terrible is it?
D: Uh, th- as worse it could be
(0.8)
R: W- y'mean Edna?
D: Uh yah
R: Whad she do, die?
D: Mm:mh

5. From Terasaki (1976: 28)

D: Didju hear the terrible news?
R: No. What?
D: Y'know your Grandpa Bill's brother Dan?
R: He died?
D: Yeah
Less dramatically, but more importantly and perennially, these
sequential constraints help to explain how the often near-vacuous nature
of what is actually 'coded' in conversation can carry so much meaning. In
the following extract, for example, co-members of a band are haggling
about how much they ought to practise together, and something as
vacuous as 'Yeah I know but I mean' can serve to suggest that R's excuses
for avoiding the next session really are not good enough:
 Interactional biases in human thinking 237

6. Tape 'Vicar' 144

c: Yeah but I mean we'll be working all night
(1.0)
R: Uh[hh(Isee)
c: er ( )(.) quite late
Well I mean it'- it's up to you I suppose
t
R: yeah
R: But I mean I've got the exam tomorrow so I can't

c: I mean I've
c: Yeah I know but I mean
(1.5)
R: Yeah alright yes =
c: = You understand what I mean
R: Yeah, do you want me to bring my guitar or not =
c: = Yeah
The limiting case is provided by the absence of speech altogether,
which can alone be sufficient to engender detailed inferences, as in the
following example where the speaker takes the absence of response to
signify a clear negative answer:

7. Tape: 'Oscillomink'
c: So: u::m (0.2) I was wondering would you be in your office (0.63)
on Monday (0.42)
by any cha:nce?
(1.86)
probably not
This example illustrates another important feature of conversational
organization, namely that it has very precise temporal characteristics.
Here, C has produced a pre-request in the form of a question, and here, as
generally in English conversation, a pause of over half a second after such
a question may be taken to indicate that the desired response cannot be
easily produced. Due to such temporal characteristics, quite minute
pauses can be most symbolic.
How does all this work? In the case of the utterance-type heuristics, it
only works because speaker and recipient(s) agree that, other things being
equal, there is a normal way to say things. That being so, a normal
description can be taken to implicate that all the normal conditions apply,
in all their empirical specificity: if I say 'John drove off, but he'd forgotten
to loosen the hand brake', you envision a motor car and all the mechanical
238 Stephen C. Levinson

consequences that such a failure of action would entail in such a

mechanism. You know that I know that you will so imagine; you can
therefore take me to be intending that you so imagine; and I can rely on
you so imagining. You would be amazed if it later transpired that John
drove off in a coach and four, or even a tractor! The same goes for the
conversational sequences: you know that I know that you know that I
expect an answer to my question within, say, 500 milliseconds; when you
don't provide it, you know that I know that you have a problem - say, the
desired answer can't be produced. Knowing that, I know from your
silence that the answer is 'no', also of course that you are reluctant to give
it, as you know that I know you should be ...
These examples, informally sketched, will suffice, I hope, to indicate
the peculiar inferential richness that can be extracted by the combination
of reflexive intentional reasoning and a handful of detailed mutual
expectations. Conversational inferences have a number of very special
properties: they are speedy, they are non-monotonic (the same premises
can give different conclusions in different contexts), they are ampliative
(you get more information out than went in) and they are subjectively
determinate. The last point is important: when John says 'I'd like some
water' we don't come away with a feeling that there's a 65 per cent
probability that he had a glass of drinking water in mind, and a 35 per cent
chance he was praying for rain; we come to a definite conclusion, which
may of course turn out to be wrong (but then he'll tell us).26
In all this, conversational inferences are different from logical or
monotonic inferences on the one hand, and inductive ones on the other.
Inferring what is meant in conversation is much more like solving a slot in
a crossword puzzle: such inferences have the rather special property of
having been designed to be solved and the clues have been designed to be
just sufficient to yield such a determinate solution. We might dub this
central feature of language understanding the whatdoyoucallit property of
language, in honour of the magical efficacy of that phrase.
Let me sum up these remarks about language so far. Linguistic
communication is fundamentally parasitic on the kind of reasoning about
others' intentions that Schelling and Grice have drawn attention to: no-
one says what they mean, and indeed they couldn't - the specificity and
detail of ordinary communicated contents lies beyond the capabilities of
the linguistic channel: speech is a much too slow and semantically
undifferentiated medium to fill that role alone. But the study of linguistic
pragmatics reveals that there are detailed ways in which such specific
content can be suggested - by relying on some simple heuristics about the
'normal way of putting things' on the one hand, and the feedback
potential and sequential constraints of conversational exchange on the
other. The astounding speed of conversational inference is something
 Interactional biases in human thinking 239

that should also be noted; these are inferences clearly made well before
responses can be composed, yet responses are, at least a third of the time,
separated by less than 200 milliseconds, and on-line testing shows
pragmatic inferences already well under way immediately after the
relevant word or expression.27

Conclusion to section I
We are now in a position to try and unravel the mystery. Recollect we
concluded that the computational problem posed by Gricean com-
munication or Schelling games looks simply intractable, largely because a
system of inference from intention to behaviour tells us nothing about
how to compute the reverse inference from behaviour to intent. And yet
we routinely manage these things. The pragmatic heuristics may give us
the clue to a solution. The inference to communicative intention from
overt behaviour is so constrained by these heuristics or expectations that
it is possible to select a unique path from within the interminable possible
teleological explanations for the behaviour.
For example, if you know that I know that you know that, for principled
and general reasons, a pause after a question seeking a 'yes'-answer will
suggest a reluctance to provide it, then you know that your pause will lead
me to think that you intended that I think that the answer is 'no'. Both
knowing this, we both know that if you don't do something to correct the
impression, then I'll feel sure that you wanted me so to think. Thus even
the absence of a behaviour may be sufficient to yield the determinate
attribution of an intention.
Or, you say: 'Put some bread and butter on the plates'. What do you
intend? The stereotypical dispositions of course - not, for instance, a well-
buttered plate, or bread on half the plates and butter on the other half.
Even if there are only two plates, you're likely (in England anyway) to end
up with buttered bread, not a plate of bread and a plate of butter. I don't
need to ask you what you intended; I know that you know that we'll both
be oriented to the heuristic authorizing inferences to the stereotype; so
both you and I know that if you want something other than the usual,
you'll have to make warning noises. You haven't; so the probability that
what you wanted was buttered bread is now, for all current purposes, a
dead certainty.28
How might this generalize from linguistic interaction to other forms of
social interaction? In all cases, intention-attribution will be crucial, and
the actual chunk of behaviour will be insufficient evidence alone for the
attribution of an intent. We can carry out mental simulations: I can ask
myself 'What would I be intending in these circumstance were I not me
but him?' But that won't necessarily help me decide whether the
240 Stephen C. Levinson

outstretched arm is a greeting or a reach for the drink or the beginning of a

swipe at afly;there are too many possibilities, too unconstrained a mental
life in the other. What we need, just like the linguistic cases, is some basis
for default presumptions - it really doesn't matter whether these are
wrong or right, arbitrary or well motivated, etc., because once the
expectations are in place you will know that I will know that you will think
that I will use them to attribute an intention to your action, and then you
can go about encouraging or discouraging that presumption (by modulat-
ing the behavioural manifestation).
It is in this context that Esther Goody's suggestion that social roles may
play a special role in interactional coordination is, I think, important (see
also Schelling 1960: 92). The point is made rather well by an artificial
intelligence program designed by Allen (1983), one of the first to draw
attention to the need for intention-attribution in the design of intelligent
responses. The problem was to program an artificial railway-station
information clerk. People might come to 'him' with questions like 'When
does the train to Windsor leave?' or just 'The train to Windsor?' He ought
to answer '11.15 at platform 5': that is, he ought to reconstruct the
intentions behind the often elliptical question, e.g. that the traveller
intends to find the train to catch, and then provide all the information
relevant to that goal. How did Allen solve our intractable intention-
attribution problem? Just by presuming that clients would come to the
clerk presuming that his role was to answer travel questions, and they'd
come with just two of their own goals in mind - catching a train or meeting
one. Thus, by guessing the goals in advance, the program could simulate
the plan-generation that might have led the client to say what he did, see if
he could find an intention chain that culminated in that observable
output, and then assume the client had those nested goals which the
simulation used to arrive at the output (the client's question). In short, a
presumption of the rights and duties of each party to the transaction made
it possible to run the practical reasoning forwards, instead of in the
impossible direction, backwards.
In the same sort of way, social roles may play a crucial role in ascribing
intentions to our co-interactants. Often they won't have the intentions so
ascribed - but that doesn't matter: by setting up the expectational
background, interactants will know equally that they will have to do
something rather special to escape it.
It is tempting, and not altogether implausible, to go the whole hog:
what else is Culture, one might ask, other than a set of heuristics for
intention-attribution? That clearly encompasses language usage, social
roles (as just argued), and a host of heuristics for the interpretation of
mundane and artistic productions. And why else do we feel so at sea in an
 Interactional biases in human thinking 241

alien culture? We may understand the coded content of verbal interaction

and fail to understand the import, observe behaviour but fail to compre-
hend its wellsprings, see mumbo-jumbo where we know there must be
sense, and so on. Any facile definition of 'Culture' with a capital C
deserves, no doubt, a certain modicum of derision, but I can think of
definitions deserving louder hoots.
To sum up: human interaction, and thus communication, depends on
intention-ascription. Achieving this is a computational miracle: inference
must be made way beyond the available data. It is an abductive process of
hypothesis formation, yet it appears subjectively as fast and certain - the
inferences seem determinate, though we are happy to revise them when
forced to do so. The extraordinary thing is that it seems, for all practical
purposes, to work most of the time.
The question is: how? The best answer that we seem to be able to give at
the moment is to take the Schelling games as model: there is an
extraordinary shift in our thinking when we start to act intending that our
actions should be coordinated with - then we have to design our actions so
that they are self-evidently perspicuous. The crucial ingredients are (1)
computations over reflexive intentions and mutual beliefs, and (2) the
ability to settle on identical heuristics, mutually shared, which will yield
default presumptions of intent. Without the heuristics, such coordination
would not be possible: we have to agree in advance, as it were, what the
salient features of the situation are, what any 'reasonable man' would
think such a behaviour betokened, what one would 'normally' mean by
saying 'He's at the door', and so on.
This kind of thinking turns mere probabilities into near certainties.
Example: I try to guess your social security number (chances near to
zero). I try to guess the seven-figure number that you have secretly chosen
in the hope that I can guess it (chances over 0.9).29 We can beat the odds.
Otherwise humans couldn't coordinate in interaction. But it only works
because we think there is a determinate solution, which we only have to
find, like in a crossword puzzle.
That is the peculiar kind of thinking intrinsic to interactional intelli-
gence. If interactional intelligence was the root of human intelligence in
general, which is the idea we are exploring, then we would expect to find
'spill-over' effects in other task domains. For example, when thinking
about (non-human) 'nature', we might expect to find 'nature' treated as a
crossword puzzle, designed (by super-human agency perhaps) to be
decoded and understood. And when humans come to think about chance,
they should fail rather miserably to come to grips with the absence of non-
deterministic solutions, with the fact that apparent patterns are in fact
random assemblages, that a chance exemplar is not custom-made to be an
242 Stephen C. Levinson

exemplar, that a sample could be unrepresentative, etc. It seems that there

is in fact rather a lot of evidence from cognitive psychology in that
direction, which I now review.

11 Biases in human thinking: psychological studies

'Judgement under uncertainty': Tversky and Kahneman

Tversky and Kahneman (1977; Kahneman et at. 1982) conducted a series
of now classic experiments on judgement under uncertainty - in effect
intuitive responses to probabilistic problems. They found that despite the
overwhelming everyday evidence to support basic principles of prob-
ability, people tend to follow other principles that yield incorrect conclu-
sions. For example, everyday reasoning seems to ignore (or only partially
take into account) the following basic statistical principles: (1) the prior
probability of outcomes, (2) the confidence attached to large samples, (3)
the potential independence of properties, (4) the possible chance occur-
rence of an expected outcome, (5) regression towards the mean, and so on.
More concretely, the following examples may give a clearer idea of the
kind of errors systematically repeated:

1. Neglect of prior probability

If subjects are told, e.g., that X is meek, shy and very tidy, they'll guess X
to be a librarian rather than a farmer even when told that there are twenty
times as many farmers as librarians.

2. Neglect of sample size

If subjects are asked what is the better evidence that an urn contains § red
balls and J white balls: (a) a sample of 4 red and 1 white or (b) a sample of
12 red and 8 white, they favour the smaller sample with the stronger
proportion (even though the odds are half as good).

3. Gambler's fallacy
Given a sequence of 'heads-heads-heads-heads' even professional
gamblers often presume a * tails' must now be almost certain (the
'gambler's fallacy' (Tversky and Kahneman 1977: 330)).

4. Neglect of regression towards the mean

Trainers of airplane pilots come repeatedly to the conclusion that
punishing bad landings has a much more powerful effect than rewarding
good ones - failing to take into account that by natural regression the
chances are that a really good landing will be followed by a worse one, and
a terrible one by a better one (Tversky and Kahneman 1977: 332;
 Interactional biases in human thinking 243

Kahneman et al. 1982: 67-8). It seems not to have been noted by

sociologists that this simple failure of statistical reasoning might be
responsible for the vast asymmetry in the size of our penal codes
compared to our system of honours!
Kahneman and Tversky attribute the majority of these 'mistakes' to a
systematic bias to 'representativeness', i.e., 'a representative sample is
one in which the essential characteristics of the parent population are
represented not only globally in the entire sample, but also locally in each
of its parts' (Kahneman et al. 1982: 36). This mistaken assumption of a
representative sample explains the neglect of sample size in everyday
reasoning: a small hospital's obstetrics ward is felt to reflect the sex ratio of
newborns just as accurately as a large one. It also explains the ignoring of
base-rate probabilities: if asked what is the probability of Mr X being a
librarian, when X has all the stereotypical properties of librarians, then
the fact that X is a typical representative candidate overwhelms thoughts
about the rarity of librarians in the population at large.
Representativeness seems also to offer an explanation for the consistent
and rather astounding tendency for people to ignore the most basic law of
probability, Bayesian conjunction, whereby the probability of a joint
event of greater specificity cannot be more than the probability of one of
them alone. Thus the probability of John being both an accountant and a
jazz player cannot be more than the probability of John being a jazz player
- but subjects told that John is a compulsive person with mathematical
skills and no interest in the humanities, feel the conjunction is more
probable (which at least includes the representative profession) than the
single attribute of being a jazz player (which alone seems unrepresentative
(ibid.: 92ff., 496)). Representatives may also explain the gambler's fallacy:
the feeling that having lost three times in a row, one is bound to win next
time - the feeling being based on the expectation that a short run of dice
should exhibit the same randomness of pattern found in a much longer
run.
In short, single facts or small samples overdetermine conclusions
because they are not considered in the larger picture of likely distribution.
Instead, the subject's focus is on typicality, even though typicality and
probability can obviously part company dramatically (e.g., an adult male
weight of 157.851b is highly typical, but less likely than a rough untypical
weight of around 1351b (ibid.: 89)). These mistakes are intriguing
because, not only do they fool the statistically naive, but also (sometimes
only in less transparent examples) dedicated statisticians. It seems
therefore that 'the bias cannot be unlearned', 'since related biases, such as
the gambler's fallacy, survive considerable contradictory evidence' (ibid.:
30).
Kahneman and Tversky detect other, partially related biases. For
244 Stephen C. Levinson

example, there is a tendency to presume causal relations, and to find it

easier to infer effects from causes than causes from effects {ibid.: 118).
There is also a strong bias to what the authors call availability (Tversky
and Kahneman 1977:333ff.), i.e., to salience or ease of recall, so that if it is
easy to think of instances, then the event type may be thought to be
frequent. For example, people overestimate the number of words begin-
ning with R over the number with R in third place, because it is relatively
much easier to retrieve words beginning with a letter than having such a
letter in third position (Kahneman et al. 1982: 166). And quickly made
associations are often presumed to be accurate correlations, despite
evidence to the contrary: if recurrent correlation is one source of mental
association, it is nevertheless of course illegitimate to assume that all
associations are based on correlations (Tversky and Kahneman 1977:
335). Availability is thus a matter oi focus, people overestimate the
importance of what is in focus and underestimate what is out of focus:
preoccupied with winning the lottery or with the thought of an aircrash,
they overestimate the probabilities of both. One aspect of this directly
relevant to agonistic interaction is the tendency to underestimate the
opponent (Kahneman et al. 1982: 177).
Kahneman and Tversky conclude: 'In his evaluation of evidence, man
is apparently not a conservative Bayesian: he is not Bayesian at all' (ibid.:
46). If this is correct, one must wonder why; after all, we live in a world
dominated by chance events. It is self-evident that in the period in which
our genetic makeup was laid down the dependence on chance, uncertain
events, from the success of the hunt or harvest to the health of the chief or
leader, must have been much greater than in the cybernetically controlled
western world of today. How could we afford mental biases in such non-
adaptive directions?
Kahneman and Tversky themselves offer no speculations, but they do
offer us some tantalizing clues as to how their observations might tie into
the biases or tendencies that are prerequisites for human communication
- the overdeterministic mode of thought typical of, and necessary for,
interactional coordination. They too notice the 'illusion of validity', 'the
unwarranted confidence' which 'persists even when the judge is aware of
the factors that limit the accuracy of his predictions' (Tversky and
Kahneman 1977: 331). One such connecting clue is the obvious relation
between their notion of representativeness and the notion of prototypi-
cality (Kahneman et al. 1982: 86-9), as that latter notion has been
explored in linguistic categorization. A representative individual would
indeed be a prototypical one, and thus a special case of a representative
sample, which should represent in microcosm the population as a whole,
and thus also mirror its variability.
 Interactional biases in human thinking 245

If the reliance on representativeness leads to systematic errors, why do people use

this relation as a basis for prediction and judgement? ... Modern research on
categorization (Mervis and Rosch 1981; Rosch 1978) suggests that conceptual
knowledge is often organized and processed in terms of prototypes or representa-
tive examples. Consequently wefindit easier to evaluate the representativeness of
an instance to a class than to assess its conditional probability. (Kahneman et al.
1982: 89.)
Our earlier suggestion, recollect, was that prototypicality or stereotypi-
cality plays an essential role in communication and action coordination by
providing a salient coordination point to which parties to interaction can
each be sure the other will attend. Hence I can assume that you will
understand what 'The pencil is IN the cup' would stereotypically suggest
- if it lies in the bottom in broken pieces, I'd better tell you so. Likewise, if
you move aside at the door, deference may be presumed as the motive
rather than, say, fear. Stereotypicality provides the heuristic for a solution
to the intention-attribution problem: each can assume that the other will
use this heuristic and will therefore act accordingly, thus giving a
deterministic solution to an otherwise impossible problem.
Other clues for connecting Kahneman and Tversky's observations to
interactional intelligence may be found in their remarks about intuitive
patterns of randomness, salience ('availability'), causality, sequentiality
and so on. First, people find it unintuitive that short highly patterned
sequences could be random. For example (where H = heads, T = tails),
that HHH could be random, let alone H T H T H T or H H H T T T is
counter-intuitive (Kahneman et al. 1982: 37). For even randomness is
expected to be 'representative', i.e., exhibited over short stretches as an
unpatterned sequence, for which one could write, for example, no little
generative grammar. The corollary is: we see design in randomness, think
we can detect signals from outer space in stellar X-rays, suspect some
doodles on archaeological artefacts to constitute an undiscovered code,
detect hidden structures in Amazonian myths. If we are attuned to think
that way, then that is perhaps further evidence for the biases of interac-
tional intelligence: in the interactional arena, we must take all behaviour
to be specifically designed to reveal its intentional source. Second, people
seem to favour higher probabilities where a causal or teleological connec-
tion can be posited, which is as might be expected from an interest in the
wellsprings of action (ibid.: ch. 8); and they attribute quasi-intentionality
to random processes when they act as if such processes were self-
correcting (ibid.: 24). Further, they believe that they can somehow
exercise control over chance, as we can over our fellows, as when they
prefer a lottery ticket they have selected over one given out (ibid.: 236).
Third, salience or mental 'availability', which Kahneman and Tversky
246 Stephen C. Levinson

construe as a separate bias, plays a special role in the solution to Schelling

games and communicational coordination: when we manage to meet
again in a foreign city after getting accidentally separated, we do so by
each thinking where the other will go, mutually deciding that a particular
location (e.g., the cafe we were last together in) will be to each of us the
most salient meeting place. Finally, when we find significance in the
pattern of coin tosses THTH or think that after T T T we must have an H,
we exhibit a sheer preoccupation with sequential pattern, where sequen-
tial patterning was one of the essential heuristics that we listed as making
communication possible.
Consider again our treatment of communication as a 'crossword
puzzle': there are multiple constraints on the 'slot' in which a communica-
tive action is fitted, and the communicative act itself is only a clue to its
proper interpretation. But it is a determinative clue: once you have it, you
have it - you don't generally have it to 65 per cent certainty or the like; for
it's taken to have been designed to yield a single, determinate interpre-
tation. It's also taken to have sequential implications of a determinative
sort - other communicative acts should now be opened up. The kind of
thinking required in communication is a mental search for a salient - for
example, stereotypical - interpretation, the psychological prominence of
which is the best guarantee that this interpretation is indeed the mutually
intended one. All the biases that Kahneman and Tversky list:
1. determinism, overconfidence, representativeness;
2. prototypicality;
3. sequentiality;
4. the ascription of teleology (e.g., in the belief in self-correcting
random sequences),
would seem to be relatable to a communicational mode of thought - on
this hypothesis they would be the side-effects of an interactional
intelligence.
Of course one radical possibility is that Kahneman and Tversky's
results are entirely a byproduct of the communicational context in which
the experiments were carried out. Instead of focusing on the tasks as real-
world problems, perhaps the subjects see the experimental tasks as
communicative crossword puzzles: the experimenter has given clues as to
what he wants - the subject must guess the desired outcome which has
been designed to be guessed, like a problem in the classroom. Recently
Kahneman and Tversky (postscript to Kahneman et al. 1982: 502)
themselves have come to see that Gricean implicatures may play a role in
their results through biasing the experimental description. However,
they continue to underestimate that possibility severely. For example,
they fail to note (ibid.: 497-8) that the famous Wason four card problem30
 Interactional biases in human thinking 247

is entirely explained by what linguists call 'conditional perfection', the

Gricean conversational implicature from 'If to 'If and only if.31

8. Wason four card problem

Instruction: Given the rule 'If a card has a vowel on one side, it has an even
number on the other', which of the following cards should you turn over
to test whether it is correct: A, B, 4, 7?
Correct Answer: A and 7. (Since the rules states 'If it's an A, it's
even', or symbolically: 'A = even' which implies by modus tollens 'not-
even = not-A'.)
Predominant answers: A and 4.
Gricean explanation: 'A = even' implicates 'A = even', thus 'even = A' (i.e.
saying 'If it's an A it's even' implicates 'If and only if it's an A, it's even',
from which it follows 'If it's even it's an A'.)
Similarly, there may be implicatural reasons for the failure to operate
the Bayesian conjunction rule (that the probability of A and B cannot
exceed either the probability of A or the probability of B). Consider the
following task:

9. Written background detail (Kahneman et al. 1982: 496)

Linda is 31, single, outspoken and very bright. She majored in Philos-
ophy. As a student, she was deeply concerned with issues of discrim-
ination and social justice, and also participated in anti-nuclear
demonstrations.
Instruction: Which is more likely:
A: Linda is a bank teller;
B: Linda is a bank teller who is active in the feminist movement?
Correct answer. A
Predominant answers: B
The Gricean explanation here would rely on the presumption of a
cooperative experimenter, who has produced (just as is always expected in
conversation) only the relevant facts in the background description. But if
judgement A is correct, then the facts must be irrelevant; since ex
hypothesi the facts are relevant, A cannot be correct, so the only given
alternative (B) must be right.
However, in addition to neglecting the possibility of a Gricean explana-
tion of their alleged biases in thinking, Kahneman and Tversky have also
failed to note the formal similarity between the basis of Gricean implica-
tures and some of their biases - those that we have noted under the rubrics
of salience (availability), prototypicality, representativeness, etc. There
248 Stephen C. Levinson

are thus two possibilities: the entire Kahneman and Tversky research
programme is vitiated by the failure to consider the biases introduced by
Gricean heuristics due to the verbal nature of the task-setting, or there is a
real non-communicative bias in thinking, but one which mimics Gricean
patterns because communicative heuristics inhabit, as it were, the deeper
reaches of our minds. Given the breadth of experimental data, and the fact
that at least some of the findings do not allow Gricean explanations
directly, the second possibility seems the more likely interpretation of the
facts.

