Poetics8 (1979) 193-211

@North-Holland Publishing Company

EXTENSIONALANDINTENSIONALNARRATIVEWORLDS

LUBOMiRDOLE2EL

Using Frege’s distinction of reference (extension) and sense (intension), as well as some
ideas of possible world semantics, the paper proposes to treat narrative macrostructures in
terms of narrative worlds. The referent of the narrative text is given in the extensional primary
world. This world is linked with an abstract secondary world through individuating functions
which provide a generic (typological) interpretation of the referent. On the other hand, the
extensional primary world is projected into an intensional world through the application of
intensional functions. These functions expressing the regularities of texture are credited with
the macroorganization of sense. Two basic intensional functions - the authentication function
and the explicitness/implicitness function - are discussed in some detail. While the first func-
tion is connected with the problem of truth in narrative worlds, the second one is used to
explain the difference between ‘visible’ and ‘invisible’ narrative worlds.

Theory of literature has been concerned with two fundamental questions: (a)
how are works of literature ‘made’, i.e. what are their formal properties? (b) what
do works of literature mean, i.e. what kind of information do they convey? Lip-ser-
vice has often been paid to the principle of the unity of form and meaning, but
explicit explanations of this unity are hard to come by. The formalist tradition,
focusing on the first question, and the hermeneufic tradition, offering answers to
the second, have been in general developing in isolation, without mutual contact.
This unfortunate situation is recently coming to an end, especially concerning the
theory of narrative texts. It has always been obvious that a theoretical account of
narratives has to incorporate an account of meaning, an account of ‘what happens’
in a story; once students of narrative also realized that narrative works are ‘made’,
i.e. that they display specific narrative forms, the way for a synthetic nurrafive
semantics was open. Unlike traditional hermeneutics, narrative semantics is a study
of meaning expressed in forms, a study of formally organized meaning. Unlike for-
malism, narrative semantics does not claim that the aesthetic effect results from
‘pure’ forms; rather, its source is located in the totality of the literary work.
If meaning is given in the forms of literary texts, then it is an intrinsic property
of these texts. The formal base of meaning warrants the objectivity (inter-subjec-
tivity) of textual meaning. The task of narrative semantics is to recover and to for-
mulate the intrinsic meaning of narrative texts. In this respect, narrative semantics
is in absntllte contrast to subjectivist criticism, which is, in principle, an act of

193
194 L, Doleiel / Extensional and intensional narrative worlds

meaningassignmenr. For the subjectivist critic, the text is just a pretext for triggering
a virtually infinite set of possible ‘interpretations’. For literary semantics, however,
a literary text is a system of restrictions articulating a particular meaning. If, as it is
often claimed by subjectivist criticism, a text can mean anything, then all texts
necessarily mean the same. Literary semantics, conceived as the theory of semantic
restrictions, avoids this counter-intuitive consequence.
A narrative text is a highly complex network of semantic restrictions. Of neces-
sity, narrative semantics formulates, in the first approach, partial models of the
semantic structure of narrative texts. At present, we can hardly see the day when
these partial models will be integrated into a complete model of narrative seman-
tics. ’ In this paper I want to concentrate on some aspects of the macrostructural
model, i.e. on the system of concepts designed for the description of the global
semantic restrictions operating in narrative texts. Logically, macrostructural models
have a priority over microstructural models: a theory of elementary semantic con-
stituents is to be developed in conformity with the general framework provided by
the chosen macrostructual model.

1. General properties of narrative worlds

This paper is based on the assumption that the global semantic restrictions oper-
ative in narrative texts can be described in terms of narrative worlds. The concept
of narrative world is the basic concept of the macrostructural model of narrative
semantics. The term ‘world’ has been used in literary criticism (in such phrases as:
‘the world of science-fiction’, ‘the Proustian world/universe’, ‘the world of Crime
and f’unishment’, etc.) as one of many critical metaphors. At present we are in a
position to transform this metaphor into a sound theoretical concept by relating it
to the logico-semantic notion of possible worlds. Some recent studies in narrative
theory (van Dijk 1974/75; Doleiel 1976; Pave1 1976) have already opened up
exploration of this link. However, in any confrontation of literary and logical con-
cepts, the different scopes and aims of literary and logical semantics must be borne
in mind. The prime concern of logical semantics has been the notion of truth;
accordingly, the notion of possible worlds has been introduced in order to formal-
ize the assignment of truth-values to propositions (atomic formulae) in modal con-
texts (Hintikka 1973: 198). * In literary semantics, as will be pointed out later, the

t In his recent criticism of some narrative models, van der Eng has emphasized the necessity of
taking into account dynamic aspects of narrative texts (van der Eng 1978). While it is true
that in narrative texts the aspect of change is essential (changes of situation, changes in the
properties and relations of characters), it would be a mistake to underestimate the importance
of ‘static’ models: by formulating the general concepts of narrative semantics, ‘static’ models
offer a theoretical base for ‘dynamic’ models.
* A formal outline of possible-world semantics can be found in Kripke 1963; for literary
semantics, Kripke 1972, and Hintikka 1975, are especially relevant.
 L. Doleiel /Extensional and intensional narrative worlds 195

concept of truth is problematic; in any case, meaning in literary texts (and in natu-
ral language texts in general) cannot be reduced to the concept of truth or truth-
conditions.
The necessity of formulating a broader theory of meaning for natural languages
was clearly recognized by G. Frege; his introduction of the well-known distinction
between ‘reference’ (Bedeutung) and ‘sense’ @inn) was meant to provide a basis for
such a theory (Frege 1892). The difference between the logical, truth-based seman-
tics and the Fregean semantics has been explicitly formulated as follows: “From the
standpoint of logic as such, we need an account of the working of language only as
it relates to truth ... . Frege’s philosophical concerns go a long way beyond anything
that is the proper concern of the logician” (Dummett 1973: 83). 3 It is especially
significant for students of literature that Frege singled out works of verbal art
(exemplified by epic poems) as lying outside the domain of truth: “In hearing an
epic poem ... we are interested only in the sense of the sentences and the images
and feelings thereby aroused. The question of truth would cause us to abandon aes-
thetic delight for an attitude of scientific investigation” (Frege 1892: 125; cf. also
Gabriel 1970).
In spite of the essential difference between logical and literary semantics, two
ideas of possible-world semantics seem to me fundamental for formulating the gen-
eral properties of narrative worlds:

