1

CAUSES AND AFFECTS

PAUL J. HOPPER
Chicago Linguistic Society
Special Session on Agency and Causativity
April 1986

1. Introduction

A long-standing informal characterization of the Object relationship in which a

noun phrase can stand to a verb involves the semantic notion of affectedness. The
Object of the verb, if all else fails, can be designated as the participant which is
altered, moved, or in some other way affected by the verb's action. Examples of this
semantic relationship are not hard to construct: pushed the cart, cracked the nut, and
so on. In such predicates, an object is impinged upon by an action, and is a Patient;
the clause itself is close to being cardinally, or prototypically, transitive.
The notion of a prototypically transitive clause stems from the work of Lakoff
(l977) and Hopper and Thompson (1980). In the latter work, a number of specific
parameters which appeared universally to condition cardinal, i.e. prototypical,
transitivity were hypothesized. Clauses tend to be marked grammatically as transitive
to the degree that they are positive for the following features (Hopper and Thompson
1980: 252):

1. Two noun phrase participants, an agent and an object;

2. Kinesis, denoting an action or event;
3. Telic aspect, i.e. sense of a completed goal;
4. Punctuality, denoting a sudden action;
5. Volitionality, denoting a deliberate action;
6. Affirmation--an affirmative clause;
7. Modality--the clause is in the realis mode;
8. Agency--the agent is human or otherwise autonomous;
9. Object affectedness--the object is changed in some way;
10. Object individuation, i.e. definiteness-referentiality

The “degree of affectedness” of the object of the verb is perhaps one of the more
important parameters by which prototypical transitivity is measured. In our pre-
theoretical understanding of “transitivity”, the impingement of the action on an Object
is the initial criterion by which we judge this relationship, and is probably central to a
preliminary definition of it (cf. Lyons 1968: 359), even though facts about grammar
will later require the extension of this criterion to other types of relationship in
particular languages. For this semantic relationship of verb to object, then, an
observation of traditional grammar that some objects of action verbs are not changed
but are instead brought about by the action of the verb causes some difficulties.
Under the heading “Objects of Result”, Jespersen (1933: 109) contrasts

(1) She lights the lamp

with
(2) She lights the fire

In the same section Jefferson points out that numerous other semantic relationships
of Objects to Verbs exist which vitiate any attempt to assign a unitary, invariant
semantic property to the verb-object relationship. A somewhat fuller discussion of
Effective Objects is found in Quirk and Greenbaum (1973: 174-75), where an Effected
Object is defined as one that “refers to something which exists only by virtue of the
activity indicated by the verb.” Like Jespersen, Quirk and Greenbaum include in the
category of Effected Objects also “cognate” objects, those which repeat the meaning of
the verb, as in sing a song. They point out (426-27), moreover, that phrasal verbs like
make an attempt, have/take a bath, give a kick also belong in this general semantic
category. The same theme was taken up by Fillmore (1968), who examined a number
of the different semantic relationships between verbs and objects with a view to
positing deep semantic “cases” as universal primes. One of these cases, the Factitive,
is the case for resultative, or what I shall here, following the more established usage,
call Effected objects. Like Jespersen and Quirk and Greenbaum, Fillmore includes
among the uses of the Factitive the “cognate object”, as in

(3) I dreamt a curious dream. (Jespersen 1933: 109)