Dorner's planning and decision-making with complex systems

Kahneman and Tversky's tasks are presented verbally, thus opening
them up to a critique in terms of communicative bias rather than cognitive
bias. They also perhaps suffer from a parlour-game quality that may
introduce Schelling's game coordination reasoning irrelevant to the
design of the tasks. But there are other lines of psychological research that
tend in the same direction. We turn now to just one further example.
Dorner (1990) has been exploring how subjects try to cope with
complex, dynamic systems, often with hidden interconnected variables,
delayed responses, patterns not available on short-term inspection, etc.
Such systems, he argues, form an important part of the human decision-
making environment, with good examples being agricultural and ecologi-
cal systems, politico-economic systems, industrial plants, and so on.
The kind of problems posed by Dorner's simulations seem typical both
of complex natural systems (like ecologies, predator-prey relationships)
and complex artificial systems (like economies, polities, armies, industrial
plant and other cybernetic machines). Yet our failures to understand
quite simple indirect causations can have quite dramatic consequences (as
with the Chernobyl accident, where increasing the flow of cooling water
indirectly caused the removal of graphite moderators (Reason 1987)).
Just as Kahneman and Tversky's results show a failure to understand the
most elementary aspects of our largely random world, it is again rather
striking that humans seem ill-adapted to coping with such complex
systems, the manipulation of which might have been expected to have co-
evolved with human intelligence. Surely, one might argue, a hunting-
gathering prehistory might have led us to have intellectual mastery over
predator-prey and other ecological systems, whereas in fact, of course,
our understanding here has proved lamentable. Indeed, we carry one
prototypical complex system permanently around with ourselves, namely
our bodies, and yet it is notable how little we naively understand the
complex system in which we are thus imprisoned - for example, how is it
possible that the relation between sepsis and lack of cleanliness had to
await discovery in nineteenth-century Europe?32
 Interactional biases in human thinking 249

In Dorner's paradigm of research, unlike the Kahneman and Tversky

one, there is often no mathematical or other precise way to measure
optimal behaviour - in even a simple ecological system the variables are
too many and their interaction too complex. Instead, one can measure the
relative success or failure of individuals against the performance of their
peers, in computer simulations of the relevant domains.
What Dorner has found is that when performance is bad, this is
attributable to a number of recurrent 'errors', for example:
1. Failure to cope with delayed responses or long-term time series:
e.g., failure to regulate a thermostat with a time-delayed, damped
oscillation pattern (1990: 20). Subjects had a great tendency to
react to the immediate state of the system, without taking into
account underlying trends.
2. There was a tendency to interdigitate action and analysis, instead
of doing prolonged analysis first.
3. There was a tendency to not look back and check the current
consequences of past actions or guide them further.
4. While subjects often focus too hard on a salient problem, thus
neglecting parallel problems and side-effects of the focal
problem, they nevertheless jump from problem to problem,
often without seeing the problem through to the end.
5. As subjects fail to control a complex system, they often take
increasingly ill-adaptive measures - they seek for continued
confirmation of failing hypotheses, become entrenched in their
thinking and less observant of changes in the system.
Dorner attributes many of these 'failures' to a relatively small number
of mental tendencies (1990). First, there are apparently irrational33 or
emotive factors, which he believes can be attributed largely to the desire
to 'guard one's feeling of competence', especially by avoiding retrospec-
tive examinations of failures. Hence one finds, especially as the loss of
control over a system builds up, a failure to check consequences of past
actions, a 'dogmatic entrenchment' of failed hypotheses, a fleeting
attention paid to one problem after another - a kind of mental 'panic'
accompanied by behavioural rigidity, all a kind of escape behaviour, as
subjects avoid facing up to the facts of loss of control of the system.
A second major source of failure he ascribes to sheer overburdening of
mental processing. Hence subjects tend to seek single central explana-
tions, or to judge a single variable to be the one determinative factor
('central reduction' (1987: 21)); and they fail to pursue consolidated prior
information gathering.
A third factor may be the sheer inability to perceive patterns distri-
buted over time because of forgetting; hence the failure to perceive time
series, exponential growth patterns, etc.34 Afinalfactor Dorner isolates is
250 Stephen C. Levinson

overemphasis on the current problem, with consequent neglect of side-

effects and long-term effects, which he attributes to an attentional
mechanism.
Dorner's explanations of the human failure to come to terms with
complex systems can be pushed one stage further by asking why the
'errors' of thinking that he detects should be there in the first place. By
innuendo, like Kahneman and Tversky, these mental failings are held up
as natural deficiencies, as it were, for which no explanation is required.
Thus 'human beings are "creatures of space" not "creatures of time5"
and hence they fail to 'see' time-series or exponential growth patterns
(1987: 22). The overall explanation is given in terms of a mixture of
processing deficiencies and irrational self-deception.
Instead, accepting Dorner's analysis for a moment, one might try to
find an explanation of the failings in our hypothesized interactional
intelligence. One of his factors, the protection of one's feelings of
competence, is easy enough to relate to interactional concerns. Self-
esteem is not generally won or lost by encounters with 'nature', but rather
by encounters with the fellow members of our species. We care about
admitting mistakes to ourselves because we care about admitting them to
others.35 Thus if the 'irrational' factors involved in poor performance in
struggles with complex systems can be rightly attributed to this preserva-
tion of the self from its failures, then these would seem to be deeply tied
into factors at the heart of social interaction (although not the ones we
focused on in section I).
The other key factors, the processing deficiencies, have certain striking
similarities with the Kahneman and Tversky findings. Kahneman and
Tversky's 'representativeness' carries with it a tendency to ignore the
larger distributional pattern, and to focus on local patterns as if they were
truly representative of the larger picture. Dorner's findings about re-
stricted information-collection, restricted planning and the entrench-
ment of old hypotheses closely echo the Kahneman and Tversky findings,
including that same insistent confidence in transparently erroneous
inferences. For example, Dorner has found subjects to become totally
preoccupied with the most salient problem, just as Kahneman and
Tversky found naive thinking about uncertainty to be partially deter-
mined by 'availability' - i.e., salience, recoverability and focus. The same
remarks about a possible source for such a focus in interactional intelli-
gence carry over to Dorner's work: intention-recovery relies on coordina-
tion on a solution to the interactional 'crossword puzzle' - there must be a
mutual focus on a single interpretation, and so the interactant must be
forever seeking the single, determinative key to the intention-recovery
problem. To this we can therefore assimilate also the tendency to seek just
one single all-explaining factor, the critical variable on which the whole
 Interactional biases in human thinking 251

complex system is thought to hang. Interaction requires single-solution

thinking; complex systems require multiple-solution thinking - but
humans are only good at the former.
Recollect that we suggested that there are two critical kinds of heuris-
tics that make intention-recovery possible. One is the kind that yields a
default interpretation for an action - the kind exemplified in verbal
interaction by heuristics that legitimize the assumption .of stereotypical
attributes, and so on. The other is the kind based on sequential infor-
mation. This latter kind shows up in the Kahneman and Tver sky material
as a presumption of the highly structured nature of any short sequence -
thus accounting for the gambler's error, and the converse effect of the
refusal to consider a patterned sequence as possibly random. Interaction
sequences display certain clear properties:
1. The last action in an action chain is usually the focus of response
(but not always, e.g. after an embedded 'insertion sequence').
2. Human interaction sequences which form canonical linear struc-
tures are usually rather short - for example a pre-sequence of four
turns is relatively long (as in A: 'Doing anything tonight?', B:
'No, why?', A: 'How about having a meal together?', B: 'Great
idea'.).
3. Interaction sequences are generally single-stranded - that is, one
doesn't run two or more equally prominent chains of interaction
simultaneously (there are of course exceptions, as with stock-
brokers simultaneously dealing on phone and floor, etc.).
4. Interaction chains are characterized by rapid turn-taking, with
short (average 400 millisecond) intervals between them.36
In Dorner's material one can find perhaps echoes of each of these
expectations. The echo of property (1) is the tendency to focus on just the
matter in hand, while being prepared to rapidly switch to other concerns.
Echoes of property (2) would be the failure to discern much longer time
series - the action-reaction style of human interaction ill prepares us for a
complex system that reacts suddenly and catastrophically after a long
time delay (as when an environment suddenly degrades, or when our
bodies react to an earlier ingestion or infection). It also ill prepares us for
reactions with relatively small time delays, but just sufficient to be beyond
the normal action-reaction time span, as the subjects found in their
unsuccessful attempts to control a simple time-delayed thermostat.37
Echoes of property (3) (single-stranded interaction chains) are Dorner's
observations about the single-strandedness of thinking, the preoccu-
pation with only one causal chain at a time. Echoes of property (4) can
perhaps be found not only in the temporal properties of the human
attention span, but also in the tendency, for example, to collect minimal
252 Stephen C. Levinson

information, then act, collect, act and so on, hoping to learn from
interdigitated action-reaction rather than exhaustive prior analysis.
Given the possibilities of interactional feedback, that is the right way to
operate in the interactional domain.
Dorner's work, unlike Kahneman and Tversky's, is concerned with
human abilities to cope with dynamic systems as they react over time - it
thus serves to bring out some of the possible biases in temporal thinking
that are not illuminated by the Kahneman and Tver sky paradigm. It
provides therefore a different kind of ammunition for the protagonists of
the primacy of interactional intelligence. Those protagonists would be in
trouble if Dorner's findings had indicated that people are good, say, at
dealing with time-delayed reactions; orfindlong-term oscillations easy to
discern; or can easily cope with multiple strands of sequential events. But
Dorner's findings are all comfortably in the right direction - towards the
conclusion that humans are good at dealing with single-stranded teleo-
logical or causal chains, with immediate action-reaction expectations
which require immediate attention or allow only a small 'push-down
stack', four or five 'plates' deep.
Kahneman and Tversky's results are vulnerable to the charge that all
the observed biases are introduced by the linguistic communication that
sets the task for subjects. Dorner's non-verbal tasks where the subject
wrestles with a computer simulation escape that charge. However, they
are arguably vulnerable to a parallel charge in just the area of most
interest, the temporal characteristics of human behaviour: perhaps in
setting up the computer simulations we have unwittingly introduced
properties of human-human interaction into the design of human-
computer interaction. For example, a keyboard or other input device will
typically control only one variable at a time, nor are commands normally
set to act at remote time intervals or atfixeddelays to govern future states.
To introduce such a system of relations between human and machine
would be 'un-naturaP - the very structure of the machines and programs
we make for humans to interact with reflects (often to the rather lowly
limitations of the engineer's imagination) the temporal properties that
we, as humans, find it comfortable to work with. Thus the very set of
biases we seek to illuminate in an objective way by setting subjects tasks
that have intrinsically good solutions, may in fact have been built into the
structure of the tasks themselves.

Conclusions
I have argued that intersubjectivity requires peculiar computational
properties, which may then bias many aspects of human thinking. On the
one hand, onefindsthe presumption of deterministic solutions, what one
 Interactional biases in human thinking 253

may call the 'crossword puzzle effect' (problems are treated as if they were
designed to be solved): hence the presumption that patterns can't be
random, exemplars are prototypical, samples are 'representative' and
conclusions can be certain. On the other hand, one finds some evidence
that attention and memory are geared to interaction tempo: humans
presume single-stranded causal chains, respond (usually) to the immedi-
ately previous event, expect brief action-response intervals and very short
sequential patterns.
The first group of biases can be plausibly related to the necessity of
having mutual orientation to the kind of heuristics we discussed as
essential to language-understanding, e.g., the kind that gives us the
strong readings of a preposition like at, according to the relata. The
second group of biases, of attention and memory, may be related to the
sequential heuristics for the attribution of intent in interaction - when
talking, we are mutually oriented to the potential for immediate correc-
tion, and to canonical sequences of certain kinds.
Without such an explanation, the kind of biases noted by Kahneman
and Tversky on the one hand, and Dorner on the other, would be puzzling
indeed from an evolutionary perspective. The ability to make objective
estimates of probability would offer immediate adaptive advantage, e.g.,
to a hunter faced with a decision to go after one kind of game or another.38
Likewise the ability to comprehend complex systems, whether natural
(like our own bodies, or the ecologies we live in) or socio-political, ought
to offer significant adaptive advantages. It seems reasonable to suppose
that, instead, there must be some greater adaptive advantage to thinking
in the ways we actually do, and my suggestion is that these biases are
essential ingredients for intersubjective reasoning. The corollary would
be that the main evolutionary pressures on our species have been intra-
specific. That accords with at least the views collected in Byrne and
Whiten (1988), who have urged us to substitute a 'Lord of the Flies'
scenario for a 'Robinson Crusoe' scenario for human adaptation. To that
view, the speculations in this chapter add, hopefully, a corrective: it is
cooperative, mutual inter subjectivity that is the computationally complex
task that we seem especially adapted to. Machiavellian intelligence merely
exploits this underlying Humeian intelligence that makes intersubject-
ivity possible. One needs too to stress that it is this cooperative intersub-
jective background that makes language interpretation possible (as shown
by the need for all those heuristics) - not, as non-semanticists may
assume, language which makes inter subjectivity possible (although it
obviously vastly increases its scope).
In this chapter, I have stressed two pervading characteristics of human
thought - attribution of intentionality and overdeterminism - which may
be directly related to interactional intelligence. For without that over-
254 Stephen C. Levinson

determinism, we would never have the heuristics that make it possible to

ascribe intentions to human behaviour. As Peirce and many more recent
writers have been keen to emphasize, deduction and induction are
relatively trivial human skills, of no great computational complexity; it is
abduction or theory construction which is the outstanding characteristic
of human intelligence.39 Abduction is the leap of faith from data to the
theory that explains it, just like the leap of imagination from observed
behaviour to others' intentions. While most explicit human theories or
abductions are wrong, our implicit ones about interactional others are
mostly good enough for current purposes. Both, though, come with that
striking element of overconfldence, overdeterminism (even when we
know, as in the case of scientific theories, that the half-life of the theory is
only a year or two). Which allows me to end on a paradox: were we to feel
any confidence that the roots of abductive ability (and it's peculiar
phenomenology of certainty) lay indeed in interactional intelligence, and
thus any confidence at all in the thesis of this chapter, then we could
ascribe that feeling entirely to the overdeterminism of interactional
intelligence itself.

Acknowledgements
This chapter owes much to Esther Goody, who provided the stimulus to
crystallize these thoughts. My second important debt is to London Transport,
since some of the ideas here transcribed arose out of a long conversation with
Dietrich Dorner while perforce walking the streets of London to and from
meetings of the Royal Society during the transport strike, June 1989 (see Dorner
1990). Other ideas in this paper were first tried out at a meeting of the British
Psychological Association (Levinson 1985), and I thank various commentators
there, especially David Good, L. Jonathan Cohen and Phil Johnson-Laird. I have
had helpful comments on this chapter from Penny Brown, Dietrich Dorner,
Esther Goody, and Alex Wearing, for which I am most grateful. Much further
back, Esther Goody (1978a) first pointed out to me (and us anthropologists
generally I suspect) the relevance of the study of social interaction for theories
about the evolution of human intelligence.

Notes
1 I see the notion of interactional intelligence contrasting in specifity with other
related notions. Social intelligence (or social cognition), as used for example in
Flavell and Ross (1981), is an altogether broader conception, including the
apprehension of morality, dominance, friendship and appropriate social role
and affect. The Machiavellian intelligence of Byrne and Whiten (1988) is also
wider, encompassing social knowledge, problem-solving in a world of flexible
andficklesocial relations, and so on (pp. 50ff.). By interactional intelligence, I
have in mind just and only the core ability to attribute intention to other
 Interactional biases in human thinking 255

agent's actions, communicative or otherwise, and to respond appropriately in

interdigitated sequences of actions; and I want to emphasize particularly the
computational intractability of intention-attribution. I take this ability to be
the bedrock feature of all the other, wider concepts, as recognized clearly in
Esther Goody's term anticipatory interactive planning (AIP), which differs
from my notion, I think, only in breadth and emphasis. All of these modified
uses of the term intelligence, Alex Wearing points out to me, refer to a faculty or
ability, and not the inherently comparative notion that the unadorned noun
refers to.
2 For the prey orientation of the owl auditory system, see Schone (1984: 212).
For the matching of speech signal and auditory system in humans see, e.g.,
Lafon (1968: 81, fig. 2). However, in the human case there is more than mere
matching of frequency and amplitude between speech signal and auditory
discrimination; there is also a special kind of neural processing that clocks in
when speech sounds are heard (Lieberman and Blumstein 1988:148ff.). There
are also fairly clear patterns of matching between properties of speech and
properties of short-term memory (see note 37). Unfortunately, there is no one
locus where all these sorts of facts are laid out for non-specialists, although
they are essential background to speculations about the evolution of language.
3 See, e.g., Kolb and Whishaw 1990: 237-41; for a wider-ranging popular
account see Landau 1989.
4 If there is such an assemblage of abilities that we can call interactional
intelligence, why has it been so neglected in the wide range of disciplines (from
anthropology to neurology) that might have studied it? Presumably, partly
because of the tendency to take for granted what humans are naturally good at.
We do not cherish bipedalism in the same way that we celebrate our ability to
do calculus. The corollary is that we value that which we are not very good at:
dancing au point, calculating decontextualized syllogisms, democracy, etc. But
there may be another reason for the neglect of the study of interaction,
namely inhibition or repression. It is not only that (as every transcriber of a
conversation knows) friendly interaction is, on minute inspection, replete with
nasty little jabs. It's also that certain human skills only run fluidly out of
conscious awareness. Just as it is awfully hard to drive when taking a driving
test, walk in a straight line when arraigned on suspicion of drunken driving, or
appreciate a symphony when trying too hard to appreciate it, so self-conscious
interactants generally do themselves a disservice (see, e.g., Field 1955 [1934]).
If so, the repressive mechanism that aids our daily interaction may be
responsible for making us equally reluctant to look at it scientifically. (On the
role of inhibition in controlling, e.g., our perceptual world of smells, see O.
Sacks (1985: 151).)
5 Dietrich Dorner suggests Homo interagens.
6 J.Z. Young (1951: 3) reporting Lord Keynes's comments on looking through
Newton's alchemical papers: 'Newton was not so much one of the first men of
the age of reason as the last of the magicians. He seems to have thought of the
universe as a riddle posed by God, which could be solved if one looked hard
enough for the clues. Some of the clues were to be found in nature, others had
been revealed in sacred and occult writings.'
256 Stephen C. Levinson

7 On speech to the self, see Goffman (1978) and Levinson (1988).

8 However, I return briefly to one social aspect in the review of Dorner's work
below.
9 See, e.g., Goffman 1981; Levinson 1983: ch. 6; Clark 1992.
10 Humphrey's (1976: 19) seminal paper on the social function of the intellect
uses the zero-sum game as a model of the computational demands of social life
which 'asks for a level of intelligence . . . unparalleled in any other sphere of
living'. My point is that zero-sum games merely require decision trees for
different contingencies; coordination games require deep reflexive thinking
about other minds, and constitute a much more demanding intellectual task.
In Schelling's (1960: 96) words: 'In the pure coordination game, the player's
objective is to make contact with the other player through some imaginative
process of introspection, of searching for shared clues; in the minimax strategy
of a zero-sum game... one's whole objective is to avoid any meeting of minds.'
11 Hence a superficial objection to the terminology of Byrne and Whiten (1988).
Actually of course what they intend is a Machiavellian intelligence superim-
posed on a Humeian one, i.e., the potential for an agonistic exploitation of a
supposedly cooperative understanding (cf. their quote (p. vi) from Machia-
velli: 'For a prince, then, it is not necessary to have all the [virtuous] qualities,
but it is very necessary to appear to have them.'). Nevertheless, one can't help
feeling that their ethology is pervaded by the very agonistic bias (vicious
struggle for survival of the fittest) that underlies the very Robinson Crusoe
model (man's mind against 'nature') which they are complaining about. We all
know cooperation is harder than conflict; it is not so obvious that one reason is
that it's computationally harder too. (By the way, the reference to Hume is to
A Treatise of Human Nature (III, ii.2.) where reflexive reasoning about the
benefits of mutual cooperation is supposed to underlie our tacit acceptance of
conventions (see Schiffer 1972: 137ff.).
12 'One of themselves, even a prophet of their own, said, the Cretans are always
liars' (St Paul's epistle to Titus 1,12). If the Cretan prophet speaks truly, then
what he says is false; if he speaks falsely, then what he says would truly
characterize him, but must nevertheless be false. The quotational aspects of
the paradox are usually abstracted away from in philosophical discussion.
13 Do humans really go through all this reasoning about what each thinks the
other thinks, and if so to what depth? The answers seem to be 'yes' and
'indefinitely deep' respectively, as is most clearly revealed where asymmetrical
beliefs at a deep level are the name of the game, as in military strategy,
paranoia, fraud and the like. Consider the beginnings of recorded western
military strategy: e.g., Hannibal beat Scipio at the battle of Trebbia by making
his centre only look like the normal thick phalanx, drawing the troops onto the
wings, so the centre would collapse and the wings wrap round. Next time
round the Romans might expect the same strategy, so this time Hannibal
might stack the centre for a central concentrated punch: Hannibal's thinking
that Scipio's thinking that Hannibal is thinking that Scipio will suspect a weak
centre; Hannibal's hoping that Scipio will think all that but not also that
Hannibal thinks Scipio will therefore weaken his centre to reinforce the wings
making the centre an obvious target. However, suppose Scipio (or his
 Interactional biases in human thinking 257

successors)figureall that out - then they will thicken up the centre. Best then
to repeat the Trebbia formation, but to bow the centre out so that it really
looks packed, but in fact is a hollow crescent, designed to crumble. So
Hannibal thought and won the battle of Cannae by another pincer movement
from theflanks(Connolly 1978). If early classical military strategy went that
deep, how deep was the reflexive thinking that, e.g., Kennedy and Krushchev
got into over the Berlin wall/Cuban-missile crisis of 1961-2 (Gelb 1986)? For
depth in cooperative reflexive reasoning, consider, e.g., irony and double
irony (see Penelope Brown's contribution to this volume (Chapter 7)).
14 See Cohen et al. (1990) for some recent ideas here.
15 This has not of course prevented computational attempts to circumvent the
problem (see, e.g., Allen 1983; Perrault, 1987; Pollack 1986a, b; and papers
(especially by Kautz and Pollack) in Cohen et al. 1990.)
16 Lest this seem too academic a possibility, an anecdote: the Germans are great
hand-shakers; when we were living in Berlin, our Hausmeister, for example,
descended on one, regardless of one's current preoccupations, to grasp the
hand on first and last sighting of the day. But Germans more used to casual
Anglo-Saxon ways curb the custom. Puzzled at first, I found myself inspect-
ing every hand-jerk during greeting/parting moments as a possible candidate
for a prof erred hand, only tofindit turn, more often than not, into a buttoning
of the coat or a struggle with a sleeve!
17 Since conversational response can routinely fall within 200 milliseconds of the
prior utterance, if one modestly ascribes half of that delay to planning of the
response, then that leaves only the other half for comprehension, including
intention- or plan-recognition, of the prior utterance.
18 One is struck too by how our abilities here are not greatly helped by
ratiocinative leisure. For example, historians make a modest, and lawyers
an immodest, living out of pondering on, and quarrelling about,
intention-attribution.
19 In papers circulated prior to the conference behind this volume, Esther Goody
argues that, although primate interactive intelligence presumably preceded
the origins of language, it is language that has projected us beyond our primate
counterparts by allowing the management and codification of social interac-
tion. If one thinks about linguistic ability as a relatively encapsulated human
skill, then its acquisition might be an explanation for our zoom into a sapient
state. But if, as this section sketches, linguistic ability is necessarily and
essentially parasitic on highly evolved interactive reasoning, then language is not
the evolutionary rocket fuel; it's the rocket (see here Sperber and Wilson
1986). One must then accept a synergistic explanation: higher levels of AIP
make higher levels of communication possible, but equally vice versa.
20 If 'the language of thought' is rather independent of 'the language of
communication', then I don't see the latter playing the crucial role in internal
representation of AIP that E. Goody hypothesizes. Alex Wearing points out to
me that the phenomenon of 'gist memory' might argue against my aphorism -
thoughts bleached by time may not be so specific. But, at least when we
communicate about our immediate environs the aphorism would seem to hold
good.
258 Stephen C. Levinson