(A) While possible-world semantics takes account of the ‘actual’ world, its scope
extends over an infinite set of possihle worlds alternative to the ‘actual’ world.
These alternative worlds are not empirically observable; rather, they are postulated
as logical ‘constructs’ (Adams 1974: 227, Hintikka 1975: 28). 4 This idea clearly
applies to narrative worlds: from the beginning, narratives have been involved in
constructing possible worlds alternative to the ‘actual’ world of human experience.
Narrative worlds can be characterized as specific semiotic ‘constructs’ built from
certain narrative entities. Their existence is textual, stipulated by nothing other
than the existence of the corresponding texts. If we accept the special ontological
status of narrative worlds, two consequences become readily apparent: (a) narrative
world is fuZZydetermined by the text in which it is described (expressed). It is a
product of the text, while at the same time, it is the basis of its meaning;(b) narra-

3 It should be noted that truth-based semantics does not explain meaning in terms of truth-
values, but in terms of truth-conditions. For a recent account of truth-based semantics, see
Kempson 1977; it is symptomatic that Kempson has no use for Frege’s distinction between
sense and reference.
4 Strictly speaking, the differentiation of the empirically observable ‘actual’ world and the
constructed possible worlds is characteristic only of the ontological position designated as actu-
alisrn (Adams 1974); the opposite position - possibilism - assumes no ontological difference
between the ‘actual’ world and its possible alternatives. While the latter position might be justi-
tied for some purposes of format logic, only the former position offers a sound basis for literary
(as well as scientific) semantics.
196 L. Dolez'el/ Extensional and intensional narrative uorlds

tive world is an aufonomous semiotic entity existing independently of the ‘actual’

world. To be sure, narrative worlds are in many respects and in different ways
ielated to the ‘actual’ world; however, this is an exterior relationship between two
autonomous macrostructures. In principle, a narrative world does not have to con-
form to the structure of the ‘actual’ world any more than the world of Euclidean
geometry conforms to the worlds in which non-Euclidean geometries are valid.

(B) A possible world is not a random assemblage of entities; it is constructed in

accordance with certain postulated global principles. According to Kripke, “a pos-
sible world is given by the descriptive conditions we associate with it” (Kripke
1972: 267). Kripke’s ‘descriptive conditions’ are clearly analogous to the global
semantic restrictions which, as we have claimed, are responsible for the macrostruc-
tural semantic organization of narrative texts. By imposing global restrictions, we
construct a narrative world whose entities are semantically homogeneous in some
respect(s). Under this interpretation, narrative world appears as an organized macro-
system of constituents (entities). As such, it represents an eminently suitable con-
ceptual tool for expressing the idea of organized meaning.

2. Primary narrative worlds

Keeping in mind the generai ontological and structural properties of narrative

worlds, we can now proceed to introduce the concept in a more definite way, It is
obvious that every narrative text constructs its own narrative world. I propose to
call the world corresponding to a particular text primaT narrative world (primary
NW). The simplest way to introduce the primary NW is in terms of its ‘inhabitants’,
i.e. in terms of the individuals who exist in it. If the individuals of narratives are called
narrative agents (acting characters), then the primary narrative world can be defined
as a set of compossible narrative agents. ’
This definition clearly has a Leibnizian ring. Leibniz thought of possible worlds
as collections of individual concepts where each individual concept is interlocked
with all the other individual concepts of that world. Individual concepts are com-
possible if they are capable of joint realization (Mates 1968). Our definition is in
agreement with the general properties postulated for narrative worlds: the primary
narrative world is constructed from specifically narrative entities (narrative agents)
and is thus autonomous with respect to the ‘actual’ world; it has a systemic charac-
ter, given by the restriction of ‘compossibility’. This restriction differentiates one
primary NW from all other primary NWs. Indeed, while Emma Bovary is compos-
sible with Rodolphe Boulanger - these agents exist in one and the same primary
world - she is not compossible with, say, Ivan Karamazov - we are dealing with

’ In possible-world semantics the set of individuals of a particular world is called its ‘domain’
(Kripke 1963: 80.5).
 L. Doleiel /Extensional and intensional narrative worlds 197

agents from different primary worlds. Clearly, the empirical criterion of compossi-
bility is given by reference to corresponding texts: this is in accord with our general
claim that a narrative world is constructed in the text.
A full understanding of the concept of primary NW requires a brief discussion of
yet another aspect: for every narrative text, the set of its agents can be given by
simple enumeration; in other words, the set is defined by extension. Moreover, we
assume that in the primary world the agents exist as entities independently of any
kind of designation which can be assigned to them. Later, we shall consider the
problem of speaking about such purely referential entities. Now let us merely note
that the postulated properties - enumerability of its members and their indepen-
dence of designation - characterize the primary NW as an extensional concept.
6
We shall speak simply about extensional primary narrative worlds.
While cornpossibility of agents is the most general restriction imposed on pri-
mary NWs, other less general restrictions operate in such a way as to partition the
set of agents of a primary NW into distinct subsets. Thus, for example, by taking
the relationship to the ‘hero’ as the partitioning criterion, we split the set of agents
of a primary world into three subsets: the hero, his ‘allies’ and his ‘enemies’ (a
fourth set of ‘neutrals’ can be also envisaged). Let us note that this kind of parti-
tioning of the extensional world is itself extensional: the world is split into sub-
worlds by splitting the enumerable set of agents into enumerable subsets. The
extensional character of this partitioning indicates the theoretical importance of the
operation: the system of extensional subworlds with its possible relations and hier-
archies represents the organization of meaning in narrative texts on the level of
Fregean reference.
For the definition of the primary NW, the most readily available narrative entity,
that of agent, has been chosen. However, the definition can be expanded to incor-
porate actions and properties of agents as well. Every agent performs a certain set
of actions - let’s call this set his action scope, and he is assigned a certain set of
properties - his property set. ’ The expanded definition of the extensional primary
NW then includes the set of compossible agents together with the corresponding
action scopes and property sets. Actions and properties are assigned to agents in the
form of propositional functions. The extensional primary NW then appears as a set
of propositional functions defined over the set of compossible agents. ’