And in fact the actual examples examined by Fillmore (1968: 85-86) involve cognate
objects rather than true Effected (resultative) objects. Moreover, the Factitive had a
rather brief life-span in the history of Case Grammar, being replaced almost
immediately by the vaguer term “Goal”, which included all kinds of things which had
previously been thought of as Objects.
Beyond English, we find that the distinction of so-called “inner” and “outer” objects
was often made in descriptive grammars of the Classical languages. An inner object is
one which is an extension of the verb, as opposed to an outer object, which is altered or
affected by the verb. Thus Bennett (1966[1914]: 195-238) distinguishes five basic uses
of the accusative case: the Inner Object, the Thing or Person Affected, Duration of
Time, Extent of Space, and Limit of Motion. If we combine the latter three categories
into a single one of something like “spatio-temporal limit”, this leaves the ordinary
substantival objects in the two large classes which I am calling those Effected and
those Affected. It is not clear, however, that any specific grammatical correlates of the
distinction were ever described. It is noteworthy that discussion of this topic is
invariably in the context of the “meaning” of the Accusative case, not in the context of
the characterization of the notion Object.
Lyons in his book on Semantics (Lyons 1977) suggests a semantic typology of
causation and transitivity in which three major categories of relationships between
causers and effects (“valency schemata” [Lyons, 1977, vol. II: 491]) exist. These are:
I. AFFECT (AGENT, PATIENT) -- “Operative”
II. PRODUCE (CAUSE, EFFECT) -- “Factitive”
III. PRODUCE (AGENT, EFFECT) - - “Operative-factitive”
In some important discussion (490-491), Lyons observes that the product of a factitive
expression cannot properly be called a semantic Patient. He also links factitives with
causatives, so that to this extent specific grammatical correlates of “Effects” can be
inferred; but Lyons does not suggest any grammatical distinctions between noun
objects which are Patients and noun objects which are Products.
 More recently, Carol Lord (1982: 284-285) has discussed the Effective/Affective
distinction in serial verb clauses in a group of West African languages, placing the
distinction into the context of the Transitivity Hypothesis proposed by Hopper and
Thompson (1980). The present paper owes much to Lord's insightful analysis.
To my knowledge there has been no cross-linguistic study of the realization of the
potential contrast between Affected and Effected objects, and both the traditional
notion of “Object of Result” and Fillmore's "Factitive Case” remain purely semantic
concepts whose primary grammatical relevance has never been demonstrated for any
language. Yet the topic seems important enough to warrant a treatment. For example,
it is clear that resultative objects present a problem for anyone wishing to characterize
the universal notion of Object. They cannot be said to “undergo” the action of the verb,
and therefore cannot be described as Patients. On the other hand, they share certain
grammatical properties which have been said to be characteristic of Objects, the most
obvious being their typical location in the syntactic slot reserved for objects. In this
paper I will discuss some of the grammatical features which appear to distinguish
Effected Objects from Affected Objects in English and a few other languages, and will
discuss the phenomenon from a discourse perspective in Malay. These objects will be
referred to as E-objects and A-objects; I will also use the self-explanatory terms E-verb
and A-verb, E-clause and A-clause; and the notation E* (or A*) is to be read as
“ungrammatical if understood Effectively (or Affectively)”.

2. English

2.1 Phrasal Verbs with “up”.

In English it is quite often the case that pairs of verbs with and without particles are
differentiated, among other things, by transitivity. For example:

(4) He cut the meat.

(5) He cut up the meat.

In (5) the object is totally affected by the action of the verb, whereas in (4) no such
claim is made, but the meat could be only partially cut. This semantic difference of
fully vs. partially affected is quite commonly associated with transitive vs. intransitive
morphology or syntax (Hopper and Thompson 1980: 262-263).
When the action of cutting results in something, the phrasal verb with up is
inappropriate:

(6) He cut a slice of cake.

(7) E* He cut up a slice of cake.

Similarly,

(8) He carved a statue.

(9) E* He carved up a statue.

This difference between phrasal and non-phrasal verbs is not a consistent one, and it
is sensitive to individual verbs and particles. The particle up seems especially to have
an affinity with A-objects; note, for example,
(10) They built some factories.
(11) They have now built up our entire neighborhood (with factories).

The claim which I am making, then, is that in minimal pair situations involving
simplex and phrasal verb with up, if one member of the pair is affective and the other
effective, it will always be the phrasal verb which is Affective and the simplex which is
Effective. This rather modest claim supports the idea of a transitivity difference
between E-clauses and A-clauses.

2.2 “Wording-Up”.

Another set of data has to do with some phenomena which J. R. Ross (1974) has called
“wording up”. Ross notes that adding prefixes to English verbs results in new verbs
which differ in certain predictable ways from the corresponding unprefixed forms.
Ross points out that typically an unprefixed verb differs from a prefixed verb in
requiring a more concrete noun object, and in requiring non-stative predicates. A good
example of this is the following. The verb “weigh” can be followed either by a measure
or a concrete object:

(12) He weighed 100 kgs.

(13) He weighed the bananas.

But when prefixed with re- the measure predicate is no longer possible:

(14) He re-weighed the bananas.

(15) *He re-weighed 100 kgs.