21 Those who follow Chomsky in thinking of a core linguistic ability as a highly

specialized, innate mental module, must now exclude semantics from that
domain. But many of us think that central aspects of syntax too show the
stigmata of interactive reasoning (see Levinson 1987b, 1991).
22 In what follows I simplify drastically from a complex, intricate, clockwork
series of mechanisms (see Levinsom 1983: ch. 3; 1987b; 1991; forthcoming).
For an alternative version, see Horn 1989.
23 See, e.g., Sacks et al. 1974; for further references see Levinson 1983: ch. 6.
24 This trick, though, may rely on something beyond the simple mathematics of
set partition, e.g., the idea of the uniquely salient solution that lies behind
Schelling's games of coordination.
25 It was striking that in the conference at which this paper was delivered, Drew,
Streeck and myself all produced pre-sequences as prototypical examples of
interactive planning. It then struck us that such four-turn ('pre-sequence')
sequences are perhaps the longest canonical sequences observable in normal
conversation, barring the 'rituals' of greeting and parting. This is surely
striking, especially when one considers that in situations of asymmetric power
and authority (of a kind frequent enough in human societies) one might expect
'superiors' to be able to impose their multi-staged 'plans' on 'inferiors'.
(Indeed, such three- or four- stage planning hardly counts as a major
intellectual achievement for Homo sapiens — Haimoff (n.d.) arguing that
gibbon calls exhibit pre-sequential structure.) Instead of forward imposition
of structure, what one finds in conversation is a robust contingency, no-one,
almost regardless of status or rank, seems able to guarantee what will happen
beyond the turn after next! I think a good case could be made that such turn-
by-turn contingency argues for a fundamentally egalitarian state in the
Garden of Eden: we are as a species adjusted to adjusting to others.
26 There is now a burgeoning literature on non-monotonic reasoning systems
(see, e.g., Ginsberg 1987). But rather than viewing these developments as
technical solutions to how conversational (and more generally interactional)
inference might work, I view them more sceptically as systems that ape the
results of inference under mutually assumed heuristics (see next section, and
Levinson forthcoming: ch. 1). In short, conversational inference is the Ur
form of default reasoning; default reasoning is not some peculiar unmotivated
property of the human mind to be copied slavishly on machine models of
intelligence - default reasoning is a mode of thinking that arises as a necessary
solution to interactional coordination. It may then spill over to other domains
of reasoning - that's the thesis of this chapter - but it is primarily motivated by
the need to find a solution to intention-attribution.
27 See, e.g., Marslen-Wilson et al. 1982; or Tyler 1992.
28 Of course, we can enjoy the jokey qualities of such examples. But specialists in
computational language understanding don't; they are plagued by just those
'silly' misconstruals we enjoy. They have no computationally tractable system
of heuristics under reflexive intentional reasoning to rid themselves of these
(to us) 'obvious' misconstruals. A machine can have a database full of semantic
knowledge, and may be replete with knowledge about probable relations
between things in the world, and still fail to find the 'obvious'. See, e.g.,
Herskovits (1986) on spatial relations like at and in.
 Interactional biases in human thinking 259

29 Another example: 'I've lost my senile grandmother in the department store;

I've got to think what she'll do, expecting her to wander blissfully on.'
(Chances for a quick meeting: slim.) Versus: 'I've lost my wife in the
department store: she'll be thinking where I'll be thinking she'll go.' (Chances
for a quick meeting: good.)
30 See Johnson-Laird and Watson 1977a: ch. 9.
31 'Conditional perfection' was so christened by Geiss and Zwicky (1971). The
inference is often subjectively very strong, as from 'If you pay me $5, I'll mow
the lawn' to 'If you don't pay me $5, I won't mow the lawn'. In the Atlas-
Levinson (1981) scheme of pragmatic inferences this is a generalized conver-
sational implicature, attributable to the Principle of Informativeness, or
Grice's second Maxim of Quantity.
32 Ethnomedicine might provide a rich area for the comparison of cultural modes
of dealing with complex systems, the essential cross-cultural similarity of the
body providing a natural control, as it were.
33 Dorner points out that such behaviour may be perfectly rational in the sense
that there is a rational means-end relation; it is the apparent over-evaluation of
the goal that inclines us to view such behaviour with analytic pity. But
compare the importance attached to the preservation of 'face' in interaction
(Brown and Levinson 1987).
34 Forgetting, on Dorner's analysis, is not (or not only) mere mechanical failure,
as it were, but also the side-effect of abstraction, or pattern-determining
processes (shades of Galton).
35 Brown and Levinson (1987) argue that the protection of a notion of self-
esteem and the projection of esteem for alter, motivate much of the detailed
patterning of social interaction. Thus we can ground in interaction Dorner's
observation that 'In a certain sense, maintaining a positive self-image is the
requirement for acting at all' (Dorner 1987: 36).
36 See Ervin-Tripp (1979), and references cited there, for temporal properties of
turn-taking.
37 One might speculate, indeed, that the temporal characteristics of short-term
memory have evolved just to cope with the short spans posed by the action-
reaction interval, on the one hand, and the maximal conversational sequential
pattern, on the other. There is, for example, a striking parallel between the
maximum capacity of the short-lived buffer known as 'echoic memory' and
MLU (or mean length of utterance in conversation). Or, as Alex Wearing has
put it to me, the properties of short-term memory and limited information-
processing capacity (which together necessitate frequent feedback) show how
Homo sapiens is virtually hard-wired for high-frequency conversational
turn-taking.
38 Such ratiocination is ethnographically real, as we experienced when working
with Aboriginal people in Cape York, still much concerned with the success of
the hunt or fishing expedition. It is not straightforward to work out the
probabilities of whether the mullet will be running at Aylem beach and
whether the water will be clear enough to spear such fish under conditions only
half predictable from the base camp, or whether it might be better to head for
more dependable but less rewarding linefishingoff a mangrove swamp. That's
the stuff and excitement of the hunter's life.
260 Stephen C. Levinson

39 This is, of course, not the view of Piaget, who viewed the logico-mathematical
as the apical intelligence, but it must now be a commonplace in the cognitive
sciences. Computationally, bipedal locomotion is vastly more complex than
calculus. What we can do 'without thinking' we devalue as not real thinking;
hence our disregard for interactional intelligence. Curiously, though, some
logico-mathematical tasks of the highest order are performed by 'idiot savants'
who typically exhibit low IQs and gross interactional inabilities or autism (see
O. Sacks 1985: ch. 23; and more scientifically, Howe 1989). They can calculate
twelve-figure primes 'without thinking', a task for which there is no known
algorithm.
 MICHAEL CARRITHERS

12 Stories in the social and mental life of people

In this chapter I want to discuss what has been called the 'narrative mode
of understanding' (Bruner 1986), the ability to create, narrate, and
comprehend stories. This ability enables us, I suggest, to grasp a flow of
social events and to convey that grasp to others. To share stories in this
way is a particularly powerful form of interactive planning: for in
fashioning an account of what has been happening and what is happening,
we lay down the background against which future mutual action may
sensibly unfold. Stories, moreover, have the capacity to frame a markedly
intricate and elaborate flow of social events, indeed just the sort of flow
that seems even more characteristic of human than of other social primate
societies (Byrne and Whiten 1988; Whiten 1991; Carrithers 1991a, 1992).
We understand our social world by means of stories, and we use those
stories to create distinctively human society.

The general idea

As I understand it, narrative thinking differs from, but complements, the
other means of interactive planning discussed in this volume. This is
partly a matter of scale. Conversation and discourse analysis (Streeck,
Drew, Brown, this volume) work usually on two interlocutors interacting
for perhaps no more than a few seconds or minutes. On the other hand,
narrative thought may easily comprehend more - sometimes many more -
than two people and may cover days, years, or even a lifetime and beyond.
Moreover, the narrative thought transmitted in a story may depend on a
communicative genre (Luckmann, Chapter 8, this volume), and may even
be partly determined by the conventions of a genre. But stories may also
have an originality, that is, an appositeness to a particular uniqueflowof
action, which is not wholly comprehended within the convention alone.
Moreover, I hope to make clear that narrative thought can be evidenced
and conveyed in forms of speech which are not marked as stories at all.
Let me sketch roughly what I think narrative thought involves. It is a
capacity to cognize not merely immediate relations between oneself and

261
262 Michael Carrithers

another, but many-sided human interactions carried out over a period of

time. We might say: humans understand character, a notion which
embodies the understandings of rights, obligations, expectations, pro-
pensities, attitudes and intentions in oneself and many different others;
and plot, which shows the consequences of those characteristics in a
multifarious flow of actions. To put it another way, human beings can
perceive any current action within a large temporal envelope, and within
that envelope they can perceive any given action, not only as a response to
the immediate circumstances, or to the perceived current mental state of
an interlocutor or of oneself, but also as part of an unfolding story. (I owe
this latter formulation to Paul Harris (personal communication).)
I use boxhplot and character here as terms of art. I think it essential that
character be conceived very broadly, since it must comprehend simulta-
neously individuals as having statuses and roles - that is, as standing in
well-precedented relation to one another - and individuals as having
idiosyncratic histories and propensities. There must be some room for
abstraction, so that people can be understood as acting with a generic set
of obligations and rights: as, for example, a lawyer, or a king, or a mother
acts with obligations and rights toward clients, subjects, or sons and
daughters. But the particularity of one person rather than another, of
Hannah rather than Amy, must also be grasped at the same time. We must
understand not just the type of the grandfather, for example, with all the
relevant expectations about how he should act within a family, but also
this grandfather's individual propensities: mellowness or irascibility,
friendliness or aloofness, wisdom or foolishness, and so forth. Whether or
not the western notion of an individuated personality really grew out of a
much earlier sense of people as personae as Mauss suggested (see
Carrithers et al. 1985), narrative understanding must comprehend both
of those possibilities and many others as well. In this respect the notion of
character resembles that of Schutz's 'type'.
But characters, with their relationships, are also set in aflowof events, a
plot, with its sense of plans, goals, situations, acts, and outcomes. As
Bruner (1986: 13) puts it, narrative thought concerns 'intention and
action and the vicissitudes and consequences that mark their course'.
Plots embody what a character or characters did to, or about, or with,
some other character or characters, for what reasons; how people's
attitudes, beliefs, and intentions thereby changed, and what followed on
from that. To comprehend a plot is therefore to have some notion of the
changes in an inner landscape of thought in the participants as well as the
outer landscape of events. Indeed the two are inseparable, because the
metamorphosis of thoughts entails the metamorphosis of social relations
and vice versa.
This metamorphosis arises from the fact that people do things because
 Stories in social and mental life 263

of what others feel, think, and plan. I may apologize because she was angry.
Or I may buy her a Michael Innes thriller because she will enjoy it. I may
explain why I made that remark in a department meeting because my
colleague apparently misunderstood it. Savage repression may be started
because a rival faction plans to overthrow the government. Many, perhaps
all, forms of law are based upon the attribution of intentions or knowledge
to those held accountable. In war, deeds are done because of what the
enemy think or believe. And it is difficult to conceive of conducting the
most elementary interaction of everyday life without attribution of
intentions and knowledge to others: for example, even the simplest
conversation is based on mutual attribution of states of mind to each other
by interlocutors (Bennett 1976; Brown and Levinson 1987; Whiten
1991).
So when we understand a plot, we understand changes of mind and of
relationship, changes brought about by acts. Moreover, we are able to link
acts, thoughts, and their consequences together so that we grasp the
metamorphosis of each other's thoughts and each other's situations in a
flow of action. In this perspective character and plot are indivisible, for we
understand character only as it is revealed to us in theflowof action, and
we only understand plot as the consequence of characters acting with
characteristic beliefs and intentions. With such narrative understanding
people orient themselves and act in an accountable manner, sensibly,
effectively, and appropriately, creating and re-creating complex skeins of
social life.
There is one further trait of narrative thinking worth mentioning, and
that is its specificity: narrative is not, in my understanding, a generalizing
mode of thought, but one which works from particular to particular. It
may seem that this makes narrative thought rather like that great
invention of Alfred Jarry, namely 'pataphysics' or the science of the
particular, whose purpose is to produce scientific statements like 'I have
just taken a small bite out of this apple.' The delightful irony in this idea
derives from a contrast with a notion of abstract qualities being shared by
objects and of proper knowledge as knowledge couched in such abstrac-
tions. On this view it might even be thought improper to dignify with the
name 'thought' anything which did not use abstract classification. But I
suggest, to the contrary, that narrative thought, like metonymic and
metaphoric thought (see Lakoffand Johnson 1980; Lakoff 1987; Fernan-
dez 1974; Sperber 1975), works more in a pataphysical than a metaphysi-
cal way. This is perhaps clearest in the case of specific characters: to
understand, for example, the assertion that 'Saddam Hussein is a Hitler'
may require little or no abstraction at all, but only a sense of what Hitler's
character was in his own very specific plot.
I write as an anthropologist, not as a cognitive psychologist, so there are
264 Michael Carrithers

some issues on which I cannot decisively pronounce. Is narrative thought

wholly distinct or is it made up of cognitive abilities used in other ways as
well? For my purposes the suggestion of Feldman et at. (1990: 3) is
adequate: 'the cognitive processing that we use to interpret human
intentionality in stories is related to the processes we use to understand
human intentionality in life encounters with other people'. The stories
which I will discuss here are indeed closely related to the conduct of actual
life, and they seem to suggest that any sharp distinction between narrative
and other socially applied thought would be artificial.

A Jain tale
I want to set out two illustrations, ethnographic snapshots of the territory
of story in social life. The first stems from an episode which occurred
while I was doingfieldworkamong Jains in Kolhapur, Maharashtra State,
India.
Jains are a severely ascetic minority sect who hold that souls cycle in
endless torment from birth to birth. The cause of this eternal suffering is
the physical and mental pain we inflict on other beings, for by such deeds
we cause defilements to adhere to our own souls, and these defilements
lead inexorably, with law-like regularity, to further painful rebirth. We
may prevent defilement by living a strict life of vegetarianism, celibacy,
truthfulness, and non-attachment to belongings, and we can cleanse
ourselves of already accrued defilement by practising self-mortification.
Jains are thus celebrated for the doctrine of harmlessness or non-violence,
ahimsa, which had great influence on Gandhi. They are celebrated
likewise for the austerities of their munis, monks or ascetics. The munis of
the Digambar sect (the sect with whom I worked in Kolhapur) go
permanently naked, eat once a day, walk the length and breadth of India,
and from time to time remove their head hair by plucking it out with their
own hands. This strenuous spiritual heroism is held in profound rever-
ence by lay Jains.
The Jains among whom I worked were mostly urban businessmen who
were eager to speak of Jainism, and I was frequently offered a long
impromptu religio-philosophical lecture - a well-marked local communi-
cative genre. On the occasion in question I was sitting in the office of a
dealer in agricultural supplies, Mr P (the T ' stands for 'philosophy'). The
following is from my field notes. Mr P spoke in English.
He began treating me to a sermon. Did I know about Jainism? Not much. He told
me that the essence of Jainism is ahimsa. This is non-violence, and Gandhi was
really a Jain ... Did I know what ahimsa means? I did not. Ahimsa is the essence of
all religions, he said. We must do no harm, we must help all beings. Did I eat meat?
I used to, but no longer. Good, he said, that is ahimsa. Ahimsa is always a profit to
 Stories in social and mental life 265

yourself Ahimsd means that we must say ill of no-one, because we might harm
them, but we would anyway harm ourselves. Why? Because to speak ill or to lie is
to speak out of greed and hatred, and these harm ourselves. Ahimsd means no
harm to others, and that means no harm to yourself. Did I think fasting was bad
for health? I hesitated. No! he said. Fasting is good for self, fasting is ahimsd,
because it harms no-one and helps only self...
Mr P seemed to be hitting his stride when someone called on a business matter,
and Mr P asked me to stay for tea, saying that he had to go out but would be right
back. He left, and after a pause a shabby older man who had been sitting in the
corner spoke in Marathi. He was a farmer perhaps, perhaps a poor relation or had
come about a loan. Did I speak Marathi? A little. This, he said, is a story my
grandfather told me. This is very important. Write this down, he said, pointing to
my notebook. There was a great man, a hero, a mahdpurus, who lived right near
here, and one time that man went out to the bulls. While (doing something
unknown to me) to the bulls one of them stood on his hand. What did he do? He
did nothing! He waited and waited, and finally the bull's owner came and saw what
was happening! The owner struck the bull to make it move, and the great man told
him to stop, that the bull did not understand! That is dharma [true religion], he
said, that is genuine jainadharma [Jainism]!
The man - I will call him Mr S for 'story' - told the tale with marked
fervour, but fell silent when Mr P returned and did not speak again. I took
it that he was rebutting or improving on Mr P's account (which I have
very considerably abridged). I later discovered a printed biography of one
Siddhasagar which contained the episode of which Mr S spoke. Siddhasa-
gar lived into the first decade of the twentieth century. Late in life, after
many notable religious deeds as a layman, he became a muni, naked
ascetic, and continued to live an increasingly ascetic life until his death.
His printed biography (Shaha 1983) informs us that he was removing
dung from beneath the bulls when the reported incident happened.
What is the difference between Mr P's philosophical account and Mr
S's story? Let me begin by pointing out an important dimension in which
they do not differ, namely, they both rely on implicit inference. By implicit
inference I mean that these slices of talk, like all discourse, require a
substantial amount of background knowledge and alertness to context to
be intelligible. For example, neither Mr P nor Mr S actually explained the
Jain theory of rebirth. I had to know it, and to infer that it was relevant (on
relevance see Brown and Yule 1983: 68ff.; Sperber and Wilson 1986).
Moreover, when Mr S began to speak I had to infer that he was referring
to what Mr P had said and not striking out on another topic altogether,
such as 'awful things that can happen to you with bulls'. It was only later
that he made the connection outright: 'that is true religion'. Moreover,
both slices of talk presupposed a knowledge of everyday assumptions -
that food, rather than its absence, is good for you, or that one would do
just about anything to get a bull off one's hand.
266 Michael Carrithers

Explicit inference, on the other hand, marked Mr P's, but not Mr S's,
discourse. That is, Mr P's account was organized to demonstrate a
particular process of reasoning and a particular set of propositions about
the world. After the preamble (i.e., after 'we must help all beings') each
segment of the talk had roughly the form: 'you might think X is good for
you, but in fact Y is good for you, because of the entailments of the theory
of rebirth and ahimsd\ Though the talk was impromptu, it was recogniz-
ably patterned after typical Jain ethical reasoning such as has been
cultivated by Jain scholars and philosophers for more than two millennia.
And in fact Mr P was setting out an example of what Bruner (1986) called
the 'paradigmatic mode of understanding', the mode of understanding of
logic and impersonal abstraction, which Bruner contrasts with the
narrative mode of understanding.
I have already suggested that narrative thought differs from such
paradigmatic thought by its plot-like organization, its characters, and its
specificity, but now I want to go a bit further. It is my view - and I think
that of Mr S - that narrative thought is, for some purposes, superior to
paradigmatic thought. For paradigmatic thought, I suggest, cannot so
easily be applied to one's own actual life; it is less persuasive, less vivid,
and less informative. In practice, story is easier to use, conveys more, and
does so more effectively. Narratives have a capacity to move people and,
in so doing, to make things happen.
There are a number of reasons for attributing such potential to
narrative thought. Levinson suggests in this volume (Chapter 11) that
humans have an interactive bias in thought, and part of that bias is a
propensity to regard phenomena as having intentions, that is, as being, in
my terms, characters. Dennett (1987) has made a similar point in regard to
the 'intentional stance'. If this is so, then we might expect listeners to
understand a point more readily through what happens to a specific
character with his or her characteristic intentions. Moreover, Lakoff
(1987), Winograd and Flores (1987), Bourdieu (1977) and many others
have suggested that humans tend to think in specifics, in images,
metonyms, and metaphors, and in terms intelligible through corporeal
rather than ratiocinative experience. And finally Bruner (1987) has
proposed that people generally see themselves in their own lives in
narrative terms, as having a specific character and a specific life plot. For
all these reasons story seems more intimately connected to our fundamen-
tally social and embodied nature than derived forms of abstract and
general reasoning.
The writer of the preface to the biography of Siddhasagar saw the
virtue of narrative clearly. 'It is our experience', he wrote, 'that the life
stories of great men are attractive, informative, and inspiring to people ...
The readers' minds are so concentrated that they attend to nothing else.'
 Stories in social and mental life 267

(Shaha 1983: 1; my translation, my emphasis.) One part of the extra

knowledge gained here, knowledge that no ethical reasoning could
adequately convey, is that Jains find it admirable when someone goes to
such extreme lengths to achieve the project of Jainism.