6 I am using the (Carnapian) terms ‘extensional’-‘intensional’ as equivalent to the (Fregean)

terms ‘reference’-‘sense’; we are forced into this usage by the impossibility of deriving a satis-
factory adjective from ‘sense’, but it does not commit us to the indentification of Frege’s
dichotomy with Carnap’s (the identification is commonly made in possible-world semantics, cf.
Hintikka 1975: 76 passim).
’ For the sake of simplicity, the term ‘property’ is used in the broadest sense, comprising
relations between agents.
* A further expansion of the concept is possible if we include narrative objects (non-agents)
and their properties; however, these additional components of the primary extensional NW do
not affect its basic properties.
198 L. DoleZel/ Extensional and intensional narrative worlds

We have to emphasize once more that agents, their actions and their properties
qua entities of extensional worlds are given independently of any designation
(name, description, expression). However, in extensional narrative semantics we
have to speak about these entities, we have to identify and differentiate them. For
this purpose, we need a strictly extensional language characterized by one-to-one
correspondence between the narrative entities and their designations. We should
add that we are here faced with a problem which is not specific for narrative seman-
tics; it is a fundamental problem of general semantic theory. At present, the task of
constructing a purely extensional language has not been resolved satisfactorily. Per-
haps most appropriate for the purposes of narrative semantics is the system of seman-
tic representations reflecting the ‘logical forms of sentences’, as developed by linguis-
tic semantics. In this language propositional functions can be expressed in the form
of predications. If narrative predications are called motifs, 9 we arrive at the final
modification of our definition: extensional primary NW is a set of motifs with argu-
ment variables ranging over the set of compossible agents and predicate variables
ranging over their action scopes and property sets. We shall use this definition
whenever the need arises to speak about identified and differentiated narrative enti-
ties.

3. Secondary narrative worlds

The concept of primary NW is essential for describing the semantic macrostruc-

ture of particular narrative texts. However, narrative theory is interested not only in
semantic regularities of particular texts, but also in regularities which can be
revealed in text classes. We have to develop an abstract concept of narrative world
which would be appropriate for this task of narrative semantics.
The concept of primary NW allows a comparison of the semantic macrostruc-
tures of different narrative texts. Two different modes of this comparison can be
distinguished: (a) comparison without cross-world identification; (b) comparison
with cross-world identification. It seems to me that this distinction has a general
import for literary theory, corresponding on the one hand to traditional compara-
tive criticism, and on the other to literary typology.
The comparison without cross-world identification, as the term indicates, keeps
the compared agents within the scopes of their respective primary worlds; the
search for similarities and/or differences in their properties and actions is carried
out in such a way that the agents are treated as distinct and separate individuals
rather than as representatives of an abstract ‘type’. Thus, for example, we might
look for similarities and/or differences between Emma Bovary and Anna Karenina,
without reducing these agents to a ‘type’. Generally speaking, comparison without

9 For an account of motifs as predications see Doleiel 1976.

 L. Dolefel /Extensional and intensional narrative worlds 199

cross-world identification does not require the construction of abstract categories

superimposed on the set of the compared primary worlds.
On the contrary, such a procedure is required in comparison with cross-world
identification. Let us turn once again to possible-world semantics to clarify the
essence of the method. In the formal language of possible-world semantics, cross-
world identity is expressed by the so-called individuating function. Hintikka has
described the operation of this function in the following words: “Each individual in
the full sense of the word is now essentially a function which picks out from several
possible worlds a member of their domain as the embodiment of that individual in
this possible world” (Hintikka 1975: 30).
Hintikka’s ‘individual in the full sense’ corresponds to the abstract ‘type’ of nar-
rative agent called actant. Using Hintikka’s suggestion, we can define an actant as
the individuating function which picks out in several primary narrative worlds one
agent as an embodiment of that actant in that particular primary world. lo How-
ever, this concept of actant becomes fully relevant for macrostructural narrative
semantics only if we postulate that actants themselves form higher-order narrative
worlds. Narrative worlds defined as sets of cornpossible actants shall be called sec-
ondav narrative worlds. The comparison with cross-world identification is, then,
based on the construction of a secondary world embracing a set of primary narra-
tive worlds. The secondary world and the corresponding primary worlds are linked
by individuating functions, with actants as arguments and agents as values.
In order to demonstrate that this formal description really represents a well-
known method of narrative semantics, let us invoke the popular system proposed
by V. Propp (1928). Propp has determined that the variety of agents of the Russian
fairy-tales (in fact, agents of 100 narrative texts) can be subsumed under seven
actants. I1 These actants can be said to be cornpossible since they are both neces-
sary and sufficient for the formation of the ‘prototype’ of the Russian fairy-tale.
For this reason, the Proppian world can be designated as a secondary narrative
world. In this interpretation, each Proppian actant is an individuating function
which picks out from the primary narrative worlds (corresponding to the particular
fairy-tale texts) an agent embodying this actant. Thus, for example, the actant ‘vil-
lain’ is embodied in the primary narrative worlds by such agents as Baba Yaga, the
sorcerer, the step-mother, etc.
Let us note that the Proppian system is just one of several similar systems of
abstract narrative semantics; L&i-Strauss system of myth analysis, the Jungian
system of ‘archetypes’ (as well as other archetypal systems), all reveal the same

lo The term ‘actant’ originates in French linguistics (cf. Tesniere 1959); Greimas has intro-
duced it into narrative semantics (Greimas 1966). It should be emphasized that individuating
functions of narrative semantics are not necessarily binary; that means that one and the same
agent can embody more than one actant and vice versa (cf. Greimas 1970: 157-183).
*’ Many revisions of Propp’s system of actants have been proposed; however, they do not
affect the general point discussed here.
200 L. Dole?el ! Extensional and intensional narrative worlds