To these observations I would add the following: Prefixed verbs in English typically
require a definite-referential object. Not only do sentences like the following sound
strange:

(16) ? He re-weighed bananas,

but so-called "referentially opaque” clauses lose their ambiguity if the verb is prefixed:

(17) Sheila wants to marry an Australian.

(18) Sheila wants to re-marry an Australian.

That is, in (18) a particular Australian is referred to, while in (17) there is ambiguity
as to whether a specific Australian or any eligible member of the class of Australians
is referred to. Since the referentiality of the Object is an important criterion of
Transitivity, it seems indisputable that prefixes like re-, mis-, and so on increase
transitivity. This being so, it is interesting to note that putting a prefix on an E-verb
often results either in a strange meaning or an Affective one:

(19)E* He re-wrote a poem.

In sentences of this kind it is virtually impossible to get an E-meaning. Furthermore,
if the verb is ambiguous between an E-meaning and an A-meaning, prefixing it
removes the ambiguity in favor of the A-meaning:

(20) He painted a house [E or A possible],

(21) E* He re-painted a house [only A possible].

This A-meaning in turn has much to do with the referentiality of the Object. There is a
sense in which the patient of a prefixed verb is "already there", and this sense is not
restricted, as one might think, to the prefix re-, which of course suggests a repeated
action. Thus in:

(22) He tied a knot,

the knot is absent until it is created by the action of tying, but in

(23) He untied a knot

the knot must already be present in order for it to be untied. In one case (the A-clause)
the indefinite object may be referential and in the other (the E-clause) the Indefinite
object cannot be referential. Not surprisingly, prefixes and particles behave alike, in
that the object of (e.g.) “tie up” must also be affected:

(24) He tied up the boat.

(25) E* He tied up the knot.

2.3 Cleft Sentences.

A similar problem exists with the English Cleft Sentence. G. Lakoff pointed out that
sentences like the following:

(26) What he did to his autobiography was tear it up

are not possible with E-verbs:

(27) *What he did to his autobiography was write it.

(See also Quirk and Greenbaum 1973:173). It should also be noted that restrictions on
“cleft sentences” are really restrictions on the complements of the verb do to, so that
the following discourses are also anomalous:

(28) *What did he do to his autobiography? - He wrote it!

(29) *He did something terrible to his autobiography. He wrote it.

In what follows, only “cleft sentences” will be discussed. Here again the crucial
parameter may be characterized as one of Transitivity. The propositional meaning of a
cleft clause involves an agent carrying out an affective action on a specific concrete
object, and clefts are questionable precisely to the extent that these properties are
absent. Hence:
(30) (a) What he did to the princess was change her into a frog.
(b) ? What he did to a stereo set was short-circuit it.
(c) ?? What he did to a newspaper was burn one.
(d) ?? What he did to rice was steam some.
(e) ?? What he did to happiness was increase some.
(f) ?? What he did to misery was imagine some.

Examples such as these suggest that the further a clause moves from prototypical
transitivity, the less acceptable it becomes in a clefted version. Hence the behavior of
E-clauses under clefting is additional evidence for the hypothesis that E-clauses are
less transitive than A-clauses.

2.4 Enclitic Benefactives.

Effective clauses have an affinity for benefactive objects which appear as unstressed
nouns or pronouns immediately after the verb. Such objects refer to the person on
whose behalf or for whose benefit the product of the action is created:

(31) I’ll fix you some crepes.

(32) I’ll draw you a nice choo-choo train.

This affinity is a fairly strong one. For example, it has often been pointed out that
ordinary transitive A-clauses sound strange with an enclitic benefactive:

(33) ? He repaired me the motorbike.

Even more telling is the fact that the ambiguity between the E-meaning and A-
meaning in verbs like to paint is cancelled in favor of the E-meaning if a benefactive is
present:

(34) A* I’ll paint you a house.

Again, it seems that there is evidence for a lessened degree of transitivity in all
sentences in which enclitic benefactives appear. It is not just that they occur
preferentially with E-objects, but that prototypical sorts of objects, i.e. those which are
specific and affected, are virtually excluded from co-occurring with them. Hence in
sentences like

(35) I’ll call you a taxi,

(36) I’ll buy you some new shoes

the Objects are not E-Objects, but are nonetheless non-referential and cannot be said
to be materially changed by the action of the verb, but are merely viewed as being
relocated in special ways. Even clauses containing referential Objects are grammatical
with benefactives provided the Object is not affected in the strong sense of “physically
impinged upon or altered”:

(37) Harry bought me this necklace.