Short sharp stories

Let me look more closely at Mr S's discourse. For the moment I will speak
of that part of it that was straightforwardly narrative, namely the account
of Siddhasagar and the bull. The plot of this was minimal, but it was a plot
all the same because (1) it shows aflowof events, (2) it concerns specific
characters, (3) it displays the attitudes, beliefs and intentions of the
characters and (4) it reveals the relationship between events and those
intentions and attitudes.
But however spare, the story was not really straightforward. Take, to
begin with, the purpose of Mr S's discourse, which was to reveal
Siddhasagar's spiritual heroism. This evidently involved attributing
character, with typical intentions and attitudes, to Siddhasagar - yet there
was in the telling no explicit attribution to him of any mental qualities at
all. The only mental attribution, indeed, was to the bull, which 'did not
understand'. So the story, as told, required background information, and
in fact the entire import of the narration had to be inferred. However this
story meant what it meant, it did not do so by putting thoughts directly
into words.
I make this point partly because an important view of narrative, that set
out by Bruner (1986; see also Feldman et at. 1990), suggests that
narratives might be divided into different kinds according to their verbal
form: one kind possesses a 'landscape of action', the other a 'landscape of
consciousness'. On this view some stories, such as the Russian folk tales
analysed by Propp, comprise largely a description of events, but contain
few or no attributions of attitudes, beliefs, and intentions. Modernist
stream-of-consciousness stories, on the other hand, are quite the
opposite, being full of mental description and very little action.
But what I suggest is rather different. Verbal form may be decisive for
certain forms offictionor for certain genres, and these may perhaps stress
one or the other of the landscapes. But stories that are closely connected to
the conduct of life must necessarily concern both landscapes, for narrative
is about orienting people in a stream of events constituted of both action
and consciousness. And so we have a story such as that of Siddhasagar and
the bull, which on the surface consists almost solely of action but which in
fact concerns - and must concern, to be of any use - the union of
consciousness and action.
These considerations of verbal form suggest, I think, that there can be
268 Michael Carrithers

no minimal definition of what constitutes a narrative apart from its

setting. Gergen and Gergen (1984), for example, have suggested that an
utterance such as 'I thought she was my friend' could be a complete story
- but of course it presupposes an implied setting of friendship and
disappointed expectations. Similarly, the words 'I came, I saw, I con-
quered', if uttered to the right people by the right person at the right time,
are a complete story, rich with mental and character attribution
(especially by the speaker to himself).
These examples have at least a temporal reference marked by a past
tense or by word sequence in the sentence, but what about the warning
'don't be a Miss Bossy!', as issued to my daughter, with whom I had
recently read a story by that name (Hargreaves 1981)? This has no
temporal marker at all, but I suggest that it is an important example of
narrative all the same, for by mentioning a character it also evokes acts,
attitudes, and the relationship between the two. Sperber and Wilson
(1986: 48-51) and Bakhtin (in Holquist 1990: 62) give examples of slight
signals - a movement, a phatic utterance - which are barely verbal but
which are nevertheless richly communicative and precise, and I see no
reason to exclude narrative from the sorts of significance that could be so
sketchily conveyed. To accept that a story could be evoked rather than
told is no more than a corollary of the observation that implicit inference
is necessary to all stories. My point is not that such compact utterances
should be regarded solely and exclusively as narratives, but rather that to
see them as narratives is to show how they orient people to events and
intentions in aflowof action. I will refer to brief, un-story-like evocations
as minimal narratives.

Planning my future
I want now to pursue the nature of implicit inference further, and propose
that the implied meaning in Mr S's telling of the story arose in part
through the collaboration of us interlocutors rather than merely through
individual ratiocination. The flow of action of the story, that is, was
understood partly through the flow of action of its telling.
Let me begin with the events in which Mr P and I were already involved
before Mr S spoke. My own part in them had a well-established
background, in that I was understood among local Jains to be a student of
Jainism and a researcher of the Jain community, intent on recording these
important matters for posterity in a complimentary light. In the fore-
ground I demonstrated, by quick and encouraging responses and by
writing in my notebook as the encounter progressed, that I valued Mr P's
opinion. Mr P, for his part, fell into the advisory, admonitory tone of the
impromptu lecture. And so we established that his utterances were for the
 Stories in social and mental life 269

record, for my notebook; and that as such they were authoritative,

important, and reasonable. So if he said things that might seem to run
against the grain of (some assumed) common sense, then we would both
know that good reason for such assertions would be forthcoming.
When Mr P stepped out of this flow, Mr S stepped in. He said, and did,
this:
Did I speak Marathi? A little. This, he said, is a story my grandfather told me.
This is very important. Write this down, he said, pointing to my notebook. There
was a great man ... who lived right near here ...
Note first the parts of this utterance that were directed toward
continuing theflowof action and preserving the generous interpretation
of utterances in the same terms. Mr S required me to speak Marathi, to
agree on the worth of what was to be said, and to record it for posterity.
For my part I had to assent to these terms, which I did by complying. So
we established a commonality of understanding that might otherwise, had
the interlocutors been closer in experience, have been left unsaid.
The other part of Mr S's prelude was this:
This, he said, is a story my grandfather told me ... There was a great man... who
lived right near here ...
This part of the prelude might seem formulaic, but in the setting it had
a particular force. I took it at the time, and it was subsequently confirmed
by finding Siddhasagar's biography, that Siddhasagar's life may have
overlapped with that of Mr S's grandfather. Moreover Siddhasagar lived
and travelled among the villages just to the south of Kolhapur. So his was
not a story of remote, unreckonable antiquity, as so much of Jain story
literature is, but one which occurred within Mr S's known world.
Mr S was bound to his grandfather, and his grandfather was bound
through further (inferrable) connections, the connections which bore the
tellings and re-tellings of the story, to Siddhasagar himself. So in effect
the prelude was part of the story. It gave the tale further characters, the
grandfather and the younger Mr S, who were implicitly assigned
the intention to speak and listen most gravely, concerning issues of the
highest import. The appended incident was set half-way between Sidd-
hasagar's deed and the present, in Mr S's childhood, and it added to the
plot an account of how a famous deed passes among people, one
authoritative speaker to the next, as the discourse of the ends of life. In the
present telling, implied Mr S, he was playing the role toward me that his
grandfather had played toward him in the earlier telling.
So on the one hand Mr S was urging me to consider a specific past flow
of events in which present action was to be oriented. On the other hand,
this account of the past was directed toward the future as well. Mr S's
270 Michael Carrithers

intention was to produce in me, with my compliance, an understanding of

Jainism and an emotional orientation informed by Jain values. This could
reasonably be regarded as interactive planning, even though Mr S was
working not on an egalitarian but on a hierarchical notion of interaction.
And in fact it might be better to speak, not of planning, but of deuteroplan-
ning, planning how to plan, suggesting a general orientation to life that
would mould any later episode of specific planning. Mr S was not
proposing specific actions in the short or medium term, but rather a
wholesale reorientation of how to plan altogether.

Kaluli stories
I have considered this Jain material because it illustrates many things: the
distinction between narrative and other forms of thought, the allusive and
cooperative nature of story, and its power to call into relevance events
long past and to adumbrate flows of action on a very large scale. But the
mutually planned outcome was purely a change in attitude and under-
standing and so does not show one of the most important features of
narrative thought, namely its close integration with social action. For that
I turn now to E. Schieffelin's (1976) ethnography of the Kaluli, a people
of the Papua New Guinea plateau. These narratives are much closer to an
immediate flow of events and therefore illustrate more plainly how people
use narrative thought to construct their mutual life.
The ethnography of the Kaluli recapitulates many themes found
throughout Papua New Guinea. They are shifting cultivators who
supplement their diet by hunting, fishing, and pig raising. Their social
intercourse is dominated by exchange and reciprocity. Such reciprocity
may create and reaffirm friendly relations, and Schieffelin makes clear
that, in such exchange, the expression of sentiment is at least as important
as material considerations. But Kaluli also reciprocate violence with
violence, and the abiding threat of conflict influences many of their
institutions.
Kaluli live in large communal longhouses distributed over their
territory. But though the most immediate way of identifying a Kaluli is by
identifying his or her longhouse, large-scale social events are not neces-
sarily organized according to longhouse membership. Nor do the Kaluli
possess superordinate coercive institutions. Hence their domestic and
kinship arrangements are relatively fluid, and most of their collective life -
work, play, ceremony and fighting - is achieved by groups assembled ad
hoc for an occasion. Schieffelin represents the organizing of ad hoc groups
as largely a male matter, and I will therefore be concerned chiefly with
relations between adult men (for a complementary and corrective view-
point see B. Schieffelin 1990).
 Stories in social and mental life 271

Schieffelin stresses that even though the group formed on a particular

occasion may appear to come together through ties of kinship or friend-
ship, there is no sense in which mutual aid is obligatory or automatic. On
the contrary, Kaluli styles of interaction assume that interlocutors are
autonomous, that participants in a common activity are equal, and that
any cooperation arises out of freely chosen fellow feeling. Thus a group
gathered for a collective purpose, such as damming a river for fishing,
takes cognizance of itself according to the task to be performed, not the social
relationships involved. The result, in most group activity, is visible in the posture
of autonomy preserved by each man engaged in the common project. (1976: 132.)
How do the Kaluli manage collective action in such an aggressively
egalitarian milieu, where 'any appearance of compulsion infringes on
one's sense of personal autonomy and invites an angry reaction, even
between children and adults' (p. 129)? Schieffelin explains that Kaluli
have two modalities or styles of addressing requests for action to each
other: assertion and appeal. Assertion takes the form of loud, imperative
statements which have the appearance of orders or demands, even though
in fact they can count as no more than forceful requests. Assertion appears
to be used most frequently with close relatives and friends. It works most
effectively, suggests Schieffelin, against a background of excitement,
helping to initiate action by being itself'exciting, noisy, and dramatic' (p.
129). Appeal, on the other hand, is most often used between people who
are more distant, or in moods or circumstances that are more sombre, and
it may be conveyed either explicitly or nonverbally, as by sighs or a
dejected posture, for example. But however a request for help is framed,
Kaluli describe their reasons for responding 'in terms of sympathy and
compassion - "I felt sorry for him so I went" - or supportive outrage - " I
was angry when those people killed his pig"' (p. 130).
Schieffelin provides us with an example of these processes in action. To
understand what follows it is necessary to know that a sei is a witch, a
person deemed responsible for causing or promoting another's death by
supernatural means. Before pacification, seis used to be killed by bereaved
relatives of their victims.
When Dasemi of Olabia [longhouse] died from snake bite, Jomo of Ferisa was
publicly identified as the sei responsible. At the funeral ... Dasemi's husband,
Beli, sat forlornly in the rubbish near the back of the [longhouse]... Around him
visiting men seated on the sleeping platforms shouted speeches of outraged
sympathy. The children had no mother now and Beli had no one to cook for him,
they yelled at me in anger and dismay. Others shouted at newcomers that Jomo
had taken Beli's wife and should be made to give return. If it were not for the
government, they declared, they would go right out and kill him. This was how
support for the murder of [an alleged sei] had materialized almost of itself twelve
years before, (p. 130.)
272 Michael Carrithers

Here I am concerned with the narrative thought conveyed in the

assertions made to Schieffelin and the newcomers. One assertion was that,
because of Jomo's act, 'the children had no mother now and Beli had no
one to cook for him'. The other was that 'Jomo had taken Beli's wife and
should be made to give return. If it were not for the government... they
would go right out and kill him.'
Of course these are not stories in a conventional sense. Unlike the story
of Siddhasagar and the bull, these utterances are not preceded by a 'once
upon a time ...' marker. Nor are they placed entirely in the past tense:
some of the utterances concern recent past events, but others concern the
present or future. From a strictly linguistic point of view, these utterances
are a different genre, quite unlike that of Siddhasagar and the bull. But on
the other hand they do act as I have suggested narrative must: they show
how people coming upon the scene were informed of the flow of events in
terms of previous events - Jomo's mystical deed, Dasemi's death - and
previous and present mental states - Jomo's evil attitude, Beli's grief,
others' anger. In other words, these speeches evince narrative thought
without having narrative form. To that extent they may be regarded as
(among other things) minimal narratives, and as minimal narratives they
connect closely to ongoing events (as did that other minimal narrative, 'I
thought she was my friend.').
Minimal narratives plainly depend on inference, and some of the
intentions evoked by these were certainly implicit: for example, to speak
of sei at all is to evoke his or her evil attitude. But some mental states were
made explicit, and were perhaps more explicit in the event than in
Schieffelin's brief depiction of it. For as he makes clear elsewhere, and
may suggest here with the words 'they yelled at me in anger and dismay',
Kaluli both display their feelings openly and say explicitly that they feel
'angry' or 'sad' (see E. Schieffelin 1976: 135).
It would be quite contrary to what Kaluli say and do on such an
occasion to divide speech marked by reference to the past - Dasemi's
death, Jomo's alleged activity as a sei- from that which describes present
circumstances or proposes future action - people's anger, a proposed
demand for compensation. As Schieffelin depicts the speeches, they move
effortlessly and naturally from showing what has happened to what
people now feel, and from what people now feel to what should happen.
Here, in medias res, narrative thought welds the past seamlessly to the
present and future, such that what to do seems to follow inevitably from
what has happened. Indeed by suggesting appropriate action (to go to
Jomo and demand compensation) Kaluli were commenting on the past
(the evil of Jomo's deed) as much as they were deriving present plans from
that past. Similarly, it would be a distortion to divide too sharply the
saying from the doing: where Schieffelin writes that men 'shouted
 Stories in social and mental life 273

speeches of outraged sympathy', I suppose him to reflect that people were

both showing themselves to be emotionally excited and going on record as
being angry.
I make these points in order to suggest how closely narrative thought
can be bound up with action. The immediately proposed plan, to go to
Jomo with demands, derived its sense from the flow of events which led
up to it. Without understanding these events, present proposals would be
senseless. Moreover people were both acting angrily and immediately
reflecting that anger as part of their narrative of events, so their anger
continually became a recorded, significant part of the flow. These
continual re-tellings then pointed forward compellingly to the rest of the
story, the future, just as Siddhasagar's story pointed toward a desired
result. We might say: as people composed the immediate past and present
for each other, they composed the future as well. Their speech had a
cognitive and an emotive or evaluative element, but also a conative
element, an element of will and determination. In the excitement, each re-
telling of events demonstrated the speaker's own willingness to face the
consequences. Each invited others to join in, and each contributed to the
growing consensus over what had, and what should, happen. Esther
Goody suggested to me the term 'confabulation' to cover this process.

The long term

Schieffelin argues not just that Kaluli confabulate their response to
events, but that they do so in a particular way, and I want now to look at
their style in the more leisurely creation of long-term relationships.
Schieffelin makes abundantly clear that material gifts and material help
are regarded by them straightforwardly as signs of love, affection and
esteem. Thus their sentimental dimension is much more salient than we
would easily recognize. He also shows that relatively formal gifts between
future in-laws initiate a marriage, and that continued gifts sustain both
the marriage and the newly created relationship between brothers-in-law,
real and classificatory. Indeed the greater part of work done by Kaluli in
mature life to gain their livelihood and feed their children is done by men
and women labouring together within the affinal relationship thus
created. The relationship between marriage partners and between affines
is therefore both sentimental and urgently practical. The consequence of
all this, writes Schieffelin (p. 25), is that
A man's relations with, interactions with, and affection for his affines ... are so
bound up with situations of reciprocal gift-giving and mutual help that he tends to
think of his life with his wife, on reflection, in terms of the situations in which he
worked cooperatively with her and reciprocally with her relatives and exchanged
countless minor gifts of food.
274 Michael Carrithers

In this light - and 'on reflection' - an older man, Sialo, of Bona

longhouse, told Schieffelin the story of his marriage:
(After a ceremony at Tusuku longhouse I met a woman.) I told her to wait. Later
Tusuku (people - his future in-laws) brought her to (my longhouse) Bona for a
formal wedding.
Well, we went down and planted pandanus and bananas. Then my brother-in-
law Kiliyae called us to plant bananas. We went to Tusuku and planted many,
many bananas, and cut many trees.
Later Tusuku brought a cassowary. I brought them bananas I cut from my
garden. I went to (Mt.) Bosavi and hunted for animals. I got many, many animals,
and also sago grubs and brought them to Tusuku. Then Tusuku went to (Mt.)
Bosavi and hunted animals and got sago grubs and brought them to Bona as
return. That's the way it was!
Well we went down to Alimsak (a garden place). We planted okari nuts. We
called Tusuku, 'Come plant bananas!' We planted bananas one day, two days,
five, six days. Then we cut trees. Being hungry we said, 'You take this sago.' 'You
go pick bananas and pandanus.'
We were hungry for meat. We (invited Tusuku) to the Walu (stream). We
caught crayfish. We dammed the Isawa, and beat poison and caught many, many
fish. We went to the gardens. Tusuku said, 'Bring bananas!' So we brought
bananas. 'Bring more'; so we brought more.
At Balasawel the woman died. I gave it up. {Ibid.: 26.)
B. Schieffelin (1990: 246) gives a clue to the narrative style of Kaluli in
general. She remarks that Kaluli story-tellers 'can evoke the sentiments of
the participants in a compressed discursive space ... and while there is
provided an outcome to which a response is certain, response is not given
by an authoritative or judgmental stance of the teller'. The style of Kaluli
narrative is not to draw moral points explicitly, but to show the texture of
human life by a muted narrative, a relatively bare reflection. They are
more modernist than homiletic, more in the style of Hemingway than of
the preacher Jonathan Edwards. So a story such as that of Sialo is very
differently framed from that of Mr S, for whom the establishing of
authority was one of the main concerns. But not just any evocation would
do, and as B. Schieffelin makes clear, and as E. Schieffelin (1976) and Feld
(1982) agree, the theme of interdependence and its difficulties would be
uppermost in Kaluli reception of the story. They would understand the
importance of these events to Sialo, the way in which they stood as
landmarks on the course of his life.
The measure of that importance depends critically on the sheer scale of
what is being narrated. Sialo's story frames the whole of his adult life. The
gain of a wife and in-laws created Sialo's very character as an adult, and
their loss marked his transition to another stage of life: 'I gave it up.'
Within that frame, moreover, acts of reciprocity and their accompanying
intentions of amity created not just a series of atomic situations, but rather
 Stories in social and mental life 275

a developing narrative of increasing affection, esteem, and mutual reliabi-

lity, of characters mutually made and tested in work and appreciation. It
was the cumulative effect of all those acts of trust in the developing plot of
their relationship which created the sentiment which bound Sialo, his
wife and her brothers; and it was a sense of that amassed value of feeling
that created the sense of loss with which the story ends. If the relation-
ships described in the story were not more than the sum of their individual
occasions, the story could not have had such force.
How was this narrative understanding used interactively? Schieffelin
reports (personal communication) that Sialo told him the story early in
their acquaintance when they were getting to know one another. He also
reports that the story was consistent with such a situation, namely that it
was offered as something like a brief curriculum vitae. The incidents
chosen to illustrate his life by Sialo were conventional in that they were
the sort of events commonly used by Kaluli as chronological markers in
veridical narratives, though other stories would be likely to be more
digressive and less concentrated, less full of such markers. The effect was
to orient Schieffelin to the plot of Sialo's life and thereby to inform him of
Sialo's character. This, Sialo implied, is the way a Kaluli life, my life,
works out, this is the way I laboured and showed proper affection to my
fellows, this is the way relationships, and gardens, are rightly made and
cultivated.
So Sialo's autobiography was offered as part of a mutual informing
which would subsequently allow the two to work together - as, says
Schieffelin, they eventually did. It was a self-revelation which formed the
basis of a bond. We can understand, too, that this cannot have been the
only such occasion: there would have been other, earlier, fragmented
tellings of the autobiography in medias res, tellings of successful
exchanges by Sialo and his in-laws while he was young and his wife was
prospering. Those tellings would have drawn others in, carrying forward
the process of forming and cementing relationships just as this one did.

Confabulation
So people confabulate society using stories. The word 'confabulate' in its
common acceptation means just chatting together, with a suggestion
perhaps of yarning, gossiping, or story-telling. I suggest that we should
add a further implication to its sense, namely that of making together,
'confabrication' if you will. What is thus made is characteristically human
social action, with its large temporal and causal perspective and its
cognizance of many related characters with their richly imagined mental
states. Godelier remarks that 'human beings, in contrast to other social
animals, do not just live in society, they produce society in order to live'
2 76 Michael Carrithers

(Godelier 1988:1). The sense of this is well illustrated by the Kaluli, who
form work groups along affinal lines to produce their livelihood. But the
Kaluli also produce stories in order to live in society. Stories of conflict
constitute society through a conflictual understanding, stories of amity
constitute it through a cooperative understanding. In either case, the
mutual knowledge created through narrative thought ensures that subse-
quent action by partners to a relationship will act appropriately and with
mutual regard - even where that regard is mutually destructive.
Finally I want to point out that stories and story consciousness help to
confabulate not just action, but more or less predictable action in a
familiar style. I have argued that stories and narrative consciousness are
specific, particular, effectively 'pataphysical' in their form and appli-
cation. It has however also been implicit in my presentation that there are
differences of culture or social structure which set apart one style of story
or story consciousness from another. The Kaluli have a different view of
what makes a good or intelligible story, and of what makes an appropriate
response in life, than do the Jains, and the ethnographic literature could
be consulted for many more variants. E. Schieffelin has captured the style
of story and action specific to a society in the phrase 'cultural scenario',
which is recognizable in the way 'a people repeatedly approach and
interpret diverse situations and carry them to similar types of resolution'
(1976: 3). In this respect Kaluli or Jain stories have a generic character,
not just a specific, pataphysical one.
Now the features of a cultural scenario may be largely implicit for the
people themselves, as they seem to be for Kaluli. In that case, ethnogra-
phers are left to produce a theoretical gloss, their own paradigmatic
thought, which describes the generic element in Kaluli narrative and
social dispositions. In other cases, such as among the Jains, the people
themselves may cultivate their own commentary with great industry. But
in either case, narrative thought allows people to weld the general
disposition to the particular action, the general aesthetic feeling for
outcomes to the particular story, the generic role to the particular player.
In this perspective the central characteristic of story is neither its
predictability nor its originality, but rather its ability to combine the two.
Stories must be intelligible, but within the bounds of intelligibility they
may still tell a story which interlocutors have never heard before. Indeed,
as Chafe argues, people tell stories in order to say something new:
'narratives that entirely fit expectations are not really narratives at all'.
(Chafe 1990: 83) Consequently, I suggest, we may in future want to look
more closely to story to see more clearly how people mutually construct a
social world that resembles the past while still creating new inventions
and new responses.
 Consolidated bibliography

Abrahams, R.G. (1989) Law and order and the State in the Nyamwezi and
Sukuma area of Tanzania. Africa 59: 356-70.
Ahem, E.M. (1973) The Cult of the Dead in a Chinese Village. Stanford, Calif.
Alhonsaari, A. (1973) Prayer; an analysis of theological society. Publication of the
Luther-Agricola Society: B8. Helsinki.
Allen, J. (1983) Recognizing intentions from natural language utterances. In M.
Brady and R.C. Berwick (eds.) Computational Models of Discourse. Cam-
bridge, Mass.
Alverez, T. (1980) Living with God: St. Theresa's concept of prayer, edited by T.
Curran. Dublin.
Anderson, J.R. (1984) Monkeys with mirrors: some questions for primate
psychology. Journal of Comparative Psychology 99: 211-17.
Anesaki, M. (1918) Buddhist prayer. In J. Hastings (ed.) Encyclopedia of Religion
and Ethics, Vol. X. Edinburgh.
Argyle, M. (1968) The Psychology of Interpersonal Behaviour. London.
Aristotle [1962] The Nichomachean Ethics. Indianapolis, Ind.
Arundale, R.B. (1993) Cultural assumptions in conceptualizing face. Paper
presented to the International Communication Association, Washington
D.C., May.
Atkinson, J.M. (1984) Our Master's Voices: the language and body language of
politics. London/New York.
Atkinson, J.M. and Drew, P. (1979) Order in Court: the organisation of verbal
interaction in judicial settings. London.
Atkinson, J.M. and Heritage, J. (eds.) (1986) Structures of Social Actions: studies
in conversation analysis. Cambridge.
Atlas, J. (1989) Philosophy without Ambiguity. Oxford.
Atlas, J. and Levinson, S. (1973) The importance of practical reasoning in
language usage: an explanation of conversational implicature. Unpublished
manuscript, Department of Linguistics, University of Cambridge.
(1981) It-clefts, informativeness and logical form. In P. Cole (ed.) Radical
Pragmatics. New York.
Austin, J.L. (1962) How to do Things with Words. Oxford.
Averna, G. (1977) Italian and Venetian profanity. Maledicta 1.
Avramides, A. (1989) Meaning and Mind. Cambridge, Mass.