formal properties. In all these systems, the potentially unlimited number of narra-
tive agents of a potentially unlimited set of primary narrative worlds is reduced to a
limited number of actants of a specified
__~_~ _.._~ secondary world. And in all these systems,
narrative agents receive an interpretution in terms of actants, and primary narrative
worlds in terms of the postulated secondary world. The ‘interpretative’ function is
nothing else than the converse of the individuating function, a function from agents
to actants, from primary worlds to secondary worlds. Clearly, all ‘interpretative’
systems of narrative semantics are based on cross-world identification formally des-
cribed by the concept of secondary worlds with corresponding individuating func-
tions.
In the case of secondary narrative worlds, their extensional character is even
more readily apparent than in the case of primary worlds: the set of actants is
clearly enumerable since it is postulated by the analyst. The designations of the sec-
ondary world entities, designations of actants, ‘functions’, ‘motifemes’, etc., are
purely extensional, since they are given as unambiguous terms of the proposed sys-
tem. ‘Interpretative’ systems of narrative semantics are mappings of one extensional
world into another extensional world; they do not leave the space of Fregean
reference.
It should be noted that the theoretical concepts just discussed do not provide
any empirical criteria for ‘interpretative’ narrative semantics. In considering the
primary NW, we could formulate an empirical criterion for determining the set of
its agents: this set is equivalent to the set of agents given in the corresponding text.
Secondary narrative worlds do not correspond to particular narrative texts; rather,
they correspond to text classes. But the set of actants of a secondary world cannot
be given as the set of all agents of the text class; in such a case, we would abandon
the operation of cross-world indentification. Consequently, the set of actants has to
be suggested as a hypothesis. What is more embarrassing is that we have no empiri-
cal criteria of individuating functions; the decision as to which agent(s) of the pri-
mary narrative worlds will be selected as the embodiment of a particular actant is
largely intuitive. Consequently, no exact procedure for testing a proposed ‘inter-
pretative’ system is available. For these reasons, ‘interpretative’ narrative semantics
is a field where sound hypotheses easily mix with products of pure imagination.
Our model does not offer a cure for these empirical defects; however, by formulat-
ing the theoretical framework of ‘interpretative’ narrative semantics, it calls atten-
tion to these defects.

4. Intensional narrative worlds

We have based the concepts of primary narrative world, as well as that of the sec-
ondary world, on the principles of extensionality. These concepts have been
defined in such a way as to fall within the space of Fregean reference. But since we
have accepted Frege’s idea about the prime importance of sense in literature, we are
 L. Doleiel /Extensional and intensional narrative worIds 201

faced with the formidable task of complementing extensional narrative macrose-

mantics with an intensional narrative macrosemantics.
Before we proceed, a few clarifying remarks about the notion of sense are in
place: (a) the semantics of sense has been sketched primarily with respect to ele-
mentary semantic units, words and word-equivalent lexical items (such as ‘morning
star’). The sense of more complex expressions (for example, sentences) is explained
in Fregean semantics as an accumulation of senses of its constituents (Dummett
1973: 152f.). This simplistic view is of little help in developing a theory of sense for
the most complex verbal unit. the text. Obviously, a theory of text sense will have
to offer more satisfactory explanations of sense formation. (b) We have to empha-
size that the Fregean concept of sense is inextricably linked to expression; indeed,
sense is determined by the form of expression. Sense for Frege (1892: 112) was
“the mode of presentation” of the reference; l2 this mode, we assume, is given in
the form of the corresponding expression. Similarly as the sense of the lower level
semantic units is determined by the form of words, phrases and sentences, the sense
of the text is determined by the form of texture. Texture is our term for the ‘word-
ing’ of the text and it comprises all lower-level units of expression as its constitu-
ents. The study of text sense is, therefore, closely linked with the study of the pos-
sible forms of texture.
In principle, every constituent of the texture contributes to the determination of
text sense. However, the macrosemantic approach is not concerned with the partic-
ular sense constituents; rather, it is interested in global principles ofsense organiza-
tion. Since we assume that text sense is determined by the forms of the texture,
principles of sense organization will be determined by global regularities of texture.
Consequently, the macrotheory of text sense formulates the principles of sense
organization in revealing the global regularities of texture.
In order to prepare the ground for a more explicit discussion of the essential
relationship between textual regularities and sense organization, let us start with an
example. We know that the ‘inhabitants’ of the extensional primary NW are narra-
tive agents. In the texture agents can be designated primarily by proper names
(Odysseus) or definite descriptions (the king of Ithaca). In some narrative texts we
observe a regularity in the distribution of proper names and definite descriptions
over the set of agents. Such is, for example, the case in- Defoe’s Robinson Crusoe:
while the hero and some of his close associates are given proper names in the tex-
ture (Robinson, Friday, Xury), secondary agents are designated by definite descrip-
tions (my father, the Portugal captain, the English captain’s widow, Friday’s father,
etc.). Such a regularity of texture can be described formally as a two-value function

12 In this respect Frege’s example from the language of geometry is even more instructive than
his famous ‘morning star’-‘evening star’ example. In view of this, it is difficult to understand
the claim, according to which “senses.. . are not linguistic entities” (Grossmann 1969: 156);
not surprisingly, Grossmann comes to the conclusion “that there are no such entities as senses”
(1969: 164).
202 L. Dokfel /Extensional and intensional narrative worlds