To sum up, the distribution of enclitic benefactives provides some further support for
the alignment of E-clauses with less than prototypically transitive clauses.

2.5 Verbs of Saying.

Speech act verbs comprise perhaps the largest and most coherent set of E-verbs in
English and presumably any language. It has been argued by Pamela Munro (1982:
301-318) that clauses directly reporting an act of saying, i.e. quotatives, are
universally coded in some way as less transitive than other kinds of verbs. The same
is true, perhaps to a lesser extent, of indirect speech reports. As regards English,
Munro notes the following phenomena which point to decreased transitivity:
A. Quotations are not readily passivized.
B. The syntax of quotations may resemble that of intransitive rather than transitive
constructions. Thus, for example, inversion of subject and verb is possible with various
kinds of intransitives, but impossible with ordinary transitives.
C. In very colloquial English, perhaps restricted to juvenile idiom, the (intransitive)
verb go serves as a substitute for say. Used in this way, go supports only direct
quotation, never indirect; cf. the following pair of examples, suggested by Munro (306):

(38) And then Mary went/goes, "Well, you'll have to come with us.
(39) *And then Mary went that we'd have to come too.

D. Not mentioned by Munro is the comparable behavior of the (archaicizing) English

verb quoth, a frozen past tense form which can have only direct speech as its
complement:

(40)(a) “Lackaday!” quoth Robin, “I prithee turn down thy stereo.”

(b) * Alan a Dale quoth that Friar Tuck was a varlet.

There are, then, compelling reasons for assuming (1) that verbs of sayling are less
than prototypically transitive, and (2) that the more a verb of saying is restricted in its
complement to direct quotation, the further it is from the transitive center. Again
there is a correlation between Effectivity and Intransitivity.

3. The Affected-Effected Distinction in Universal Grammar

3.1 Introduction.

It is a commonplace in the study of Universal Grammar to find that distinctions which

are manifested in peripheral or purely quantitative ways in one language are
represented in other languages by grammaticized morphology or syntax. Having seen
that in English those areas of grammar and lexicon which involve the
Affected/Effected distinction show E-clauses to be less transitive than A-clauses, there
is an obvious motivation to examine other languages and discover whether a
comparable correlation can be established. The following is a rather random set of
observations which suggest the widespread nature of the relevance of Transitivity to
the Effective/Affective distinction.

3.2 Serial Verb Languages

There is by now a considerable literature on languages which permit multi-verb
propositions whose arguments are distributed singly over the verbs. The article by
Carol Lord (Lord 1982) referred to above has examined such constructions in a group
of West African languages, the Kra group, from the perspective of the Transitivity
Hypothesis; this article also explicitly raises the question of the Affective/Effective
distinction, and must be credited with being the first to recognize the relevance of the
Transitivity Hypothesis to the distinction. In this group of languages a verb meaning
“take” (not etymologically relatable from language to language) functions as a
preposed object marker in constructions such as:

(41) Idoma [Ghana] o l uwa nu

she take them drive-away
“She drove them away”

Some languages, such as Idoma, are indifferent to the effected/affected object

distinction, and permit the serial verb transitive construction with “take” for both
types:

(42) Idoma o l Oyi ma

she take child bear
“She bore a child”

In other languages of the same group, however, only Affected objects may participate
in the "take” serial verb construction. Thus in Akan [Ghana], A-objects may appear
with de “take”:

(43) Akan Kofi de nwoma no a - ba

Kofl take book that PF come
“Kofi brought that book”

With E-objects, however, the “take”' construction is not possible, i.e. the equivalent of
“He take letter write” in the sense of “He wrote the letter” is not possible. Such
messages must instead be expressed in the less transitive mode, with a single verb:

(44) w - a - kyerEw me nhoma

he PF write me letter
“He wrote me a letter”

Similarly, in Ga [Ghana]:

(45) Ga e kE wolo ngmesi

she take book put down
“She put the book down”,

but:

(46) *e kE wOlO ngme

she take egg lay
“She laid an egg”