277
278 Consolidated bibliography

Bailey, R. et al. (1989) Hunting and gathering in tropical rain forest: is it possible?
American Anthropologist 91: 59-82.
Bakhtin, M.M. (1965) Rabelais and his World. English translation, 1968. Cam-
bridge, Mass.
(1986) Speech Genres and Other Late Essays. Austin, Texas. (Partial translation
of Estetika slovesnogo tvorchestva. Moscow, 1979.)
(1986) [1952-3] The problem of speech genres. I n C . Emerson and M. Holquist
(eds.) Speech Genres and Other Late Essays. Austin, Texas.
Baldi, C. (1623) Delle Mentite. Bologna.
Barnard, A. (1981) Universal categorization in four Bushmen societies. L'Uomo 5:
219-37.
Barnes, J.A. (1994) A Pack of Lies. Cambridge.
Baron-Cohen, S., Leslie, A.M. and Frith, U. (1985) Does the autistic child have a
theory of mind? Cognition 21: 37-46.
Barth, F. (1966) Models of Social Organization. Royal Anthropology Institute
occasional paper no. 23. Glasgow.
Bascom, W.R. (1954) Four functions of folklore. Journal of American Folklore 67:
333-49.
(1969) If a Divination: communication between gods and men in West Africa.
Bloomington, Ind.
Bauman, R. (1983) Let Your Words Be Few: symbolism of speech and silence among
seventeenth-century Quakers. Cambridge.
Bauman, R. and Sherzer, J. (eds.) (1974) Explorations in the Ethnography of
Speaking. Cambridge.
Beattie, G.W. (1979) Planning units in spontaneous speech: some evidence from
hesitation in speech and speaker gaze direction in conversation. Linguistics
17: 61-78.
(1983) Talk. Milton Keynes.
Beattie, J.H.M. (1964) Divination in Bunyoro, Uganda. Sociologious 14: 44-61.
(1966) Consulting a diviner in Bunyoro: a text. Ethnology 5: 202-17.
(1967) Consulting a Nyoro diviner: the ethnologist as client. Ethnology 6:
57-65.
Beavin, J.H., Jackson, D.D. and Watzlawick, P. (1969) Menschliche Kommunika-
tion> Formen, Storungen, Paradoxien. Bern, Stuttgart and Vienna.
Bell, D.E., Raiffa, H. and Tversky, A. (eds.) (1988) Decision Making: descriptive,
normative and prescriptive interactions. Cambridge.
Bennett, J. (1976) Linguistic Behaviour. Cambridge.
Berger, P.L. and Luckmann, T. (1966) The Social Construction of Reality: a
treatise in the sociology of knowledge. New York.
Bergmann, J.R. (1993) Gossip. New York. German edition, 1987. Berlin.
Bickerton, D. (1990) Language and Species. Chicago.
Binsbergen, W. van and Schoffeleers, M. (eds.) (1985) Theoretical Explorations in
African Religion. London.
Bird, N. (subsequently Bird-David) (1982) 'Inside' and 'outside' in kinship
usage: the hunter-gatherer Naiken of South India. Cambridge Anthropology
7(1): 47-57.
 Consolidated bibliography 279

(1983a) Conjugal families and single persons: an analysis of the Naiken social
system. Ph.D. dissertation, University of Cambridge.
(1983b) Wage-gathering: socio-economic change and the case of the Naiken of
South India. In P. Robb (ed.) Rural South Asia: linkages, change and
development. London.
Bird-David, N. (1987a) The Kurumbas of the Nilgiris: an ethnographic myth?
Modern Asian Studies 21: 173-89.
(1987b) Single persons and social cohesion in a hunter-gatherer society. In P.
Hockings (ed.) Dimensions of Social Life: essays in honour of David Mandel-
baum. Berlin.
(1988) Hunter-gatherers and other people: a re-examination. In T. Ingold, D.
Riches and J. Woodburn (eds.) Hunters and Gatherers. Vol. I: History,
Evolution and Social Change. Oxford.
(1989) An introduction to the Naikens: the people and the ethnographic myth.
In P. Hockings (ed.) Blue Mountains: the ethnography and biogeography of a
South Indian region. New Delhi.
(1990) Giving environment: another perspective on the economic system of
gatherer-hunters. Current Anthropology 31: 189-96.
(1992a) Beyond 'the original affluent society': a culturalist reformulation.
Current Anthropology 33: 189-96.
(1992b) Beyond 'the hunting and gathering mode of subsistence': observations
on Nayaka and other modern hunter-gatherers. Man (n.s.) 27: 19-44.
(1994) The Nayaka. Encyclopedia of World Cultures. Human Relations Area
File: New Haven, Conn.
Boesch, C. (1991) Teaching in wild chimpanzees. Animal Behaviour 41: 530-2.
(forthcoming) Teaching in Wild Chimpanzees.
Boesch, C. and Boesch, H. (1989) Cooperative hunting among wild chimpanzees
in the Tai National Park. American Journal of Physical Anthropology 78:
547-73.
Bourdieu, P. (1977) Outline of a Theory of Practice. Cambridge.
Boyce, M. (1979) Zoroastrians: their religious beliefs and practices. London.
Boyd, R. and Richardson, P. (1985) Culture and the Evolutionary Process.
Chicago.
Briggs, J.L. (1970) Never in Anger: portrait of an eskimo family. Cambridge, Mass.
Britton, B. and Pellegrini, A. (eds.) (1990) Narrative Thought and Narrative
Language. Hillsdale, N J.
Brother, O.P. (1979) The Little Way: the spirituality of Therese of Lisieux.
London.
Brown, G. and Yule, G. (1983) Discourse Analysis. Cambridge.
Brown, P. (1979) Language, interaction and sex roles in a Mayan community: a
study of politeness and the position of women. Ph.D. dissertation, University
of California, Berkeley.
(1980) How and why are women more polite: some evidence from a Mayan
community. In S. McConnell-Ginet, R. Borker and N. Furman (eds.)
Women and Language in Literature and Society. New York.
(1990) Gender, politeness and confrontation in Tenejapa. In D. Tannen (ed.)
280 Consolidated bibliography

Gender and Conversational Interaction, special issue of Discourse Processes

(Spring). Reprinted in D. Tannen (ed.) Gender and Conversational Interac-
tion. Oxford.
Brown, P. and Levinson, S. (1978) Universals in language use: politeness
phenomena. In E. Goody (ed.) Questions and Politeness: strategies in social
interaction. Cambridge.
(1987) Politeness: some universals in language usage. Cambridge.
Bruner, J.S. (1972) Nature and uses of immaturity. American Psychologist 27:
687-708.
(1978) From communication to language: a psychological perspective. In I.
Markova (ed.) The Social Context of Language. Chichester.
(1986) Actual Minds, Possible Worlds. London.
(1987) Life as narrative. Social Research 32: 11-32.
(1990) Acts of Meaning. Cambridge, Mass.
Bruner, J.S. and Sherwood, V. (1976) Peekaboo and the learning of rule
structures. In J.S. Bruner, A. Jolly and K. Sylva (eds.) Play: its role in
development and evolution. Harmondsworth.
Buber, M. (1900) land Thou.
(1970) Between Man and Man. New York.
Burke, K. (1945) A Grammar of Motives. New York.
Burke, P. (1987) Insult and blasphemy. In P. Burke, The Historical Anthropology
of Early Modern Italy: essays on perception and communication. Cambridge.
Byrne, R.W. and Whiten, A. (eds.) (1988) Machiavellian Intelligence: social
expertise and the evolution of intellect in monkeys, apes and humans. Oxford.
(1990) Tactical deception in primates: the 1990 database. Primate Report 27:
1-101.
(1991) Computation and mindreading in primate tactile deception. In A.
Whiten (ed.) Natural Theories of Mind: evolution, development and simulation
of everyday mindreading. Oxford.
(1992) Cognitive evolution in primates: evidence from tactical deception. Man
(n.s.) 27: 609-27.
Carrithers, M. (1988) Passions of nation and community in the Bahubali Affair.
Modern Asian Studies 22: 815-44.
(1991a) Narrativity: mindreading and making societies. In A. Whiten (ed.)
Natural Theories of Mind: evolution, development and simulation of everyday
mindreading. Oxford.
(1991b). The foundations of community among southern Digambar Jains: an
essay on rhetoric and experience. In M. Carrithers and C. Humphrey (eds.)
The Assembly of Listeners: Jains in society. Cambridge.
(1992) Why Humans have Cultures: explaining anthropology and social diversity.
Oxford.
Carrithers, M., Collins, S. and Lukes, S. (eds.) (1985) The Category of the Person:
anthropology, philosophy, history. Cambridge.
Chafe, W. (1990) Some things that narrative tells us about the mind. In B. Britton
and A. Pellegrini (eds.) Narrative Thought and Narrative Language. Hills-
dale, N J .
 Consolidated bibliography 281

Chauvin, Y. (1988) Symbol acquisition in humans and neural (PDP) networks.

Ph.D. dissertation, University of California, San Diego.
Chaytor, M. (1980) Household and kinship. History Workshop Journal 10: 25-51.
Cheney, D.L. and Seyfarth, R.M. (1988) Social and non-social knowledge in
vervet monkeys. In R.W. Byrne and A. Whiten (eds.) Machiavellian Intelli-
gence: social expertise and the evolution of intellect in monkeys, apes and humans.
Oxford.
(1990) How Monkeys see the World. Chicago.
Chomsky, N. (1966) Cartesian Linguistics. New York.
(1968) Language and Mind. New York.
(1980) Rules and Representations. New York.
Cicero, Q. [1923] De Senectute, De Amicitia, De Divinatione, translated by W.A.
Falconer: Loeb Classical Library, London.
Clark, H. (1992) Arenas of Language Use. Chicago.
Clark, H. and Carlson, T. (1982) Speech acts and hearer's beliefs. In N. Smith
(ed.) Mutual Knowledge. London.
Clark, H. and Marshall, C. (1981) Definite reference and mutual knowledge. In A.
Joshi, I. Sag and B. Webber (eds.) Elements of Discourse Understanding.
Cambridge.
Clutton-Brock, T.H. and Harvey, P.H. (1980) Primates, brains and ecology.
Journal of Zoology 190: 309-23.
Cohen, P., Morgan, J. and Pollock, M. (1990) Intentions in Communication.
Cambridge, Mass.
Cole, M. (1985) The zone of proximal development: where culture and cognition
create each other. In J.V. Wertsch (ed.) Culture, Communication and Cogni-
tion. Cambridge.
Connolly, P. (1978) Hannibal and the Enemies of Rome. London.
Corum, C. (1975) Basques, particles and babytalk: a case for pragmatics.
Proceedings of the Berkeley Linguistics Society 1: 90-9.
Coulmas, F. (ed.) (1981a) Conversational Routine: explorations in standardized
communication situations andprepatterned speech. The Hague, Paris and New
York.
(1981b) Poison to your soul, thanks and apologies contrastively viewed. In F.
Coulmas (ed.) Conversational Routine: explorations in standardized communi-
cation situations and prepatterned speech. The Hague, Paris and New York.
Cranach, M. von, Foppa, K., Lepenies, W. and Ploog, D. (eds.) (1979) Human
Ethology: claims and limits of a new discipline. Cambridge.
D'Andrade, R. (1981) The cultural part of cognition. Cognitive Science 5:179-95.
(1984) Cultural meaning systems. In R.A. Shweder and R.A. LeVine (eds.)
Culture Theory: essays on mind, self and emotion. Cambridge.
(1990) Some propositions about the relations between culture and human
cognition. In J.W. Stigler, R.A. Shweder and G. Herdt (eds.) Cultural
Psychology. Cambridge.
Darwin, C. (1964) [1859] On the Origin of the Species. Cambridge, Mass.
Dasser, V. (1988) Mapping social concepts in monkeys. In R.W. Byrne and A.
Whiten (eds.), Machiavellian Intelligence: social expertise and the evolution of
282 Consolidated bibliography

intellect in monkeys, apes and humans. Oxford,

de Waal, F. (1982) Chimpanzee Politics. London.
Dennett, D.C. (1983) Intentional systems in cognitive ethology: the Panglossian
paradigm defended. Behavioral and Brain Sciences 6: 343-90.
(1987) The Intentional Stance. Cambridge, Mass.
(1988) The intentional stance in theory and practice. In R.W. Byrne and A.
Whiten (eds.) Machiavellian Intelligence: social expertise and the evolution of
intellect in monkeys, apes and humans. Oxford.
Devisch, R. (1985) Perspectives on divination in contemporary sub-Saharan
Africa. In W. van Binsbergen and M. Schoffeleers (eds.) Theoretical Explo-
rations in African Religion. London.
Dhalla, M.N. (1922) Zoroastrian Civilization: from the earliest times to the
downfall of the last Zoroastrian Empire 651 AD. New York.
Diaz, R.M. (1986) The union of thought and language in children's private
speech. Quarterly Newsletter of the Laboratory of Human Cognition 8:
90-7.
Donald, M. (1991) Origins of the Modern Mind. Cambridge, Mass.
Donin, H.H. (1980) To Pray as a Jew. New York.
Dorner, D. (1987) On the Logic of Failure: thinking, planning and decision making
in uncertainty and complexity. Working paper no. 54, Lehrstuhl Psychologie
II, University of Bemburg.
(1990) The logic of failure. Philosophical Transactions of the Royal Society
B327: 463-73. (Originally given as address to the Royal Society meeting:
Human factors in high-risk situations, June 1989.)
Drew, P. (1984) Speakers' reportings in invitation sequences. In J.M. Atkinson
and J. Heritage (eds.) Structures of Social Actions: studies in conversation
analysis. Cambridge.
(1987) Po-faced receipts of teases. Linguistics 25: 219-53.
(1991) Asymmetries of knowledge in conversational interactions. In I. Mar-
kova and K. Foppa (eds.) Asymmetries of dialogue. Hemel Hempstead.
Drew, P. and Holt, E. (1988) Complainable matters: the use of idiomatic
expressions in making complaints. Social Problems 35: 398-417.
Du Bois, J.W. (1987) Meaning without intention: lessons from divination. IPJA
papers in pragmatics 1(2): 80-112.
Dunbar, R.I.M. (1992a) Why gossip is good for you. New Scientist 136 (1848):
28-31.
(1992b) Neocortex size as a constraint on group size in primates. Journal of
Human Evolution 20: 469-93.
(1993) Coevolution of neocortical size, group size and language in humans.
Behavioral and Brain Sciences 16: 681-735.
Enkvist, N.E. (1984) Text and discourse linguistics, rhetoric, and stylistics. In T.
van Dijk (ed.) Discourse and Literature. Amsterdam.
Ervin-Tripp, S. (1979) Children's verbal turn-taking. In E. Ochs and B.
Schieffelin (eds.) Developmental Pragmatics. New York.
Evans-Pritchard, E.E. (1937) Witchcraft, Oracles and Magic among the Azande.
Oxford.
(1956) Nuer Religion. Oxford.
Fauconnier, G. (1985) Mental Spaces. Cambridge, Mass.
 Consolidated bibliography 283

Feld, S. (1982) Sound and Sentiment: birds, weeping, poetics and song in Kaluli
expression. Philadelphia.
Feldman, C , Bruner, J., Renderer, B. and Spitzer, S. (1990) Narrative compre-
hension. In B. Britton and A. Pellegrini (eds.) Narrative Thought and
Narrative Language. Hillsdale, N J .
Ferguson, C.A. (1981) The structure and use of politeness formulas. In
F. Coulmas (ed.) Conversational Routine: explorations in standardized
communication situations and prepatterned speech. The Hague, Paris and
New York.
Fernandez, J. (1974) The mission of metaphor in expressive culture. Current
Anthropology 15: 119^45.
Field, J. (1955) [1934] A Life of One's Own. London.
Finnegan, R. (1969) How to do things with words: performative utterances
among the Limba of Sierra Leone. Man (n.s.) 4.
Flavell, J.H. and Ross, L. (eds.) (1981) Social Cognitive Development. Cambridge.
Fodor, J. (1983) Modularity of Mind. Cambridge, Mass.
Fortes, M.F. (1983) Rules and the Emergence of Society. Royal Anthropological
Institute, occasional paper no. 39. London.
(1987) Ritual and office. In J.R. Goody (ed.) Religion, Morality and the Person.
Cambridge.
Fouts, R.S., Fouts, D.H. and Van Cantfort, T.E. (1989) The infant Loulis learns
signs from cross fostered chimpanzees. In R.A. Gardner, B.T. Gardner and
T.E. Van Cantfort (eds.) Teaching Sign Language to Chimpanzees. New
York.
Frankel, R. (1983) From sentence to sequence: understanding the medical
encounter through microinteractional analysis. Discourse Processes 7:135-70.
Gallup, G.G. (1970) Chimpanzees: self-recognition. Science 167: 86-7.
Gardner, H. (1983) Frames of Mind. New York.
Gardner, P. (1966) Symmetric response and memorate knowledge: the structure
and ecology of individualistic culture. Southwestern Journal of Anthropology
22: 389-415.
(1991) Foragers' pursuit of individual autonomy. Current Anthropology 32:
543-73.
Garfinkel, H. (1967) Studies in Ethnomethodology. Englewood Cliffs, N.J.
(1984) [1967] Commonsense knowledge of social structures: the documentary
method of interpretation in lay and professional fact finding. In H. Garfinkel
Studies in Ethnomethodology. Cambridge.
Garfinkel, H., Lynch, M. and Livingston, E. (1981) The work of a discovering
science construed with materials from the optically discovered pulsar.
Philosophy of the Social Sciences 11: 131-58.
Geiss, M. and Zwicky, A. (1971) On invited inferences. Linguistic Inquiry 2:
561-5.
Gelb, N. (1986) The Berlin Wall. New York.
Gellner, E. (1975) Ethnomethodology: the re-enchantment industry or the
Calif.n way of subjectivity. Philosophy of the Social Sciences 5: 431-50.
Gelman, R. and Spelke, E. (1981) The development of thought about animate and
inanimate objects: implications for research on social cognition. In J.H.
Flavell, and L. Ross (eds.) Social Cognitive Development. Cambridge.
284 Consolidated bibliography

Gergen, M. and Gergen, K. (1984) The social construction of narrative accounts.

In K. Gergen and M. Gergen (eds.) Historical Social Psychology. Hillsdale.
Gibson, T. (1986) Sacrifice and Sharing in the Philippine Highlands: religion and
society among the Buid of Mindoro. London School of Economics mono-
graphs on social anthropology, no. 57. London.
Gill, S.D. (1981) Sacred Words: a study of Navajo religion and prayer. Westport,
Conn.
Ginsberg, M.L. (1987) Readings in Non-monotonic Logic. Los Altos, Calif.
Glickman, S.E. and Scroges, R.W. (1966) Curiosity in zoo animals. Behaviour 26:
151-8.
Godelier, M. (1988) The Mental and the Material. London.
Goffman, E. (1959) The Presentation of Self in Everyday Life. New York.
(1974) Frame analysis. Cambridge, Mass.
(1978) Response cries. Language 54: 787-815.
(1981) Forms of Talk. Philadelphia.
(1983) The interaction order. American Sociological Association 1982 Presi-
dential Address. American Sociological Review 48: 1-17.
Good, D.A. (1989) The viability of conversational grammars. In M. Taylor, F.
Neel and D. Bouwhuis (eds.) Multi-modal aspects of Dialogue. Berlin.
Goodall, J. (1986) The Chimpanzees of Gombe: patterns of behavior. Cambridge,
Mass.
Goodenough, W.H. (1990) Evolution of the human capacity for beliefs. American
Anthropologist 92: 597-612.
Goodwin, C. (1979) The interactional construction of a sentence. In G. Psathas
(ed.) Everyday Language. New York.
(1981) Conversational Organization: interaction between speakers and hearers.
New York.
(1986) Gesture as a resource for the organization of mutual orientation.
Semiotica 62(1-2): 29-49.
(1993) Perception, technology and interaction on a scientific research vessel.
Unpublished manuscript, University of South Carolina.
Goodwin, C. and Goodwin, M.H. (1986) Gesture and coparticipation in the
activity of searching for a word. Semiotica 62(1-2): 51-75.
Goodwin, M.H. (1981) Searching for a word as an interactive activity. In J. Deely
and M.D. Lenkart (eds.) Semiotics. New York.
Goody, E. (1972) Greetings, begging and the presentation of respect. In J.S.
LaFontaine (ed.) The Interpretation of Ritual. London.
(1978a) Towards a theory of questions. In E. Goody (ed.) Questions and
Politeness: strategies in social interaction. Cambridge.
(1978b) Introduction. In E. Goody (ed.) Questions and Politeness: strategies in
social interaction. Cambridge.
(1982) Daboya weavers: relation of production, dependence and reciprocity. In
E. Goody (ed.) From Craft to Industry. Cambridge.
(1987) Why must might be right? Observations on sexual Herrschaft. Quarterly
Newsletter of the Laboratory of Comparative Human Cognition 9: 54-76.
(1989) Learning, apprenticeship and the division of labour. In M. Coy (ed.)
Apprenticeship: from theory to method and back again. Albany, N.Y.
 Consolidated bibliography 285

(1993) Informal learning of adult roles in Baale. In M. Fielous, J. Lombard and

J-M. Kambou-Ferrand (eds.) Images d'Afrique et Sciences Sociales: les pays
lobi, birifor et dagara. Paris.
(n.d.) Working papers in implications of a social origin of human intelligence.
Berlin. Unpublished manuscript.
Goody, J.R. (1962) Death, Property and the Ancestors. Stanford, Calif.
(1977) The Domestication of the Savage Mind. Cambridge.
(1987) The Interface between the Written and the Oral. Cambridge.
Greenfield, P. (1991) Language, tools and brain: the ontogeny and phylogeny of
hierarchically organized sequential behavior. Behavioral and Brain Sciences
14: 531-95.
Grice, H.P. (1957) Meaning. Philosophical Review 66: 377-88.
(1989) Ways with Words. Cambridge, Mass.
Gumperz, J.J. and Hymes, D. (eds.) (1964) The Ethnography of Communication.
Special publication of American Anthropologist, vol. 66.
Haack, S. (1978) Philosophy of Logics. Cambridge.
Haiman, J. (1990) Sarcasm as theater. Cognitive Linguistics 1-2: 181-205.
Haimoff, E. (n.d.) Video analysis of Siamang call bouts. Unpublished manu-
script, University of California, Los Angeles.
Hall, E.T. (1959) The Silent Language. Greenwich, Conn.
Harcourt, A.H. (1988) Alliances in contests and social intelligence. In R.W.
Byrne and A. Whiten (eds.) Machiavellian Intelligence: social expertise and
the evolution of intellect in monkeys, apes and humans. Oxford.
Harcourt, A.H. and de Waal, F. (eds.) (1992) Coalitions and Alliances among
Humans and other Animals. Oxford.
Hargreaves, R. (1981) Little Miss Bossy. Bromley, Kent.
Harman, G. (1974) review of Meaning by S. Schiffer. Journal of Philosophy 71(7):
224-9.
Havelock, E.A. (1963) Preface to Plato. Cambridge, Mass.
Haviland, J.B. (1977) Gossip, Reputation and Knowledge in Zinacantan. Chicago.
Hayes, C. (1951) The Ape in our House. New York.
Hayes, K J . and Hayes, C. (1952) Imitation in a home-raised chimpanzee. Journal
of Comparative Physiological Psychology 45: 450-9.
Headland, T.N. and Reid, L. (1989) Hunter-gatherers and their neighbours from
prehistory to the present. Current Anthropology 30: 43-67.
Heath, S.B. (1983) Ways with Words: language, life and work in communities and
classrooms. Cambridge.
Henn-Schmolders, C. (1975) Ars Conversationis. Zur Geschichte des sprachli-
chen Umgangs. Arcadia 10: 16-33.
Heritage, J. (1984) Garfinkel and Ethnomethodology. Cambridge.
(1990/91) Intention, meaning and strategy: observations on constraints in
interactional analysis. Research in Language and Social Interaction 24:
311-22.
Herskovits, A. (1986) Language and Spatial Cognition. Cambridge.
Hobson, R.P. (1984) Early childhood autism and the question of egocentrism.
Journal of Autism and Developmental Disorders 14: 85-104.
Hollis, M. and Lukes, S. (eds.) (1982) Rationality and Relativism. Oxford.
286 Consolidated bibliography

Holquist, M. (1990) Dialogism: Bakhtin and his world. London.