Extensional Intensional
world world

~~~~J

Fig. 1. X represents an agent, l is an agent expressed by a proper name. o an agent expressed by

a definite description.

which assigns to each agent of the extensional primary NW a proper name or a

definite description in the texture. We claim that functions of this type project the
extensional primary NW into its intensional correlate; therefore, they are called
in tensional functions. I3 A schematic representation of this mapping is given in
fig. 1.
There are two fundamental reasons for our claim that we have reached the do-
main of sense: (a) one and the same extensional primary world can be projected
into many possible intensional worlds by the application of different intensional
functions. This corresponds to the general feature of Fregean semantics: a variety
of senses corresponds to one and the same referent, the various senses ‘illuminating’
particular aspects of the referent (Frege 1892: 119f.). (b) The intcnsional world
and its structure is fully determined by the forms of texture, by the texture regular-
ities expressed in intensional functions. This determination of sense by the forms of
expression is the essence of the Fcegean notion of sense, as we have emphasized.

5. A scheme of narrative macrosemantics

A complete model of narrative macrosemantics is represented in fig. 2; it is a

simplified representation, since it shows neither the possible set of primary exten-
sional worlds corresponding to one secondary world nor the potential variety of
intensional worlds assigned to one extensional primary world. Individuating func-
tions are assumed to be binary, intensional functions two-valued. (The entities of
the secondary world can be interpreted as actants, the entities of the primary
world as agents and those of the intensional world as ‘named’ agents. The corres-
ponding forms of language representation: artificial language of ‘interpretative’ nar-
rative semantics; standardized language of semantic representations; natural language
texture.)

13 Intensional functions are not necessarily two-valued; many-valued intensional functions arise
whenever the choice between more than two forms of expression is given.
 L. DoleIe/ / Extensional and intensional narrative \corlds 203

Extensional Intensional
Secondary worlds Primary worlds
world
O individ? ($ Inten!: 0

Fig. 2.

We began our discussion by introducing the concept of the extensional primary

NW; the domain of this world is given by the set of compossible agents and the
world includes the action scopes and property sets in the form of propositional
functions. The extensional primary NW is assumed to have two functional links: (a)
the individuating/interpretative functions link the primary extensional NW to
abstract secondary worlds; these worlds are constructed as sets of cornpossible
actants. Agents of the primary worlds are ‘interpreted’ as actants, their actions as
‘functions’ (in the Proppian sense), etc. Secondary worlds are hypothetical ‘inter-
pretative’ systems for the description of generic(typologica1) properties of narrative
texts. (b) On the other side, intensional functions representing the global regular-
ities of texture project the primary extensional NW into one of possible intensional
worlds. By the application of these functions, narrative texts are assigned their sin-
gular sense organization. In such a way, our model offers a framework for describing
both the generic and the singular aspects of semantic macrostructures, following the
double task of narrative semantics which has been in the center of its attention
from the beginning of theoretical research.

6. Basic intensional functions

Since the text sense is inextricably linked to forms of texture, a description of

sense in terms of a paraphrase is inadmissible; as a tool for capturing sense, para-
phrase is, indeed, a ‘heresy’ (cf. Brooks 1947). However, rejecting a certain method
of sense description does not mean abandoning the task of formulating a theory of
text sense. Observable regularities of texture provide a clue to the organization of
sense and, therefore, this organization is amenable to theoretical and empirical
study. In fact, the method of capturing sense through the forms of its expression
has been practiced with respect to such intensional micro-phenomena as metaphor,
symbol, etc. A theory of the macro-sense of the text based on intensional functions
is clearly analogous to these micro-structural approaches.
In introducing the concept of intensional function, we have used a schematiza-
204 L. Dolriel /Extensional and intensional narrative worlds

tion in order to facilitate comprehension. The representation given in fig. 1 has to

be supplemented by an account which corresponds more adequately to the actual
process of semantic interpretation of narrative texts. An interpreter (reader) is pre-
sented with the text in the form of its texture. On the basis of observed texture
regularities, he constructs the principles of sense organization in the form of an
intensional NW. From the intensional NW he derives the extensional primary NW
by following intensional functions from the intensional to the extensional world. In
such a way, the intensional narrative world is the prime target of semantic interpre-
tation, while the extensional world represents the derived ‘background’.
This consideration confirms the primacy of sense in narratives and in literature
in general. Literary texts are, first of all, specific mechanisms for producing new
senses. Excluding the case of obsequious imitation, every narrative text generates
new sense by discovering new possibilities of sense organization through the appli-
cation of new intensional functions. Since the set of possible intensional functions
is open, the possibilities of sense organization are practically unlimited.
The fundamental task of intensional narrative macrosemantics is to uncover and
to describe intensional functions operative in different narrative texts. For the mo-
ment, let us just concentrate on two of these functions which seem to play a funda-
mental role in the sense organization of narrative texts: (a) the authentication func-
tion; (b) the explicitness/implicitness function.

The authentication function

The authentication function is universal: it operates in all narrative texts as a
basic world-forming procedure. The function is closely related to the problem of
truth in narratives and, therefore, we have to turn briefly to this complicated and
controversial topic.
When we inspect the use of the term ‘truth’ in discussions of literature, we see a
rather confusing picture. Many literary critics do not hesitate to use the term quite
frequently, sometimes as a fundamental term of their critical metalanguage. At the
same time, they take it at its face value, without any analysis of its meaning. It is, in
fact, a convenient but meaningless or highly ambiguous label. The task of clarifying
the concept of truth and distinguishing its various meanings is the prime task of the
semantic theory of truth. In its classical formulation (Tarski 1944), semantic theory
defined in a precise way (at least with respect to languages whose structure had
been exactly specified) the concept of truth as was known in the so-called corres-
pondence theory. The essence of Tarski’s definitions can be presented as follows: to
say that a statement (a sentence) is true is equivalent to saying that it designates an
existing state of affairs. This means that sentences (utterances) are “pronounced to
be true or false, on empirical grounds, in the light of the relevant states of affairs
in the world” QVarnock 1963: 741).
It seems that Tarski’s definition of truth (and the whole correspondence theory)
cannot be applied to sentences of literature, at least not without serious modifica-
tions. In fact, those philosophers and literary theoreticians who took a hard look at
 L. DoleZel /Extensional and intensional narrative worlds 205

the problem, expressed rather skeptical opinions. I4 Two varieties of this skepticism
can be distinguished: (a) the Fregean position assumes that literary sentences are
“neither true, nor false” (cf. Gabriel 1970: 1 If.). This position is generally inter-
preted in such a way that a three-value logic is postulated (with values: true; false;
neither true, nor false). Is (b) The classical structuralist position finds no use for
the concept of truth in literary theory: “the question of truthfulness does not make
any sense in poetry” (Mukaiovsky 1948: vol. I, 82). This position corresponds
closely to Austin’s radical solution: there are classes of sentences which cannot
be said to be true or false and, therefore, cannot be called statements; specifi-
cally, sentences which are ‘part of a work of fiction’ are excluded from the class of
statements and from the domain of truth values (Austin 1961: 99).
It seems that after hearing these opinions, we can hardly find a place for the con-
cept of truth of the corresponding theory in literary theory. There is, however, one
aspect of narrative sentences to which the assigment of truth value (in some sense)
seems quite natural; recently, this aspect has been discussed by Pave1 (1976) and,
independently, by Chateaux (forthcoming). According to Chateaux the sentence
Emma BovaTyjinit par se suicider has to be characterized as true, while the sentence
Charles Bovary est unijambiste must be designated as false. Pavel, in fact, makes this
account more precise by stating that ‘true’ or ‘false’ can be said only of ersatz-sen-
tences, i.e. sentences taken outside their literary context. Significantly, both
authors relate the concept of truth to possible-world semantics: statements (atomic
formulae) are designated as true or false in or of a possible world. The first
Chateaux’s example can be said to be true in the ‘fictional’ world of Emma Bovay,
while the second example is false in it.
It is clear that the relativized concept of truth in possible-world semantics is not
equivalent to truth ‘on empirical grounds’. Actual ‘states of affairs’ exist indepen-
dently of the statements (sentences) which can be made about them. In contrast,
alternative possible worlds (including the ‘fictional’ worlds of narrative) are ‘artifi-
cial’ constructs. To say that a proposition is true in or of a possible world and, at
the same time, to say that a possible world is constructed from propositions true in
it, brings us to the brink of circularity.
What is obviously needed is an account of the concept of existence (in a possi-
ble world) without recourse to the concept of truth. I believe that for narrative
semantics such an account can be given by the theory of authentication. The theory
postulates a special intensional function - called authentication function - which
operates in such a way as to assign to narrative sentences either the value ‘authentic’
or the value ‘non-authentic’. By claiming that authentication has the status of inten-

l4 In pronouncements of this kind, we find often the term ‘poetry’; it should be understood in
its broad sense as meaning ‘artistic literature’ (corresponding to the German term Dichtung).
Is In fact, such an interpretation cannot be derived from Fregc (1892); here, Frege insists that
“the truth-value of a sentence . . . is true or false. There are no further truth values” (1892:
123). Literary sentences are without truth-value, since they lack reference.
206 L. Dole51 / Extensiotlal and intcnsional narrative \rorlds

sional function, we imply that it is given by certain regularities of texture and,

further, that authenticity values are a component of the sense of narrative sen-
tences.
Viewing authentication in this framework requires, first of all, specifying the
texture regularities which determine the authenticity values of narrative sentences.
These regularities are to be sought for in specific properties of narrative speech-acts.
It is well known that narrative texts present a rich variety of speech-acts which can
be described by a scale ranging from the ‘objective’ Er-form narrative to the ‘subjec-
tive’ speech-acts of narrative agents. There is no need to describe this scale here,
since I have done so elsewhere (Doleiel 1973). What is important for our topic is
rather the possibility of associating with the scale of narrative speech-acts a corre-
sponding scale of authentication authority. The procedure of authentication is
based on the fact that the ‘cbjective’ Er-form narrative has the highest, absolute
authentication authority; on the other end of the scale, ‘subjective’ speech-acts of
narrative agents show the lowest degree of authentication authority. l6 The basic
rule of authentication in narrative texts can be formulated as follows: narrative sen-
tences stated in the ‘objective’ Er-form narrative are automatically assigned the
value ‘authentic’; sentences stated in other narrative speech-acts are ‘non-authentic’,
unless they are authenticated by the Er-form narrative or by other specific
devices. ”
I do not want to present a detailed discussion of the various possible authentica-
tion devices; it should be clear from our outline that these devices are given in the
texture, specifically in the forms of narrative speech-acts conveying varying degrees
of authentication authority. This justifies our claim that authenticity values are
intensional concepts. By the operation of the authentication function, the inten-
sional narrative world is split into authentic and non-authentic subworlds. This par-
titioning reveals the organization of narrative sense, resulting from the application
of the authentication function.
The authenticity values of narrative sentences as components of sense are deter-
mined by the texture and, therefore, are inseparable from it. On the other hand, the
ersarz-sentences which provoked our discussion are obviously independent of tex-

I6 ‘Authentication authority’ is analogous to the authority of a speaker uttering a performative

speech-act; thus, for example, a ship can be named only by “the person properly authorized to
name it” (Austin 1963: 14). Following Austin, we can say that only the anonymous source of
the Er-form narrative is ‘properly authorized’ to make authentic statements. Note that in both
cases, the authority of the speaker is based on convention: in Austin’s example on social con-
vention, in narrative speech-acts on artistic convention.
” In our discussion, the problems of authentication in Ich-form narratives are left aside; they
require a special consideration. It seems obvious that the degree of the authentication authority
of the Ich-form narrator is different from that of the &-form narrative. At the same time, how-
ever, it should not be reduced to the degree of authentication authority of narrative agents; the
lch-narrator obviously has a higher degree of authentication authority than the agents of his
story.
 L. Dolefell Extensional and intensional narrative worlds 201

ture, they are paraphrases expressing extensional meaning. However, we are now in