Holton, G. (1981) Comments on Professor Harold Garfinkel's paper. Philosophy
of the Social Sciences 11: 159-61.
Horn, L. (1989) A Natural History of Negation. Chicago.
Horton, R. (1967) African traditional thought and Western science. Africa
38: 50-71 and 155-87.
(1982) Tradition and modernity revisited. In M. Hollis and S. Lukes (eds.)
Rationality and Relativism. Oxford.
Howe, M. (1989) Fragments of Genius: the strange feats of idiots savants. London.
Humboldt, W. von (1988) [1836] On Language. Cambridge.
Humphrey, N.K. (1976) The social function of intellect. In P.P.G. Bateson and
R.A. Hinde (eds.) Growing Points in Ethology. Cambridge. Reprinted in
R.W. Byrne and A. Whiten (eds.) Machiavellian Intelligence: social expertise
and the evolution of intellect in monkeys, apes and humans. Oxford, 1988.
(1983) Consciousness Regained. Oxford.
(1986) The Inner Eye. London.
Hutchins, E. (1991) Social organization of distributed cognition. In L. Resnick,
J. Levine and S. Teasley (eds.) Perspectives on Socially Shared Cognition.
Washington, D.C.
(in press) Cognition in the Wild. Cambridge, Mass.
Hutchins, E. and Hazlehurst, B. (1991) Learning in the cultural process. In C.
Langton, C. Taylor, D. Farmer and S. Rasmussen (eds.) Artificial Life, Vol.
II. Redwood City, Calif.
Hutchins, E. and Klausen, T. (1990) Distributed Cognition in an Airline Cockpit.
Department of Cognitive Science, University of California, San Diego.
Ingold, T. (1987) The Appropriation of Nature: essays on human ecology and social
relations. Iowa.
Irvine, J. (1974) Strategies of status manipulation in Wolof greeting. In R.
Bauman and J. Sherzer (eds.) Explorations in the Ethnography of Speaking.
London.
Jackson, M. (1978) An approach to Kuranko divination. Human Relations 31:
117-38.
Jammer, M. (1974) The Philosophy of Quantum Mechanics. New York.
Jauss, H.-R. (1972) Theorie der Gattungen und Literatur des Mittelalters. In
H.-R. Jauss and E. Kohler (eds.) Grundriss der romanischen Literatur en des
Mittelalters, Vol. I: Generalities. Heidelberg.
Jefferson, G. (1975) Error correction as an international resource. Language in
Society 4: 181-99.
(1978) Sequential aspects of storytelling in conversation. In J. Schenkein (ed.)
Studies in the Organization of Conversational Interaction. New York.
(1984) On step wise transition from talk about a trouble to inappropriately next-
positioned matters. In J.M. Atkinson and J. Heritage (eds.) Structures of
Social Action. Cambridge.
(1988a) On the sequential organization of troubles-talk in ordinary conver-
sation. Social Problems 35: 418-41.
(1988b) Remarks on 'non-correction' in conversation. Unpublished paper:
lecture given at Helsingen Yliopisto, Suomen Kielen Laitos, Helsinki (to
appear in Text).
 Consolidated bibliography 287

(1992) List-construction as a task and resource. In G. Psathas (ed.) Interactio-

nal Competence. Hillsdale, N.J.
Jerison, H J . (1973) Evolution of the Brain and Intelligence. New York.
Johnson-Laird, P.N. (1983) Mental Models: towards a cognitive science of lan-
guage, inference and consciousness. Cambridge.
Johnson-Laird, P.N. and Wason, P.C. (eds.) (1977a) Thinking: readings in
cognitive science. Cambridge.
(1977b) A theoretical analysis of insight into a reasoning task, and Postscript. In
P.N. Johnson-Laird and P.C. Wason (eds.) Thinking: readings in cognitive
science. Cambridge.
Jolles, A. (1972) Einfache Formen: Legende, Sage, My the, Rdtsel, Spruch, Kasus,
Memorabile, Mdrchen, Witz. Tubingen.
Jolly, A. (1966a) Lemur social behaviour and primate intelligence. Science 153:
501-6. Reprinted in R.W. Byrne and A. Whiten (eds.) Machiavellian
Intelligence: social expertise and the evolution of intellect in monkeys, apes and
humans. Oxford.
(1966b) Lemur Behaviour. Chicago.
Judge, P.G. (1982) Redirection of aggression based on kinship in a captive group
of pigtail macaques. International Journal of Primatology 3: 301.
Kahneman, D., Slovic, P. and Tversky, A. (1982) Judgement under Uncertainty:
heuristics and biases. Cambridge.
Kant, I. (1934) Religion Within the Limits of Reason Alone. New York.
Katz, R. (1982) Boiling Energy: community healing among the Kalahari IKung.
Cambridge, Mass.
Keenan, E.O. (1977) Why look at unplanned and planned discourse? In E.O.
Keenan and T.L. Bernett (eds.) Discourse across Time and Space. Los
Angeles.
Keesing, R.M. (1982) Kwaio Religion: the living and the dead in a Solomon Island
society. New York.
Kendon, A. (1976) Some functions of the face in a kissing round. Semiotica 17:
299-334.
Kenny, A. (1966) Practical inference. Analysis 26(3): 65-75.
Kolb, B. and Whishaw, I. (1990) Fundamentals of Human Neuropsychology. New
York.
Kummer, H. (1967) Tripartite relations in hamadryas baboons. In S.A. Altman
(ed.) Social Communication among Primates. Chicago.
Lafon, J.-C. (1968) Auditory basis of phonetics. In B. Malmberg (ed.) Manual of
Phonetics. Amsterdam.
LaFontaine, J.S. (1978) Introduction. In J.S. LaFontaine (ed.) Sex and Age as
Principles of Social Differentiation. London.
Lakoff, G. (1987) Women, Fire and Dangerous Things. Chicago.
Lakoff, G. and Johnson, M. (1980) Metaphors We Live By. Chicago.
Landau, T. (1989) About Faces: the evolution of the human face. New York.
Langacker, R. W. (1986) An introduction to cognitive grammar. Cognitive Science
10: 1-40.
(1987) Foundations of Cognitive Grammar, Vol. I: Theoretical Prerequisites.
Stanford, Calif.
Lave, J., Murtaugh, M. and de la Rocha, O. (1984) The dialectic of arithmetic in
288 Consolidated bibliography

grocery shopping. In B. Rogoff and J. Lave (eds.) Everyday Cognition: its

development in social context. Cambridge, Mass.
Laver, J.D.M.H. (1981) Linguistic routines and politeness in greeting and
parting. In F. Coulmas (ed.) Conversational Routine. The Hague, Paris and
New York.
Lee, R.B. (1979) The IKung San: men, women and work in a foraging society. New
York.
(1984) The Dobe IKung. New York.
Lee, R.B. and DeVore, I. (eds.) (1968) Man the Hunter. Chicago.
Lerner, G.H. (1987) Collaborative turn sequences: sentence construction and
social action. Ph.D. dissertation. University of California, Irvine.
Leslie, A.M. (1987) Pretence and representation in infancy: the origins of'theory
of mind'. Psychological Review 94: 412-26.
Levine, L.W. (1977) Black Culture and Black Consciousness: Afro-American folk
thought from slavery to freedom. New York.
Levinson, S. (1983) Pragmatics. Cambridge.
(1985) What's special about conversational inference? Paper presented to the
Annual Conference of the British Psychological Society, Swansea, April.
(1987a) Minimization and conversational inference. In M. Pappi and J.
Verscheuren (eds.) The Pragmatic Perspective. Amsterdam.
(1987b) Pragmatics and the grammar of anaphora. Journal of Linguistics 23:
379^34.
(1988) Putting linguistics on a proper footing: explorations in Goffman's
participation framework. In P. Drew and A. Wootton (eds.) Erving Goffman:
exploring the interaction order. Cambridge.
(1989) Review of Sperber and Wilson's Relevance. Journal of Linguistics 25:
455-72.
(1991) Pragmatic reduction of the binding conditions revisited. Journal of
Linguistics 21 \ 107-61.
(in press) Three levels of meaning. In F. Palmer (ed.) Grammar and Meaning.
Cambridge.
(forthcoming) Generalized Conversational Implicature. Cambridge.
Levi-Strauss, C. (1969) The Elementary Structure of Kinship. London. French
edition, 1949.
Lewis, D. (1969) Convention. Cambridge, Mass.
Lewis, I.M. (1971) Ecstatic Religion. Harmondsworth.
Liberman, R. (1985) Understanding Interaction in Central Australia. London.
Lieberman, P. (1968) Primate vocalizations and human linguistic ability. Journal
of the Acoustical Society of America 44: 1157-64.
(1991) Uniquely Human: the evolution of speech, thought and selfless behavior.
Cambridge, Mass.
Lieberman, P. and Blumstein, S. (1988) Speech Physiology, Speech Perception,
and Acoustic Phonetics. Cambridge.
Lienhardt, G. (1961) Divinity and Experience: the religion of the Dinka. Oxford.
Lindblom, B., MacNeilage, P.F. and Studdert-Kennedy, M. (in press) Evolution
of Spoken Language. New York.
Lloyd, G.E.R. (1979) Magic, Reason and Experience. Cambridge.
 Consolidated bibliography 289

Luce, R.D. and Raiffa, H. (1957) Games and Decisions. New York.
Luckmann, T. (1979) Soziologie der Sprache. In R. Konig (ed.) Handbuch der
empirischen Sozialforschung, Vol. XIII. Stuttgart.
(1982) Individual action and social knowledge. In M. von Cranach and R.
Harre (eds.) The Analysis of Action: recent theoretical and empirical advances.
Cambridge.
(1983) Elements of a social-theory of communication. In T. Luckmann Life-
world and Social Realities. London.
(1986) Grundformen der gesellschaftlichen Vermittlung des Wissens: kommu-
nikative Gattungen. In F. Neidhardt, M. Lepsius and J. Weiss (eds.) Kultur
und Gesellschaft, Sonderheft 27 der Kolner Zeitschrift fur Soziologie and
Sozialpsychologie. (French: Les formes elementaires de la transmission
sociale du savoir: genres comunicatifs. To be published in Societes.)
(1987) Kanon und Konversion. In A. and J. Assman (eds.) Kanon und Zensur.
Munich.
(1989) Prolegomena to a social theory of communicative genres. Slovene Studies
11(1-2), Toussaint Hocevar Memorial Issue: 159-66.
(1992) On the communicative adjustment of perspectives, dialogue, and
communicative genres. In A. Heen Wold (ed.) The Biological Alternative:
towards a theory of language and mind. London.
Luckmann, T. and Keppler, A. (1989) Weissheits-vermittlung im Alltag. Wer in
den Augen eines anderen weise ist, ist weise. In W. Oelmuller (ed.)
Philosophie und Weisheit. Paderborn.
Luria, A.R. (1979) The Making of Mind, edited by M. Cole and S. Cole,
Cambridge, Mass.
(1981) Language and Cognition, edited by J.V. Wertsch. Washington, D.C.
Lynch, M. and Woolgar, S. (eds.) (1988) Representation in Scientific Practice.
Cambridge, Mass.
Lyons, J. (1988) Origins of language. In A.C. Fabian (ed.) Origins. Cambridge.
Machiavelli, N. (1966) The Prince, translated by D. Donno. London.
MacNeilage, P.F. (1987) The evolution of hemispheric specialization for manual
function and language. In S. Wyse (ed.) Higher Brain Functions: recent
explorations of the brain's emergent properties. New York.
MacNeilage, P.F., Studdert-Kennedy, M. and Lindblom, B. (1987) Primate
handedness reconsidered. Behavioral and Brain Sciences 10: 247-63.
Macphail, E.M. (1985) Vertebrate intelligence: the null hypothesis. In L.
Weiskrantz (ed.) Animal Intelligence. Oxford.
Malinowski, B. (1927) The problem of meaning in primitive languages. In C.K.
Ogden and LA. Richards (eds.) The Meaning of Meaning. London.
Mandelbaum, J. and Pomerantz, A. (1991) What drives social action? In K. Tracy
(ed.) Understanding Face-to-Face Interaction: issues linking goals and dis-
course. Hillsdale, N J.
Marett, R.R. (1909) The Threshold of Religion. London.
Marshall, L. (1957) The kin terminology system of the !Kung Bushmen. Africa
27(1): 1-25.
(1976) The IKung of Nyae Nyae. Cambridge, Mass.
Marslen-Wilson, W., Levy, E. and Tyler, L. (1982) Producing interpretable
290 Consolidated bibliography

discourse: the establishment and maintenance of reference. In R. Jarvella and

W. Klein (eds.) Speech, Place and Action. London.
Martin, C.E. (1990) Popular speech and social order in Northern Mexico, 1650-
1830. Comparative Studies in Society and History 32: 305-24.
Martin, R.D. (1990) Primate Origins and Evolution. London.
Maupoil, B. (1943) La geomancie a l'ancienne Cote des Esclaves. Trav. et Mem. de
Vlnst. d'Eth. 42.
McHugh, P. (1968) Defining the Situation: the organisation of meaning in social
interaction. New York.
McNeill, D. (1985) So you think gestures are nonverbal? Psychological Review 92:
350-71.
Mead, G.H. (1967) [1934] Mind, Self and Society. Chicago.
Mervis, C. and Rosch, E. (1981) Categorization of natural objects. Annual Review
of Psychology 32: 89-115.
Michaels, S. (1986) Narrative presentations: an oral preparation for literacy with
first graders. In J. Cook-Gumperz (ed.) The Social Construction of Literacy.
Cambridge.
Miller, G.A. (1956) The magical number seven, plus or minus two. Psychological
Review 63: 81-97.
Miller, G.A., Galanter, E. and Pribram, K.H. (1960) Plans and the Structure of
Behavior. New York.
Miller, G.A. and Johnson-Laird, P.N. (1976) Language and Perception.
Cambridge.
Miller, R.A. (1967) The Japanese Language. Chicago.
Milton, K. (1988) Foraging behaviour and the evolution of primate intelli-
gence. In R.W. Byrne and A. Whiten (eds.) Machiavellian Intelligence:
social expertise and the evolution of intellect in monkeys, apes and humans.
Oxford.
Morris, D. (1979) Gestures: their origins and distribution. London.
Moulton, J.H. (1913) Early Zoroastrianism. (Hibbert lectures at Oxford and
London 1912.) London.
Myers, F. (1982) Always ask: resource use and land ownership among Pintupi
aborigines of the Australian Western Desert. In N.M. William and E.S.
Hunn (eds.) Resource Managers: North American and Australian hunter-
gatherers. Canberra.
Nadel, S.F. (1957) The Theory of Social Structure. New York.
Nelson, K. (1981) Social cognition in a script framework. In J.H. Flavell and L.
Ross (eds.) Social Cognitive Development. Cambridge.
Nies, F. (1978) Genres mineurs. Texte zur Theorie und Geschichte nichtkanonischer
Literatur (vom 16 Jahrhundert bis zur Gegenwart). Munich.
Nishida, T. (1983) Alpha status and agonistic alliance in wild chimpanzees.
Primates 24: 318-36.
Oldman, D. (1974) Chance and skill: a study of roulette. Sociology 8: 596-603.
Oxnard, C.E. (1981) The place of man among the primates: anatomical, molecular
and morphometric evidence. Homo 32: 149-76.
Park, G.K. (1963) Divination and its social context. Journal of the Royal
Anthropological Institute 93: 195-209.
 Consolidated bibliography 291

Parkin, D. (1979) Straightening the paths from wilderness: the case of divinatory
speech. Journal of the Anthropology Society of Oxford 10(3): 147-60.
Passingham, R.E. (1981) Primate specializations in brain and intelligence.
Symposia of the Zoological Society of London 46: 361-88.
(1982) The Human Primate. New York.
Passingham, R.E. and Ettlinger, G. (1974) A comparison of cortical function in
man and other primates. International Review of Neurobiology 16: 233-99.
Patterson, F.G. and Cohn, R. (in press) Self recognition and self awareness in
lowland gorillas. In S. Parker, M. Boccia and R. Mitchell (eds.) Comparative
Reflections on Self Awareness in Animals and Humans. Cambridge.
Patterson, F.G. and Linden, E. (1981) The Education of Koko. New York.
Perrault, C.R. (1987) An Application of Default Logic to Speech Act Theory.
Stanford CSLI report.
Piaget, J. (1929) The Child's Conception of the World. London.
Pollack, M. (1986a) A Model of Plan Inference that Distinguishes between the Beliefs
of Actors and Observers. Technical note 387, AI center, Computer Science
Division, SRI International.
(1986b) Inferring Domain Plans in Question-Answering. Technical note 403, AI
center, Computer Science Division, SRI International.
Pomerantz, A. (1978) Compliment responses: notes on the cooperation of
multiple constrains. In J. Schenkein (ed.) Studies in the Organization of
Conversational Interaction. New York.
(1980) Telling my side: 'limited access' as a 'fishing' device. Sociological Inquiry
50:186-98.
(1984) Agreeing and disagreeing with assessments: some features of preferred/
dispreferred turn shapes. In J.M. Atkinson and J. Heritage (eds.) Structures
of Social Actions: studies in conversational analysis. Cambridge.
Premack, D. (1986) Gavagai! or the Future History of the Animal Language
Controversy. Cambridge, Mass.
(1988) 'Does the chimpanzee have a theory of mind?' revisited. In R.W. Byrne
and A. Whiten (eds.) Machiavellian Intelligence: social expertise and the
evolution of intellect in monkeys, apes and humans. Oxford.
Quinn, N. and Holland, D. (1987) Culture and cognition. In D. Holland and N.
Quinn (eds.) Cultural Models in Language and Thought. Cambridge.
Ransom, T.W. (1981) Beach Troop of the Gombe. East Brunswick, N.J.
Rappaport, R.A. (1988) Logos, liturgy and the evolution of humanity. Confer-
ence paper, Symposium, Zagreb, XII International Conference of Anthro-
pological and Ethnological Sciences.
Rattray, R.S. (1927) Religion and Art in Ashanti. London.
Reason, J. (1987) The Chernobyl Errors. Bulletin of the British Psychology Society
40: 201-6.
Rehfisch, F. (1960) The dynamics of multilinearity on the Mambila Plateau.
Africa 3: 246-61.
(1972) The Social Structure of a Mambila Village. Ahmadu Bello University
Sociology Department, occasional paper 2. Zaria.
Reisman, P. (1977) Freedom in Fulani Social Life: an introspective ethnography.
Chicago.
292 Consolidated bibliography

(1992) First Find your Child a Good Mother: the construction of self in two
African communities. New Brunswick, N.J.
Richards, A.I. (1956) Chisungu: a girls' initiation ceremony among the Bemba of
Northern Rhodesia. London.
Rogoff, B. (1990) Apprenticeship in Thinking: cognitive development in social
context. Oxford.
Rommetveit, R. (1990) On axiomatic features of a dialogical approach to language
and mind. In I. Markova and K. Foppa (eds.) The Dynamics of Dialogue.
Hemel Hempstead.
(1992) Outlines of a dialogically based social-cognitive approach to human
cognition and communication. In A. Heen Wold (ed.) The Dialogical
Alternative: towards a theory of language and mind. London.
Rosaldo, M. (1980) Knowledge and Passion: Ilongot notions of self and social life.
Cambridge.
(1984) Toward an anthropology of self and feeling. In R. Schweder and R.
LeVine (eds.) Culture Theory: essays on mind, self and emotion. Cambridge.
Rosaldo, R. (1980) Ilongot Headhunting, 1883—1974: a study in society and history.
Stanford, Calif.
Rosch, E. (1978) Principles of categorization. In E. Rosch and B. Floyd (eds.)
Cognition and Categorization. Hillsdale, N J .
Rumelhart, D., Hinton, G. and Williams, R. (1986) Learning internal represen-
tations by error propagation. In D. Rumelhart, J. McClelland and the PDP
Group (eds.) Parallel Distributed Processing, Vol. I: Foundations. Cambridge,
Mass.
Rumelhart, D., McClelland, J. and the PDP Group (1986) Parallel Distributed
Processing, Vol. I: Foundations. Cambridge, Mass.
Sacks, H. (1986) [1973] On the preferences for agreement and contiguity in
sequences in conversation. In G. Button and J.R.E. Lee (eds.) Talk and
Social Organization. London.
(1989) Lectures 1964-1965. Human Studies 12, special issue, edited by G.
Jefferson.
(1992) Lectures on Conversation. 2 vols. Edited by G. Jefferson. Oxford.
Sacks, H. and Schegloff, E.A. (1979) Two preferences in the organization of
reference to persons in conversation and their interaction. In G. Psathas (ed.)
Everyday language: studies in ethnomethodology. New York.
Sacks, H., Schegloff, E.A., and Jefferson, G. (1974) A simplest systematics for the
organization of turn-taking for conversation. Language 50: 696-735.
Sacks, O. (1985) The Man who Mistook his Wife for a Hat. London.
Samarin, W J . (1976) Language in Religious Practice. Georgetown, Washington,
D.C.
Sanders, R.E. (1988) Review of Sperber and Wilson's Relevance. Language in
Society 17: 604-9.
Sarich, V.M. and Wilson, A.C. (1967) Immunological time scale for hominid
evolution. Science 158: 1200-3.
Savage-Rumbaugh, E.S. (1986) Ape Language: from conditioned response to
symbol. New York. Accompanying film.
Savage-Rumbaugh, E.S. and McDonald, K. (1988) Deception and social mani-
 Consolidated bibliography 293

pulation in symbol using apes. In R.W. Byrne and A. Whiten (eds.)