a position to explain why the ersatz-sentence Emma Bovavfinished by suicide can
be proclaimed ‘true’, while the ersatz-sentence Charles Bovary was one-legged has to
be designated as ‘false’. The reason is that the first ersatz-sentence is a paraphrase of
authentic narrative sentences, while the second is not (to be exact, the second
ersatz-sentence has no equivalent narrative sentence in Flaubert’s text). The assign-
ment of truth-values to ersatz-sentences is made on the basis of authenticity values
of corresponding narrative sentences, as given in the texture. ‘Truth’ in narratives is
derived from ‘authenticity’ and, as such, it has a clearly different meaning than the
concept of truth in the correspondence theory. In the final account, the concept of
truth in narrative semantics seems redundant and dispensable.

Explicitness/implicitness function

The importance of the explicitness/implicitness distinction has always been

recognized in linguistic and literary studies; however, only recently have some
approaches to the problem been formulated (cf. Ducrot 1972; Hausenblas 1972). In
this paper I propose to treat explicitness/implicitness as an intensional function (E/I
function). Probably no one will deny that explicitness and implicitness are two con-
trary forms of expression (texture); the concept of the E/I function makes it pos-
sible to incorporate the distinction into intensional narrative semantics.
The E/I function, we assume, takes entities of the primary extensional NW as
arguments and assigns to each of them either an explicit or an implicit form of
expression. This description of the operation of the function corresponds exactly to
fig. 1. We should emphasize that any entity of the extensional world can be made
explicit or implicit; our discussion will, however, focus on the category of motifs
(narrative predications), since this case is both the most common and the most
complex. Since motif is a predication, we will say that it is expressed explicitly if
and only if it is expressed in the texture by a complete simple sentence. ‘* The
essence of the implicit form of expression is more difficult to capture. We must
assume that a motif is expressed implicitly, if: (a) it is not assigned a complete sim-
ple sentence; (b) it is nevertheless recoverable from the texture. The seeming con-
tradiction of these requirements is resolved by the technique of allusion. Allusion,
we assume, is a device of implicitly signalling a motif in the texture. I9
Allusion is an expression which is semantically defective; it creates a semantic
‘gap’, a ‘gap’ which has to be filled by the recall of the presuppositions of the sen-

*a The formulation indicates that besides explicitness and implicitness, a redundant (more than
explicit) form of expression is possible.
t9 We take the term ‘allusion’ in its broadest sense, corresponding roughly to ‘hint’; in poetics,
the term has usually been used more narrowly, designating the semantic links of one literary
text to other texts, art works, etc. (cf. Preminger 1974’: 18).
208 L. Dolr?el /Extensional and intensional narratiw worlds

tence in which the allusion is given. Allusion triggers the activation of its presuppo-
sitions, or, more exactly, the activation of the presuppositions of the sentence in
which the allusion is given. These activated presuppositions, not stated in the tex-
ture, but necessitated by it, represent the missing form of the motifs expression.
While allusion is part of the texture, its presuppositions are not. In such a way, the
seemingly contradictory conditions of implicitness, formulated earlier, are recon-
ciled: while the motif is not expressed in the texture, it is recoverable from it due
to the recall of presuppositions, triggered by the allusion given in the texture.
Since we assume that presuppositions play an essential role in the implicit forms
of texture, we need a clear concept of presuppositions. Unfortunately. in spite of
the many discussions which have been devoted to it, the concept is muddled (for a
recent survey of various approaches, see Lyons 1977: 592-606, vol. II). In order to
make the concept useful for narrative semantics, two requirements should be ob-
served: (a) presupposition should be clearly distinguished from entailment and
other possible implication relations between sentences. The best way to do it is to
respect the original logico-semantic definition of presupposition. a0 For the
purposes of narrative semantics, the criterion of presupposition can be formulated
as follows: a set of sentences which is necessitated by a narrative sentence repre-
sents the presuppositions of this sentence if and only if the set is the same for the
asserted and for the negated sentence. (b) Presuppositions should be postulated as
objective (intersubjective) semantic facts; every competent speaker has the ability
to recall the presuppositions of sentences expressed in his language. Ducrot’s fine
distinction between the presuppost of a texts and its sowentendu (the former
being a fact of langue, the latter of parole) (Ducrot 1972) reflects the postulate of
objectivity. As objective semantic facts, presuppositions can enter the formation of
the objective text sense.
In order to make our discussion more perspicuous, let us invoke one example of
the allusion technique. In Kafka’s The Trial the arrest of Joseph K. is announced in
the following performative act of one of the warders: “You are arrested. . . . BYI-
ceedings have been instituted against you”. Moreover, the warder speaks about his
authorization and his instructions. Significantly, the name of the agent(s) who gave
the authorization and instructions and who ordered Josef K.‘s arrest is witheld by
consistent manipulations of the texture. Consequently, a conspicuous semantic gap
is created. The sentences of the warder function as allusions; their full semantic
interpretation requires the recall of their presuppositions. These presuppositions
can be formulated as follows: there exists an agent (or a group of agents) different
from the warder and he (they) ordered Joseph K. ‘s arrest. These presuppositions
represent the implicit expression of the corresponding motifs of Kafka’s The Tt-i>l.
We shall keep this example in mind when we proceed to consider briefly the con-

2o The definition is given by van Fraassen: “A presupposes B if and only if: (a) A necessitates
B, (b) not-A necessitates B.” The semantic relation of necessitation means that the argument
from A to B (and, of course, also from not-A to B) is valid (van Fraassen 1968: 138).
 L. DoleIel / Extensional and intensional narrative worlds 209

sequences of the E/I function for the organization of sense in narrative texts. Obvi-
ously, the E/I function splits the intensional NW into an explicit and an implicit
subworld; this partitioning is the base of the organization of sense by the applica-
tion of the E/I function. Before we say something more specific about this sense
organization, we have to consider the possibility that the E/I function, unlike the
authentication function, is not universal. It seems that the E/I function is applied
rather unevenly in different narrative texts: many texts prefer to express all (or
practically all) of their motifs explicitly; others are full of allusions and implicit mo-
tifs. This empirical fact gives us a reason to believe that also the consequences of
the E/I function for the organization of sense are text (or text-class) specific, rather