Machiavellian Intelligence: social expertise and the evolution of intellect in
monkeys, apes and humans. Oxford.
Schank, R. and Abelson, R. (1977) Scripts, Plans, Goals and Understanding: an
enquiry into human knowledge structures. Hillsdale, N.J.
Scheflen, A. (1972) Body Language and Social Order. Englewood Cliffs, N.J.
Schegloff, E.A. (1979) The relevance of repair to syntax-to-conversation. In T.
Givon (ed.) Syntax and Semantics 12: discourse and syntax. New York.
(1980) Preliminaries to preliminaries: 'Can I ask you a question?'. Sociological
Inquiry 50(3-4): 104-51.
(1984a) On some questions and ambiguities in conversation. In J.M. Atkinson
and J. Heritage (eds.) Structures of Social Action. Cambridge.
(1984b) On some gestures' relation to talk. In J.M. Atkinson and J. Heritage
(eds.) Structures of Social Action. Cambridge.
(1986) The routine as achievement. Human Studies 9: 111-52.
(1988a) On an actual virtual servo-mechanism for guessing bad news: a single
case conjecture. Social Problems 35: 442-57.
(1988b) Presequences and indirection. Journal of Pragmatics 12: 55-62.
(1989) Harvey Sacks lectures: an introduction/memoir. In G. Jefferson (ed.)
Harvey Sacks Lectures 1964—1965, special issue of Human Studies 12.
(1992a) Repair after next turn: the last structurally provided defense of
inter subjectivity in conversation. American Journal of Sociology 97:
1295-345.
(1992b) Introduction. In G. Jefferson (ed.) Lectures on Conversation, Vol. I.
Oxford.
(1992c) Reflections on language, development, and the interactional character
of talk-in-interaction. In M. Bornstein and J.S. Bruner (eds.) Interaction in
Human Development. Hillsdale, N.J.
Schegloff, E.A., Jefferson, G. and Sacks, H. (1977) The preference for self-
correction in the organization of repair in conversation. Language 53:361-82.
Schegloff, E. and Sacks, H. (1973) Opening up closings. Semiotica 7: 289-327.
Schelling, T. (1960) The Strategy of Conflict. Cambridge, Mass.
(1988) The mind as a consuming organ. In D. Bell, H. Raiffa and A. Tversky
(eds.) Decision Making. Cambridge.
Scher, N.J.M. (ed.) (1962) Theories of Mind. New York.
Schieffelin, B. (1990) The Give and Take of Everyday Life: language socialization
of Kaluli children. Cambridge.
Schieffelin, E. (1976) The Sorrow of the Lonely and the Burning of the Dancers.
New York.
Schiffer, S.R. (1972) Meaning. Oxford.
Schone, H. (1984) Spatial Orientation. Princeton.
Schrire, C. (1984) Past and Present in Hunter-Gatherer Studies. London.
Schutz, A. (1932) Der sinnhafte Aufbau der sozialen Welt. Vienna.
(1962) Choosing among projects of action. In A. Schutz Collected Papers, Vol.
I: The Problem of Social Reality. The Hague.
(1970) On Phenomenology and Social Relations. Selected writings edited by
H.R. Wagner. Chicago.
294 Consolidated bibliography

Schutz, A. and Luckman, T. (1973) The Structures of the Life-World, Vol. I.

Evans ton, 111.
(1989) The Structures of the Life-World, Vol. II. Evanston, 111.
Searle, J.R. (1969) Speech Acts. Cambridge.
Service, E.R. (1962) Primitive Social Organization: an evolutionary perspective.
New York.
Seyfarth, R.M. and Cheney, D.L. (1988) Do monkeys understand their relations?
In R.W. Byrne and A. Whiten (eds.) Machiavellian Intelligence: social
expertise and the evolution of intellect in monkeys, apes and humans. Oxford.
Shaha, S. (1983) Sri 108 Siddhasdgar MuniMaharaj. Kolhapur.
Sharpe, J.C. (1980) Defamation and Sexual Slander in Early Modern England.
York.
Shaw, R. (1985) Gender and structuring of reality in Temne divination: an
interactive study. Africa 53: 286-303.
Sibley, C.G. and Ahlquist, J.E. (1984) The phylogeny of the hominoid primates,
as indicated by DNA-DNA hybridization. Journal of Molecular Evolution
20: 2-15.
Simon, H. (1981) The Sciences of the Artificial. Cambridge, Mass.
Smith, P.K. (1988) The cognitive demands of children's social interactions with
peers. In R.W. Byrne and A. Whiten (eds.) Machiavellian Intelligence: social
expertise and the evolution of intellect in monkeys, apes and humans. Oxford.
Smith, WJ. (1977) The Behavior of Communicating: an ethological approach.
Cambridge, Mass.
Smuts, B. (1985) Sex and Friendship in Baboons. Chicago.
Solway, J. and Lee, R.B. (1990) Foragers, genuine or spurious? Situating the
Kalahari San in history. Current Anthropology 31: 109-46.
Sperber, D. (1975) Rethinking Symbolism. Cambridge.
(1985) The epidemiology of beliefs. In C. Fraser and G. Gaskell (eds.)
Psychological Studies of Widespread Beliefs.
(1990) The evolution of the language faculty: a paradox and its solution.
Behavioral and Brain Sciences 13: 756-8.
Sperber, D. and Wilson, D. (1981) Irony and the use-mention distinction. In P.
Cole (ed.) Radical Pragmatics. New York.
(1986) Relevance: communication and cognition. Oxford.
Stern, D.N. (1977) The First Relationship: infant and mother. London.
Steward, J.H. (1936) The economic and social basis of primitive bands. In R.H.
Lowie (ed.) Essays in Anthropology Presented to A.L. Kroeber. Berkeley,
Calif.
Strathern, M. (1988) The Gender of the Gift: problems with women and problems
with society in Melanesia. Berkeley, Calif.
Strecker, L. (1988) The Social Practice of Symbolisation: an anthropological
analysis. London School of Economics monograph on social anthropology,
no. 60. London.
Streeck, J. (1988) The significance of gesture: how is it established? Papers in
Pragmatics 2: 1-2.
(1989a) The social organization of self-adaptor sequences in conversation.
 Consolidated bibliography 295

Paper presented at the annual meeting of the American Anthropological

Association, Washington, D.C.
(1989b) On self-initiated repair in Ilokano conversation. Paper presented at the
Annual Meeting of the Speech Communication Association, San Francisco.
(1992) The dispreferred other. In J. Searle et al. (On) Searle on Conversation.
Amsterdam.
(1993) Gesture as communication I: its coordination with gaze and speech.
Communication Monographs 60: 275-99.
(1994) Gesture as communication II: the audience as co-author. Research on
Language and Social Interaction (Fall) Special issue Is Gesture Communica-
tive? A. Kendon, ed.
Streeck, J. and Hartge, U. (1992) Previews: gestures at the transition place. In P.
Auer and A. di Luzio (eds.) The Contextualization of Language. Amsterdam.
Stross, B. (1974) Speaking of speaking: Tzeltal meta-linguistics. In R. Bauman
and J. Sherzer (eds.) Explorations in the Ethnography of Speaking.
Cambridge.
Strum, S.C. (1983) Use of females by male olive baboons. American Journal of
Primatology 5: 53-109.
Surgy, A. de (1981) La Geomancie et le culte d'Afa chez les Evhe du Littoral. Paris.
Talmy, L. (1985) Lexicalization patterns: semantic structure in lexical forms. In
T. Shopen (ed.) Language Typology and Syntactic Description, Vol. III.
Cambridge.
Tambiah, SJ. (1970) Buddhism and the Spirit Cults in Northeast Thailand.
Cambridge.
(1985) Culture, Thought and Social Action. Cambridge, Mass.
Taylor, W.B. (1979) Drinking, Homicide and Rebellion in Colonial Mexican
Villages. Stanford, Calif.
Terasaki, A. (1976) Pre-Announcement Sequences in Conversation. Social Sciences
working paper, no. 99. University of California, Irvine.
Thomas, W.I. (1960) On Social Organization and Social Personality. Selected
Papers, edited by M. Janowitz. Chicago.
Thompson, J.B. (1984) Studies in the Theory of Ideology. Cambridge.
Tomasello, M., Davis-Dasilva, M., Camak, L. and Bard, K. (1987) Observational
learning of tool-use by young chimpanzees. Journal of Human Evolution 2:
175-83.
Tomasello, M., Kruger, A.C. and Ratner, H.H. (1993) Cultural learning.
Behavioral and Brain Sciences 16: 495-552.
Tracy, K. (ed.) (1991) Understanding Face-to-Face Interaction: issues linking goals
and discourse. Hillsdale, N.J.
Trevarthen, C.B. (1979a) Communication and cooperation in early infancy: a
description of primary intersubjectivity. In M. Bullowa (ed.) Before Speech:
the beginning of interpersonal communication. Cambridge.
(1979b) Instincts for human understanding and for cultural cooperation: their
development in infancy. In M. von Cranach, K. Foppa, W. Lepenies and D.
Ploog (eds.) Human Ethology: claims and limits of a new discipline.
Cambridge.
296 Consolidated bibliography

(1988) Universal co-operative motives: how infants begin to know the language
and culture of their parents. In G. Jahoda and I.M. Lewis (eds.) Acquiring
Culture: cross cultural studies in child development. London.
Trexler, R.C. (1987) The Christian at Prayer: an illustrated prayer manual
attributed to Peter the Chanter (d. 1197). Birmingham, N.Y.
Turnbull, C. (1968) The importance of flux in two hunting societies. In R.B. Lee
and I. Devore (eds.) Man the Hunter. Chicago.
Tver sky, A. and Kahneman, D. (1977) Judgement and uncertainty: heuristics and
biases. In P.N. Johnson-Laird and P.C. Watson (eds.) Thinking: readings in
cognitive science. Cambridge.
Tyler, L.K. (1992) Spoken Language Comprehension. Cambridge, Mass.
Tylor, E.B. (1958) [1871] Religion in Primitive Culture. Part II of Primitive
Culture, chs. 11-19. New York.
Ulmer, B. (1988) Konversionserzahlungen als rekonstruktive Gattung. Erzahler-
ische Mittel und Strategien bei der Rekonstruktion eines Bekehrungserleb-
nisses. Zeitschrift fiir Soziologie 17: 19-33.
Velmans, M. (1991) Is human information processing conscious? Behavioral and
Brain Sciences 14: 651-726.
Vernant, J.P. (1974) Paroles et signes muets. In J.P. Vernant (ed.) Divination et
Rationalite. Paris.
Visalberghi, E. and Fragaszy, D.M. (1990) Do monkeys ape? In S. Parker and K.
Gibson (eds.) cLanguage3 and Intelligence in Monkeys and Apes: comparative
development perspectives. Cambridge.
Voosinov, V.N. (1973) Marxism and the Philosophy of Language. New York and
London.
Von Wright, G.H. (1971) Explanation and Understanding. London.
Vygotsky, L.S. (1934) Myshlenie iRech': psikhologicheskie issledovaniy (Thinking
and Speech: psychological investigations). Moscow and Leningrad.
(1962) Thought and Language. Cambridge, Mass.
(1978) Mind in Society: the development of higher psychological processes, edited
by M. Cole, V. John-Steiner, S. Scribner and E. Souberman. Cambridge,
Mass.
Warren, J.M. (1973) Learning in vertebrates. In D.A. Dewsbury and D.A.
Rethlingshafer (eds.) Comparative Psychology: a modern survey. New York.
Weber, M. (1922) Wirtschaft und Gesellschaft. Tubingen. English edition (1968)
Economy and Society. New York.
Weiser, A. (1975) How not to answer a question: purposive devices in conver-
sational strategy. Proceedings of the Chicago Linguistics Society 11: 649-60.
Welch, H. (1967) The Practice of Chinese Buddhism 1900-1950. Cambridge,
Mass.
Werbner, R.P. (1989) Tswapong wisdom divination. In R.P. Werbner Ritual
Passage, Sacred Journey. Washington, D.C.
Wertsch, J.V. (1985a) Vygotsky and the Social Formation of the Mind. Cambridge,
Mass.
(ed.) (1985b) Culture, Communication and Cognition: Vygotskian perspectives.
Cambridge.
Whiten, A. (ed.) (1991) Natural Theories of Mind: evolution, development and
simulation of everyday mindreading. Oxford.
 Consolidated bibliography 297

Whiten, A. and Byrne, R.W. (1988) Tactical deception in primates. Behavioral

and Brain Sciences 11: 233-73.
Whiten, A. and Perner, J. (1991) Fundamental issues in the multidimensional
study of mindreading. In A. Whiten (ed.) Natural Theories of Mind:
evolution, development and simulation of everyday mindreading. Oxford.
Whiting, B.B. and Edwards, C.P. (1988) Children of Different Worlds: the
formation of social behavior. Cambridge, Mass.
Wiessner, P. (1982) Risk, reciprocity, and social influence on !Kung San
economics. In E. Leacock and R. Lee (eds.) Politics and History in Band
Societies. Cambridge.
Wilkins, W. and Dumford, J. (1990) In defence of exaptation. Behavioral and
Brain Sciences 13: 763.
Williams Jackson, A.V.W. (1899) Zoroaster: the prophet of ancient Iran. New
York.
Wilmsen, E. and Denbow, J. (1991) Paradigmatic history of San-speaking
peoples and current attempts at revision. Current Anthropology 31: 489-524.
Wilson, PJ. (1980) Man, the Promising Primate: the conditions of human evolution.
New Haven, Conn.
(1988) The Domestication of the Human Species. New Haven and London.
Wimmer, H. and Perner, J. (1983) Beliefs about beliefs: representing and
constraining function of wrong beliefs in young children's understanding of
deception. Cognition 13: 103-28.
Winograd, T. and Flores, F. (1987) Understanding Computers and Cognition: a
new foundation for design. Reading, Mass.
Wood, D.J., Bruner, J.S. and Ross, G. (1976) The role of tutoring in problem
solving. Journal of Child Psychology and Psychiatry 17(2): 89-100.
Woodburn, J. (1968) Stability andflexibilityin Hazda residential groupings. In
R.B. Lee and I. DeVore (eds.) Man the Hunter. Chicago.
(1979) Minimal politics: the political organization of the Hadza of North
Tanzania. In W.A. Shack and P.S. Cohen (eds.) Politics in Leadership.
Oxford.
(1980) Hunters and gatherers today and reconstruction of the past. In E.
Gellner (ed.) Soviet and Western Anthropology. London.
(1982) Egalitarian societies. Man (n.s.) 17: 431-51.
Young, J.Z. (1951) Doubt and Certainty in Science. Oxford.
Zeitlyn, D. (1983) Contradictions and alternative logics. Man (n.s.) 18: 788-9.
(1987) Mambila divination. Cambridge Anthropology 12(1): 21-51.
(1990a) Mambila traditional religion: Sua in Somie. Ph.D. thesis, University of
Cambridge.
(1990b) Professor Garfinkel visits the soothsayers: ethnomethodology and
Mambila divination. Man (n.s.) 25: 654-66.
(1993) Spiders in and out of court or 'the long legs of the law'. Styles of spider
divination in their sociological contexts. Africa 63(2): 219-40.
Ziff, P. (1971) On H.P. Grice's account of meaning. In D. Steinberg and L.
Jakobovitz (eds.) Semantics. Cambridge.
Zvelebil, K.V. (1981) Problems of identification and classification of some Nilgiri
tribes. Anthropos 76: 467-528.
 Index

Abrahams, R.G. 32 n2l 6-7, 10, 14-15, 16, 68-9, 81-3, 183;
action; AIP as prerequisite for 219; and temporality 139, 140, 144-5, 148;
accountability of 207; joint/ see also AIP entries under separate
collaborative 6, 93-4, 106, 143-4; headings, and also agonistic planning,
logic of 230-1; primate 48-50; confabulation, foresight, goal-directed
projected/planned 23, 88-94, 98-9, interaction, interaction sequences,
104-6, 128, 176-7; social 16, 184 w2, modelling, projection, strategy.
270, 273, 275; strategies for 12, 22, apes; and closeness to humans 39; and
134; synchronization of 16, 21-2; see intelligence 42-3, 44, 45; and
also coordination, goal-directed intention-attribution 47, 48-9;
interaction, interaction, interaction teaching among 46; see also primates,
sequences, social interaction hominids
activity theory 32 n\4 approval, desire for 26-7
adjacency pairs, in conversation 115, 139, Argyle, M. 32n\7
189 Aristotle 230
agonistic planning 226, 244 Ashanti prayer 209-10
agreement, preference for in conversation Atlas, J. 259^31
117-18 Austin, J.L. 142-3
AIP see anticipatory interactive planning autism 9, 45, 48, 222
Alhonsaari, A. 215 autoassociation networks 56—8, 66 w2, w3
Allen, J. 240 autobiography 275
allometric scaling 40, 43 awareness; and language 8-9, 19; and
ambiguity; of action in primates 49; in relationship with intentionality 32
conversation 23, 25; in divination 199; wl6; self-awareness 8-9, 18, 19, 44,
of intended meaning 139, 141-3, 158, 148;- see also cognition, intelligence,
164-9 passim, 170; of pre's 89, 235; see knowledge, social intelligence
also inference, intentionality,
intention-attribution, meaning Bakhtin, M.M. 108, 185 n\2
ancestor worship 209-10 Bayesian conjunction 243, 244, 247
Anesaki, M. 216-17, 218 Beattie, J.H.M. 144
animism 218, 224-5 Berger, P.L. 68-70, 77, 80, 82, 83
anticipation limit (AL) 145, 148 bias, of human intelligence; towards
anticipatory interactive planning (AIP) animism 224-5; in cognition 221; and
2-3, 32 «16, 206; and conscious interactive intelligence 224-5, 241-2,
strategy 111, 112, 114; and control 245-7, 250-1, 253-4, 255 w6; in
7-8, 129-132; evolution of 52, 68-9, judging probability 204 w4, 242-4,
82-3, 140; and communicative 253, 259 n3S; against logical reasoning
interaction 176; and intentionality 16, 13, 29, 221; against randomness
24, 26, 112-14; and language 3, 6-9, 245-7; towards social interaction 27-9
10-13, 16, 140, 206-7; in primates 18, Bickerton, D. 3, 6, 18
50-1, 68; and projectability 119, 121, Bird-David, N. 19-20
124, 127; and social roles and rules Blumstein, S. 255 nl

298
 Index 299

brain size, in primates 3, 42-3 intersubjectivity, social interaction,

Briggs, J.L. 10 social intelligence
Brown, P. 24-5, 145, 153-4, 156, 173 w5 communicative; acts 177, 178, 179-80,
Bruner, J.S. 29, 33 »22, 45, 262, 266, 267 227; constraints 180, 190, 236; genres
Buber, M. 218 26, 176-80 passim, 181-3, 185 n 12,
buddhism 216-18, 219 261, 264; problems 179, 181, 183
Burke, K. 146 'community of minds' 53
Burke, P. 33 «24 complex systems, human difficulties with
Byrne, R.W. 17-18, 47, 137ral,253, 256 248-50, 253, 259 w32, w35
nil complexity; of intention-attribution
229-30, 239-40, 258 w28; of
Carrithers, M. 28, 29-30, 262 interactional intelligence 37, 226
categorization 14 'conditional perfection' 247, 259 w31
Chafe, W. 276 confabulation 30, 273, 275-6
character, in narrative 262-3, 266, 267 Connolly, P. 257
chess, as model for social interaction/ contradiction; in divination 28, 190,
intelligence 2, 17, 145 192-5, 199-203; reinterpreting 19?>-A
chimpanzees 2, 8-9, 18, 46-7, 48; see also conventionalization see routinization
apes, primates conversation analysis 195, 196, 261
Chomsky, N. 3, 32ralO,258 w21 cooperation, in interaction 7, 18, 20, 27,
classification 6, 7 41, 253; see also collaboration,
cognition 12, 53-4, 55, 221, 264; and AIP coordination
modelling 17-18, 22-3; coordination, of interaction; and
autoassociation networks in 56-8, 66 coordination games 226-7', 256 «10,
n2, «3; of complex systems 248-50, 256-7 «13, 259 w29; and heuristics
253, 259 «32, «35; computer 241; and language 5, 6; and
modelling of 56-9, 60-1, 67 w5; and overdeterminism 244; and reflexive
dialogue 12, 21; inferred 30ral,32 reasoning 229-30; and prototypes/
ralO; and interactional intelligence stereotypes 245; and social roles 240;
250-1; and language 6, 53^4; and see also collaboration, cooperation,
primates 48; human evolution of 37, interaction sequences
51-2, 110 w8; as non-logical 13; and court cases 146, 163-4
interdependence 19, 21; structures of courtship 163
56 (Fig. 2.1), 64; social 28-9, 5 3 ^ ; Cretan liar 256 wl2
and social interaction 2, 21, 65; see curiosity, in primates 40-1
also intelligence, interactional
intelligence, logic, social intelligence, deception; and AIP 7, 17-18, 140-2; and
thought intentionality 137 nl; and irony 174
collaboration; in conversation 93-5, 96, «8; in primates 42, 47, 48, 141; and
160, 161-2; and gestures 93-4, 98-9; reflexive reasoning 170
in narrative 268-9, 275; preference for Dennett, D.C. 8, 32 wl6, 141, 266
106-7; and pre's 87, 130; see also deutero planning 270
cooperation, coordination, Devore, I. 78-9
intersubjectivity dialogic template 12, 27, 190, 207, 218
communication 25, 30 ral, 111-12, 176, dialogue; with ancestors 210; and AIP
179, 221; universality of AIP in 206-7, 218-19 passim; and animism
176-7; crossword puzzles as model for 218; and divination 197-8, 201-2; and
232, 238, 241, 246, 250, 253; and human cognition 12, 21; individual
Gricean theory of meaning 227-8, 215-16; internal 10-12, 206-7, 220 w4,
154, 169-70; and heuristics 248, 224; and learning 10-12; and prayer
251-2, 253; and interactional 12, 28-9, 208, 209, 210, 215-17; see
intelligence 251-2, 259 «37; and also communication, interaction
language 176, 177-8, 228; phatic 8, 31 sequences
«6, 32 «12, 171; regulation of 177-8, Dinka prayer 213-14
179; sequences 251-2, 259 «37; see divination 28-9, 189-90; and AIP 202-3;
also dialogue, interaction sequences, and contradiction 28, 190, 192-5,
interactional intelligence, 199-203; and dialogue 197-8, 201-2;
300 Index

divination (cont.) protolanguage 3, 22-3, 31 w5, 109;

Mambila 28, 190, 193, 198-203, 205 and symbolism 99-104 passim; see also
«13; principles of 192-3; rationality prefatory components, interaction
debate and 196; and reinterpretation sequences, questions, repairs
193-5; study of 195-8; types of 190, Gill, S.D. 211, 212-13
196-7 gift exchange 12, 273
Donald, M. 3, 22, 107, 109, 110 «8 Ginsberg, M.L. 258 «26
Dorner, D. 248-52, 255 «5, 259 «33 goal-directed interaction; and AIP 2, 7, 12,
Drew, P. 20-1, 22, 49, 115 26-7; assumption of 24; and deception
Dunbar, R. 4, 5, 8 42, 137 wl; and negotiation 208, 219;
dyadic premise 12, 207-8 projecting 121, 175, 240; ranking of
230-1; in sequences 114, 145; and
echoic mention 156 strategies 21-2; see also intentionality,
embarrassment 165-8 intention-attribution, strategy
Ervin-Tripp, S. 259 «36 Godelier, M. 275-6
ethnomethodology; and analysis of gods, and AIP 203 n\
conversation 20-3; opacity of Good, D. 23, 24, 25, 30, 49, 145
language in 204 w6 Goodall, J. 44
Evans-Pritchard, E.E. 193, 195-6, 224 Goodwin, M.H. 99, 143-4
evolution; of AIP 52, 68-9, 82-3, 140; of Goody, E.; and AIP 68-9, 114, 257 w20;
cognition 37-9, 51-2, 110 «8; of and Gonja teasing 137 w5; and
hemispheric specialization ?r-A, 31 «3; language 8, 10, 13, 141, 257 «19; and
of intelligence 1-2, 4-6, 38, 52, 65, prayer 28-9; and sexual division of
81-2; of language 3-6, 31 «5, 65, 178; labour 14; and social roles 33 «23, 240
of primates 38 (Fig. 1.1) Goody, J. 13
gorillas 43, 44, 47, 49; see also apes,
face 146, 153, 259 w33 primates
Feldman, C. et al. 264 gossip 4, 162, 165-8, 171
Fodor, J. 223, 232 Grammar of Motives 146
form-meaning pairs 54, 55, 66 grammar and syntax, of spoken language
foresight, in conversation; and ambiguity 3-5 passim, 31 «4, n5, nl, w8, 108
141-3; and retrospective revision 23, greeting exchanges 8, 113, 122-4, 137 nl
140, 144, 145-6; see also anticipatory Grice, H.P. 156, 194, 204 w7, 229, 159
interactive planning, projection, w31; and Gricean theory of meaning
temporality 154, 169-70, 227-8, 247-8
Fortes, M. 14
Fossey, D. 49 Haiman, J. 174 «12
Frames of Mind 222 Haimoff, E. 288 «25
friendship 41, 44, 164 Hall, E.T. 32rcl7
Hargreaves, R. 268
Gandhi, Mahatma 264 Harris, P. 262
Gardner, H. 222 Hartge, U. 104
Garfinkel, H. 8, 15, 190, 191-5, 201, 204 Harton, R. 203 n\
»6, 201 Havelock, E.A. 109
gambler's fallacy 204 «4, 242, 243 Hayes, C. 46
gaze, during conversation 96, 99, 100, 103 Hazlehurst, B. 18, 20, 30
Geiss, M. 259 w31 hemispheric specialization 3-4, 31 w3
Gelman, R. 224 Heritage, J. 16, 121, 122, 143
Gergen, M. and Gergen, K. 268 Herskovits, A. 258 «28
German conversation 96, 103, 105 heuristics; communicative 251-2, 253;
gestures, in interaction; collaborative 93-4, pragmatic 239, 248; and routinization
98-9; cross-cultural similarity in 22; and stereotypicality 233-4,
103^; iconic 97-8, 100-4, 106; and 237-41; see also inference, intention-
intentionality 170; as juncture attribution, intentionality, meaning
activities 94; as linguistic action hominids; Australopithecus 4, 31 w5; and
projectors 99-100, 104-6; as pre's 21, cranial capacity 3; and development of
87, 90-2, 102-6 passim', as AIP 52, 68-9, 82-3, 140; and
 Index 301

development of cognition 37-9, 51, meaning 154, 227-8, 247-8; and

110 nS; and development of shared inferential reasoning 238, 239, 258
lexicon 54, 55, 61-3; and evolution of «26; and interaction sequences 15,
social intelligence 2-6, 31 «3, 31 n5, 251, 263; in irony 24-5, 153-4, 155,
38, 52, 65, 81-2; Homo erectus 2, 3, 4; 169-70, 172; in law 143, 263; and
Homo habilis 31 «5; as Homo politeness 24, 26, 153, 169; in
interactans 223-A; and vocal signals primates 18, 47, 48-9; as a primitive
4-5, 31 «4; see also apes, primates category system 33 n22; and
hostility 162^ sequential heuristics 253; and social
Hume, D. 256^11 intelligence 26, 27, 28, 52; see also
Humphrey, N. K. 2, 145, 256 wlO bias, heuristics, inference,
Humphrey-Jolly hypothesis 37 intentionality, meaning, projection,
hunter-gatherers; Aboriginals 259 w38; social intelligence
and AIP hypothesis 69-70, 78-83; 'intentional stance' 141, 266
and kinship 69, 70, 75-7, 79-80; and interaction protocol 59-61
pre-institutionalized interaction 70, interaction sequences (verbal); and AIP
83; social interaction among 69-70, 21-3, 111-12, 114-22 passim, 127,
78-80; and social roles and rules 19, 130, 258 «25; and avoidance 122-6,
70-8 passim, 79, 81-3; see also 134; and communicative heuristics
Nayaka, !Kung 251-2; control of 129-32, 135, 146-7;
Hutchins, E. 18, 20, 30 and gestures 97-104 passim, 105-6;
and greetings exchanges 122-4, 137
Ilokano conversation 95—6, 98-9 n2; and intentionality 15, 112-14, 130,
imagination 47-8 234-7, 251, 263; negotiation of 125-6,
inference; enrichment 234; explicit 266; 147, 169, 195; and preference for
implicit 265, 267, 268; logical 231-2; agreement 117-18; and pre-requests
pragmatic 231, 259 «31; and 122-6 passim, 231', and pre-sequences
stereotypes 233-4; see also heuristics, 89, 118-19, 120, 125, 127-8, 130;
intention-attribution, intentionality, projected 98, 115, 119-21 passim, 129,
meaning, reasoning, salience 136; and repairs 124, 125; and
insults 33 «24 routines 119, 122-4, 133-6 passim,
intelligence; artificial (AI) 66, 142, 229, 156, 161, 169; silence in 19, 117,
240; and deception 17-18; 120-1, 122, 162, 238; strategies in 21,
development of 1-2, 4-6, 38, 52; 112, 121-9 passim, 133, 134-5, 171-2;
Humeian 227, 253, 256 n\ 1; and see also anticipatory interactive
language 3-12 passim; logico- planning, communication,
mathematical 260 w39; in primates coordination, dialogue, prefatory
1-2, 16, 17, 37-8, 41-2, 45; and social components, questions, repairs, social
interdependence 1-2, 16; see also bias, interaction, talk-in-interaction, turn-
cognition, interactional intelligence, taking in conversation
logic, Machiavellian intelligence, interactional intelligence 254-5 nl; and
rationality, reasoning, social AIP 257 nl9; and agonistic planning
intelligence, thought 226; bias of 224-5, 245-7, 250-1, 253,
intentionality 128, 130, 137 wl, 170; and 255 «6; bias towards in humans
AIP 16, 112-14; and cognitive 241-2, 253^; biological basis of
processing 264; consciousness of 222-5; and cognitive 'failures' 250-1;
112-14, 122-4, 133-5; and language and communication sequences 251-2,
8-9, 11,16; premise of 24, 28, 29; see 259 w37; 'conditional perfection' and
also goal-directed interaction, 247, 259 w31; and coordination
intention-attribution, projection, 226-7, 229-30, 253^, 256 wlO, 256-7
strategy nl3; and language 232-3, 257 wl9, 258
intention-attribution; and AIP 24, 26; n2l; primacy of 252; properties of
bias against randomness 245-6; 225-32; and zero-sum games 226-7,
computational complexity of 256 nlO; see also bias, cognition,
229-30, 239-40, 258 w28; and culture interaction sequences, logic,
240-1; and gambler's fallacy 204 w4, Machiavellian intelligence,
242, 243; and Gricean theory of rationality, reasoning, social
302 Index

interactional intelligence (cont.) towards 255 nl; and intentionality

intelligence, social interaction 8-9, 11, 16, 29; and interactional
intersubjectivity; and bias in human intelligence 232-3, 257 nl9, 258 nil;
thought 252^; and communicative and kinship terms 81-2; and learning
genres 178-80; and cooperation 253; 9, 10-12, 61-3; and routinization 5, 31
and irony 155-6, 160; phylogeny of nS, 107, 108; as Secondary
language and 184 n9; see also Representational System (SRS) 6, 11;
collaboration, confabulation, and self-awareness 8-9, 19; as shared
cooperation, interactional intelligence meanings system 6, 107; and social
irony (Tzeltal) 155-6; and AIP 172; as intelligence 3-5, 7-12, 5 3 ^ , 65,
distinct from joking, mockery and 107-9; and social interaction 8, 18, 27,
teasing 173 n5; and embarrassment 109, 140; and social roles and rules
165-8; hostile 162-4; positive 159-61; 6-7, 10, 13-16, 81-2; symbolic power
and reasoning 155-6, 170; routinized/ of 108, 207; see also dialogue,
conventionalized 24-5, 154-6, 161, interaction sequences, learning
170; and shared knowledge/attitudes langue 108
155, 160, 161, 168, 169, 171; and law, and interpreting intention 143, 263
social relations 25, 171-2; structure of learning; and activity theory 32 «14;
157-9; and Tenejapan women 25, computer simulation of 58, 61-2; and
155, 157, 161^, 172; and vocal pitch language 9, 10-12, 61-3; in primates
174 ralO 45, 46; zone of proximal development
and 9; and social roles 10, 66; through
Jain narrative 264—7 internal dialogue 10-12
Japanese gestures and conversation 100-2 Lee, R.B. 79
Jarry, A. 263 Leslie, A.M. 9, 32 «16, 46
Jefferson, G. 113, 131, 138 nl Levinson, S.C.; and interactional
Johnson-Laird, P.N. 220 n\ intelligence 18, 147; and routinization
joking 160-2, 171, 173 «8 22; and intention-attribution 27-8,
juncture activities 94 29, 112, 132, 259 w31, 266; and
greeting 137 nl; and politeness 145,
Kahneman, D. 224, 242-9 passim, 250-2 153-4, 156, 173 n5
Kaluli narrative 30, 270-6 lexicon, shared 54, 55, 61-3
Kant, I. 215 Lieberman, P. 4, 32 nl, 32 w8, 255 nl
'Kenny logic' 230 Lienhardt, G. 213-14, 216
kinship; and hunter-gatherers 69, 70, 75, Lindbolm, B., MacNeilage, P.F. and
76, 79-80; Nayaka avoidance between Studdert-Kennedy, M. 4
kin 77-8; in primates 41-2, 44; and literacy 13
social roles and rules 14, 19, 68-9, logic; and inference 231-2; and 'Kenny
75-7, 80-3; terms 19, 73, 74, 80, 81-2, logic' 230; mathematical 221, 260 w39;
84 n6; see also Nayaka social bias against 13, 29, 221; see also
knowledge; abstract 263; cultural/social cognition, intelligence, interactional
107, 177, 180-2 passim; shared 155, intelligence, rationality, reasoning
160, 161, 168-9, 171, 229-30, 276 Luckmann, T.; and pre's 22; and
!Kung 79-80, 85 w6 reciprocal adjustment of perspectives
25-6; and Social Construction of
labour, sexual division of 10, 14, 76-7 Reality, The 69-70, 77, 80, 82, 83
Lafon, J.-C. 255 «2 Lyons, J. 3, 22
Langacher, R.W. 108
language (spoken); and AIP 3, 6, 7-9, Machiavelli, N. 44, 142
10-13, 16, 140, 206-7; and cognition Machiavellian intelligence; and AIP 23;
6, 53-4; and communication 176, and deception 7, 18, 142; and
177-8, 228; and deception 141-2; and Humeian intelligence 147, 227, 253,
development of shared lexicon 54, 55, 256 n\ 1; in primates 44; and social
61-3; evolution of 3-6, 31 «5, 65, 178; intelligence 37, 51-2; see also
and gestures 3, 22, 31 «5; grammar intelligence, interactional intelligence
and syntax of 3-5 passim, 31 w4, n5, McHugh, P. 190
nl, w8, 108; human auditory bias Malinowski, B. 4, 8, 31 «6
 Index 303

Mambila peoples; description of 198; networks, computer models of 56-8, 59, 66

divination (Nggam) 28, 190, 193, w2, w3
198-203; and sua 198, 199-201, 202, neural structures 31 n5
205 «13; and witchcraft 199-201
Mandelbaum, J. 121 'pataphysics' and narrative 263, 276
mapping, nonlinear 67 w5 phatic communication 8, 31 w6, 32 wl2,
Marett, R.R. 218 171
Marshall, L. 79 phylogeny, of language 184 «9
Mead, M. 19, 109 Piaget, J. 224, 260 n39
meaning; form-meaning pairs 53, 54, 55, performance 210—11
66; Gricean theory of 154, 169-70, perspectives, reciprocal adjustment of 22,
227-8, 247-8; inferred 32ralO,191; 25, 26, 178-80 passim, 183; see also
interpretation of 132, 142-3, 156, 157, reciprocal monitoring, retrospective
158, 191; negotiation of 143^4, 191, revision
195; shared 6, 18-19, 30, 65, 107; see pivotal segments 131
also cognition, heuristics, inference, Plato 13
intention-attribution, knowledge, plot, in narrative 262-3, 267
salience, social intelligence politeness; cross-cultural similarities in
meditation 29, 216, 217-18, 219 153-4; and embarrassment 165-8; and
memory 147-8, 249, 259 w37 face 145-6, 153, 259 w33; functions of
modelling; AIP and 12, 17-18, 24-5, 26, 157; and intention-attribution 24, 26,
183; cognitive 22-3, 56-9, 60-1, 67 153, 169; model of 153; strategies of
w5; communicative acts 179-80; 146, 153-4; and Tenejapan women
politeness 153-4, 156-61, 164, 155, 157, 161-^
169-72; social interaction 2, 17-18, Pomerantz, A. 117, 118, 121
145, 207, 225-6 power, and social interaction 24, 142, 153,
monkeys 39, 41-2, 44, 45, 141 see also 154
apes, primates practical reasoning 230-1
monotheism 215-16, 220 w5 prayer; and ancestor worship 209-10;
buddhist 216-17; and cognition 28—9;
Nadel, S.F. 32 «20 Dinka 213-14; as dialogue 12, 28-9,
narrative; and AIP 30, 270; and character 208, 209, 210, 215-17; and meditation
262-3, 266, 267; and collaboration of 29, 217-18, 219; in monotheistic
interlocutors 268-9, 275; and religions 215-16, 220 w5; Navajo
confabulation 30, 273, 275-6; as 211-13; and ritual 210-13, 214
'cultural scenario' 28, 30, 219, 276; prefatory components (pre's); as action
and inference 265-8; and intention- projection 23, 88-94; and AIP 128,
attribution 266; and interaction 130, 258 «25; ambiguity of 89, 235;
sequences 128-9, 261; and plot 262-3, definition of 21, 87; and gestures 87,
267; and social action 270—6; and 90-2, 102-6 passim; preface questions
society 30, 261, 276; templates for 33 88-9, 237; pre-pre's 127-8; pre-
w25; and temporality 262-3, 268, 269, sequences89, 118-19, 120, 125,
272-3, 275; thought 29, 261-2, 263-4, 127-8, 130; and routinization 22, 88,
266, 272-3 108; and social intelligence 106-9; and
Navajo ritual and prayer 211-13 stories 128-9; see also dialogue,
Nayaka 19-20; avoidance between kin gestures, interaction sequences,
77-8; description of 70-2; disputes 76; questions, repairs
division of labour 76-7; and absence Premack, D. 45
of kinship roles 75-7; kinship terms pre's see prefatory components
19, 73, 74; rituals 75-6; social Presentation of Self in Everyday Life, The
interaction of 71-2, 73-7 26
negotiation; and AIP 7, 18, 27, 206; in primates; and AIP 18, 50-1, 68; brain size
conversation 125, 126, 147, 169, 195; among 39-40; evolutionary taxonomy
of meaning 143-4, 191, 195; in social of 38 (Fig. 1.1); and friendships/
interaction 7, 18, 20, 27, 147; see also kinship 41-2, 44; and deception 42,
goal-directed interaction, projection, 47, 48, 141; and imagination 47-8;
strategy and intelligence, development of 1-2,
304 Index

primates (cont.) gestures, interaction sequences,

16, 17, 37-8; and intention- prefatory components, questions
attribution 18, 47, 48-9; interpreting representation 56; cultural 16, 88, 207,
actions of 48-50; and learning ability 216; internal 65; Secondary
45, 46; and Machiavellian intelligence Representational System 6, 11;
37, 44, 51-2; play and pretence among semantic 232-3, 257 w20; spatial 93;
45-6; and self-awareness 8-9, 44; and see also symbols
social cooperation among 18, 41; and retrospective revision; constraints of
social intelligence 41-2, 4 3 ^ , 46, 146-7; in conversation 23, 140, 144,
50-2; and social interdependence 1-2, 145-6; in divination 192-3; see also
4, 16, 41-2, 43-4; and vocal signals 4; reciprocal monitoring; perspectives,
see also apes, chimpanzees, gorillas, reciprocal adjustment of
hominids, monkeys rhetoric 158-9, 201
Prince, The 142 ritual; Dinka sacrificial 213-14; efficacy of
probability, human bias in judging 204 w4, 210; Navajo hoop ceremonial 211-13
242-4, 253, 259 w38 routinization/conventionalization; and
projection 87-110 passim; of action 88-94, AIP 25-6, 177; and communicative
100, 104-6, 134; and AIP 119, 121, genres 26, 180-1; in conversation
124, 127; and gestures 99, 100-3, sequences 21, 119, 122^, 133-6, 156,
104-6; and interaction sequences 98, 161, 169; of gestures 104; of greeting
115, 119-21 passim, 129, 136; see also 123-4; of grammar and syntax 31 «8,
anticipatory interactive planning, 108; or irony 24-5, 154-6, 161, 170;
foresight, memory, temporality and language 5, 31 «8, 107, 108; and
prothetic devices 88 pre's 22, 88, 108; and social
protolanguage 3, 4, 5, 22-3, 31 «5 interaction 15-16, 177; strategic use of
public opinion 162 133-4, 154-5, 170; see also interaction
sequences
questions; and divination 189, 191-4, 199, Rumelhart, D., Hinton G. and Williams,
202; in interaction sequences 118-19, R. 58
127, 235; preface 88-9, 237; rejection
of 201; rhetorical 155-8 passim, 159, Sacks, H. 122-3, 132, 133, 234
163, 166, 168; strategic 24, 113-14 salience; mutual 226, 230; presumption of
246; see also inference, meaning
rank, and social interaction 44, 153 sarcasm 174 n\2
ratiocination 196-7, 202, 259 «38, 266, 268 Scarry, E. 218
rationality 28, 153-4, 195-6, 250; see also 'shallow symbols' hypothesis 65
cognition, intelligence, reasoning, Schegloff, E.A. 88, 95, 109, 127, 128, 234
social intelligence Schelling, T. 221, 226-9 passim; and
reasoning; ethical 266; inferential 154, Schelling games 239, 241, 246, 256
234-5, 238-9, 258 w24, «26; and irony ralO, 258 w24
155-6, 170; logical 13, 29, 221, 230-2, SchiefTelin, B. 274
260 w39; practical 230-1; reflexive 24, SchiefTelin, E. 270—5, 276
154, 169-70, 226, 229, 238; see also Schiffer, S.R. 229-30
cognition, inference, logic, projection, Schone, H. 255 ril
rationality Schutz, A. 185 n 17, 262
reciprocal monitoring, in conversation 12, Searle, J.R. 142
23,25,26, 179, 183; see also Secondary Representational System (SRS)
retrospective revision 6, 11
religious belief 8, 208 sequences see interaction sequences
repairs (conversational) 93; and avoidance Shaha, S. 265, 267
125; and communication 179; in Siddhasagar 265, 266, 267, 269
divination 203; initiation of 95; and silence, use of in conversation 19, 117,
interaction sequences 21, 124, 125; 120-1, 122, 162, 238
and memory 147; and negotiation of Simon, H. 65
meaning 195; self-repair 95, 96, 99, Social Construction of Reality, The 68-70,
106, 147; segments 94-5; and word- 77, 80, 82, 83
search 96, 106; see also dialogue, social institutions; and communicative
 Index 305

genres 178, 181-3; emergence of Sperber, D. 30ral,32 «10, 53, 56, 156,
13-16, 69, 83, 208; see also kinship, 268
social roles and rules, social spirit possession 196-7, 214
interdependence stories see narrative
social intelligence; as communicative strategy; for action 12, 22; conscious use of
behaviour 111; and cultural resources 112-14, 122-4, 133-5; in conversation
88, 107; and deception 7, 17-18; 121-2, 122-6, 126-9, 171-2; and
dyadic premise of 207-8; and politeness 145-6, 153-4; and routines
intention-attribution 26, 27, 28, 52, 133-4, 154-5, 170; see also
112; and Machiavellian intelligence anticipatory interactive planning,
18, 51-2, 147; and language 3-5, 7-12, foresight, projection, interaction
53-4, 65, 107-9; as a network system sequences
107-8; and preference for Strategy of Conflict 226
collaboration 106-7; and pre's 106-9; Strathern, M. 32 «18
in primates 41-2, 4 3 ^ , 46, 50-2; and Streeck, J. 21, 22, 104, 114
routines 107, 133; see also cognition, strepsirhine primates 38-9
interactive intelligence sua 198-202 passim, 205 wl3
social interaction; and AIP 175, 183; as symbolic resources 107
basis for human cognition 65; and symbols; and AIP 102, 103; gestural
cooperation/negotiation 7, 18, 20, 27, 99-104; and language 54, 55, 65, 66,
147; coordination of 226-7, 229-30, 207; natural 109; see also
241, 244-5, 253-4, 256 wlO; and representation
deception 140-2, 148; denned 175,
184 wl; and desire for approval 26-7; talk-in-interaction 108-9, 111, 112, 113,
as distinct from social action 184 n2; 133
and gestures 93-4; and hunter- Tambiah, SJ. 210-11,217
gatherers 69-70, 78-80; and language teasing; collaboration in 161-2; Gonja 137
8, 18, 27, 109, 140; routinization of «5; in conversation sequences 21-2,
15-16, 177; see also interactional 115-17, 119, 137 «5, 137-8 n6
intelligence, interaction sequences, temporality; and AIP 139, 140, 144-5,
kinship, social interdependence, social 148; and conversation 237, 259 w36;
institutions, social roles and rules and narrative 262-3, 268, 269, 272-3,
social interdependence; and AIP 2-3; as 275; temporal thinking, biases in 252;
basis for primate intelligence 1-2, 16; see also memory, foresight projection,
and cognition 19, 21; and retrospective revision
intentionality 11, 16, 27; non-verbal Tenejapan; description of 172 wl; use of
monitoring of 12, 32, wl7; and social irony in 156; and joking 171-2, 173
roles and rules 13-16; in primates 1-2, w4; and lies 173-4 «8; women 25, 155,
4, 16, 41-2, 43-4; see also kinship, 157, 161-4, 172; see also irony, Tzeltal
social institutions, social roles and speech
rules Teraski, A. 125, 235, 236
social roles and rules; and AIP 6-7, 10, thought; and intersubjectivity 252^;
14-15, 16, 68-9, 81-3, 183; and logical mode 29; narrative 29, 261-2,
character 262, 269; defined through 263-^, 266, 272-3; and temporality
language 6-7, 10, 14; development of 252; see also cognition, intelligence,
81-3; and interactional coordination interactional intelligence, logic,
228, 240; and kinship 14, 19, 68-9, rationality, reasoning, social
70-8 passim, 80-3; and language 6-7, intelligence
10, 13-16, 81-2; and sexual division topics, introducing in conversation 130-2
of labour 10, 14, 76-7; and social trajectory components 99-100
interdependence 13—16; and social Trevarthen, C.B. 223, 224
regularity 15-16; see also Nayaka, turn-taking in conversation; collaboration
social institutions, social in 94; and communicative genres 182;
interdependence and gestures 99, 104; and goal-
society, human 13, 14, 182, 261, 276 achievement 145-6; and human
sound-stretches 95-6 attention span 251-2, 253; and
Spelke, E. 224 projecting sequences 111, 120; see also
306 Index

turn-taking in conversation (cont.) Wason four card problem 247

interaction sequences, Wearing, A. 257 w20, 259 «37
talk-in-interaction Weiser, A. 142
Tversky, A. 224, 242-9 passim, 250-2 Whiten, A. 7, 37, 47, 137 n\, 253, 256 wll
Tylor, E.B. 218 Wilson, D. 30 »1, 32ralO,53, 56, 156, 268
Tzeltal speech; conventionalized 156, 161, Wilson, P J. 14
169, 170; irony in 24-5, 155-73, 173 witchcraft, suspicion of 199-201, 271
rc2, «4; rhetoric 158-9; see also irony Wittgenstein, L. 221
(Tzeltal) Woodburn, J. 70, 79
word-searching, during conversation 90,
Uktu Inuit peoples 10 96-8, 106

Young, J.Z. 255 n6

Velmans, M. 32 n\6
vocal; pitch 174 n 10; signals 4—5, 31 «4; Zeitlyn, D. 27-8, 190
tract 4 zero-sum games 226-7
Vygotsky, L.S. 9, 10-11, 18, 19, 206-7, zone of proximal development 9
224 Zwicky, A. 259 «31