than universal. Such differences could be related to the different degrees of impor-
tance of the function and/or to its ‘teaming’ with other intensional functions. Thus
in some narrative texts, the explicit and the implicit intensional subworlds provide a
base for differentiating the ‘foreground’ and the ‘background’ of the story. In some
modern texts, such as in Kafka’s The Trial and The Castle, the organization of the
intensional world can be described as an opposition between the ‘visible’ and the
‘invisible’ worlds (cf. also Doleiel forthcoming). While the explicit subworld pro-
vides a base of the ‘invisible’ world, the implicit subworld, due to its vagueness and
indefiniteness, is a natural base of the ‘invisible’ world. If our interpretation is, in
principle, correct, then the sense organization effected by the E/I function might be
one of the most interesting phenomena of modem fiction.
If we assume that such semantic macrostructures as ‘visible’ and ‘invisible’
worlds are intensional phenomena, then they are, by definition, determined by the
texture. Any change of the texture affecting the E/I function will change the char-
acter and the relation of the two worlds. If the changes of texture were so radical as
to eliminate the E/I function (if everything was expressed explicitly), the ‘invisible’
world would be destroyed. This reflection leads us back to the initial and basic idea
of our paper: the inextricable link between form and meaning which is made
explicit in the theory of sense represents the most important problem of narrative
semantics. However, intensional semantics presupposes a well-established exten-
sional semantics; consequently, both the intensional and the extensional semantics
of narrative texts require systematic attention.
The difficulties of intensional semantics are well known; they are given by the
fact that sense cannot be adequately formulated in a paraphrase. The difficulties
have been obvious in the domain of literary micro-semantics, in the study of such
intensional phenomena as metaphor, symbol, etc.; they increase substantially when
we proceed to macro-structural intensional semantics, to the study of global princi-
ples of sense organization. Here, we are faced with sense constituents en masse and
with unlimited possibilies of their organization. Obviously, such problems require a
theoretical treatment in terms of sets and functional relations between sets. Our
concepts of intensional narrative world and of intensional function are based - as
can easily be seen - on these formal notions. If the perennial problem of ‘truth’ in
literature 9-4 the important contrast between explicitness and implicitness has
received some elucidation within this theoretical framework, our attempt to link
problems of narrative semantics to those of logical semantics has not been in vain.

References

Adams, Robert M. 1974. Theories of actuality. NoCs 8: 21 l-23 1.

Austin, J.L. 1961. Philosophical papers. Ohford: Clarendon Press.
Austin, J.L. 1962. How to do things with words. Cambridge, Mass.: Harvard University Press.
Austin, J.L. 1963. ‘Performative-constative’. In: Charles W. Caton, ed. Philosophy and ordinary
language. University of Illinois Press. Quoted from: J.R. Searlc, ed. The philosophy of lan-
guage (1971). London: Oxford University Press, 13-22.
Brooks, Cleanth. 1947. The well wrought urn. New York: Harcourt, Brace and \Vorld.
Chateaus, Dominique. Forthcoming. La simantique du r&it. To appear in Semiotica..
Dijk, Teun A. van. 1974/75. Action, action description, and narrative. New Literary History 6:
273-294.
Dolcicl, Lubomir. 1973. Narrative modes in Czech literature. Toronto: University of Toronto
Press.
Dolcicl, Lubomir. 1976. Narrative semantics. PTL 1: 129-151.
Doleicl, Lubomir. Forthcoming. The visible and the invisible Petersburg.
Ducrot, Oswald. 1972. Dire et ne pas dirt. Paris: Hermann.
Dummett, Michael. 1973. Frege. Philosophy of language. London: Duckworth.
Eng, Jan van der. 1978. ‘On descriptive narrative poetics’. In: Dutch studies in Russian litera-
ture, vol. 4. Lissc: Peter de Ridder Press, 9-94.
Fraassen, Bas C. van. 1968. Presupposition, implication, and self-rcfercnce. The Journal of Phi-
losophy 6.5: 136- 152.
Frege, Gottlob. 1892. uber Sinn und Bedeutung. Zeitschrift fiir Philosophie und philosophische
Kritik 100: 25-50. English transl. quoted from Zabeeh-Klemke-Jacobson 1974: 118-140.
Gabriel, Gottfried. 1970. G. Frege fiber semantische Eigenschaftcn der Dichtun:. Linguistische
Berichte 8: 10-17.
Creirnas, A.J. 1966. SGmantique structuralc. Rechcrchc dc m&thodc. Paris: Laroussc.
Greimas, A.J. 1970. Du sens: essais sCmiotiqucs. Paris: Seuil.
Grossmann, Reinhardt. 1969. Reflections on Fregc’s philosophy. Evanston: I\;orth\vestern Uni-
versity Press.
Hauscnblas, Karel. 1972. Iixplicitnost a implicitnost jazykovdho vyjadiovini. Slovo a slovesnost
33: 98-105.
tlintikka, Jaakko. 1973. ‘Grammar and logic: some borderline problems’. In: Hintikka,
Moravcsik and Suppcs, eds., Approaches to natural language. Dordrecht: Reidcl, 197-214.
Hintikka, Jaakko. 1975. The intentions of intcntionality and other new models for modalities.
Dordrccht: Rcidel.
Kempson, Ruth M. 1977. Semantic theory. Cambridge: Cambridge University Press.
Kripke, Saul A. 1963. Semantical considerations on modal logic. Acta Philosophica Fennica
XVI: 83-94. Quoted from: Zabcch, Klemkc and Jacobson 1974: 803-814.
Kripkc, Saul, A. 1972. ‘Naming and necessity’. In: D. Davidson and C. Harman, eds., Semantics
of natural language. Dordrecht: Reidel, 253-35.5.
Lyons. John. 1977. Semantics. Vols. I and Il. Cambridge: Cambridge University Press.
Mates, B. 1968. ‘Leibniz on possible worlds’. In: B. van Rootscllar and J.F. Steal, cds.. Logic,
methodology and philosophy of science, vol. III. Amsterdam: North-Holland, 507-529.
SlukaiovskC, Jan. 1948. Kapitoly z zeski poctiky. Prague: Svobods. English transl. quoted
from J. Burbank and P. Steiner, eds. and transls., The word and verbal art. New Haven and
London: Yale University Press (1977).
 L. Dolcz’cl /Extensional and intensiorlal narrative wor!ds 211

Pavcl, Thomas C. 1976. ‘Possible worlds in literary semantics. The Journal of Acsthctics and
Art Criticism 34 (2): 165-176.
Prcminger, Alex. Ed. 1974’. Princeton encyclopedia of poetry and poetics. Princeton, N.J.:
Princeton University Press.
Propp, Vladimir. 1928. Morfologija skazki. MOSCOW.
Tarski, Alfred. 1944. The semantic conception of truth and the foundations of semantics. Phi-
losophy and Phcnomcnological Research IV: 341-375. Quoted from Zabeeh. Klcmkc and
Jacobson, 677-711.
Tesnicrc, Lucicn. 1959. I:IL:mcnts de syntaxe structuralr. Paris: Klincksicck.
warnock, C.J. 1963. ‘Truth and corrcspondencc’. In: C.D. Rollins, ed., Knowlcdgc and capcri-
encc. Pittsburgh: University of Pittsburg Press, 11-20. Quoted from Zabeeh, KIcmkc and
Jacobson, 733-742.
Zabceh, Farhang, li.D. Klemke and Arthur Jacobson, eds. 1974. Readings in semantics. Urbana-
Chicago-London: University of Illinois Press.

Lzrbonrir Doleiel, formerly a rcscarch fellow of the Czechoslovak Academy of Sciences, is pro-
fessor at the University of Toronto in the Dcpartmcnt of Slavic Languages and Literatures. He
has published numerous works in stylistics, poetics and statistical linguistics. Among his recent
publications in narrative theory arc: Narrative modes in Czech liferanrre (Toronto. 1973); ‘Nar-
rative semantics’ and “Die Hundcblumc’, or poetic narrative’ (both in PTL 1976); ‘Narrative
modalities’ (Journal of Literary Semantics 1976).