In her article on Transitivity in these languages, Lord shows that it is the “take”
construction which is to be regarded as the more highly transitive one. The diachronic
development is to be seen as one from a concrete meaning of “take” to a more general
use of “take” in which this word becomes a marker of all direct objects, whether
strictly “takable” or not. It is this change which accounts for the semantic restrictions
on the “take” serial verb construction. These semantic restrictions include more than
the Effected/Affected dimension; in Dagbani [Ghana], for example, clauses in which
the object is not movable are excluded from the “take” construction; we can have the
equivalent of “take truck paint” for “paint the truck”, but not “take room paint” for
“paint the room.”
Lord's discussion suggests that in none of these languages is the Affected/Effected
dimension the only parameter for determining whether an object marker appears. On
the one hand there are “mature” serializing languages in which a concrete verb “take”
has been bleached into a grammatical marker of all direct objects; this type is
represented by Idoma. At the other extreme are languages like Dagbani, in which the
concrete meaning of “take” is still alive, and only physically manipulable objects are
marked. The restrictions on object marking in the latter case are to be seen as
semantic restrictions on the action of “taking”, which can only have as its patient
objects which are physically present and are movable. In Dagbani and languages like
it, objects which are not yet in existence do not qualify as patients of “take” verbs. Yet
in the “selective” group of languages which includes Dagbani and Ga it remains a fact
that objects which are the product of the verb may not participate in the serial verb
construction with “take”. Clauses with E-objects are thus in these languages to be
viewed as less transitive than clauses with A-objects.

3.3 Chinese.

It is worth noting that the Chinese ba construction (cf. Hopper and Thompson 1980:
274-275), which is closely comparable in both syntax and semantics to these West
African serializing constructions (Lord 1982), also appears to be incompatible with an
Effective meaning; the same is true of the “passive” in bei, as illustrated by these
examples from Chao (1968: 706; the romanization is Chao's):

(47) (a) Hwunan chu-mii “Hunan produces rice.”

(b) *Hwunan bae mii chu-le “(...take rice produce)”
(48) (a) Woo fa-shau “I have, develop a fever”
(b) *Shau bey fa gau-le “(A fever is developed by me)”
(c) *Woo bae shau fa gau-le “(I take fever develop)”

3.4 German Transitive Prefixes.

For certain German verbs, prefixed and unprefixed forms contrast as Affective and
Effective. A clear example of this is the verb malen “to paint”:

(49) Er malt ein Bild “He’s painting a picture”

(50) Er bemalt das Bild “He paints over the picture”
Here, (50) might suggest an act of defacing an already existent picture, though this is
not necessarily part of its meaning. With zeichnen “to draw”, the prefixed form
diverges in meaning:

(51) (a) Er zeichnete die Burg “He was drawing the castle”
(b) Er bezeichnete die Burg als ein Gefängnis “He described the castle as a jail”

With the speech act verbs sagen “to say” and antworten “to answer”, only the
unprefixed forms may be used with a quotative object:

(52) Er sagte ‘Du bist dran.’ “He said, ‘It’s your turn.’”
(53) Er antwortete ‘Keine’ “He answered, ‘None.’”

The prefixed form beantworten requires a noun object:

(54) Er beantwortete die Frage “He answered the question.”

The uses of the prefixed form of sagen are very restricted. By far the commonest use is
in the construction which is equivalent to English “that is not to say that...”:

(55) Das besagt allerdings nicht, dass alle Studenten faul sind.
“That is not to say, of course, that all students are lazy.”

This affinity with a negated clause is particularly interesting, in that in order to

explicitly deny the validity of a statement it must necessarily be presupposed that the
statement has already been made. Consequently the complement clause in (55) is not
a result of the verb besagen; rather, besagen comments on a statement which has
already been made and qualifies it.
The prefix be- frequently serves as a marker of increased valency. The following are a
few of many examples:

(56) wohnen “dwell, live” bewohnen “inhabit”

gehen “go” begehen “commit (an offense)”
schwören "swear" beschwören “enchant, exorcise”

and so on. Moreover, already transitive verbs, when prefixed with be-, may carry the
extra meaning that the object is completely or more drastically affected:

(57) (a) Die Artillerie schoss viele Granaten

“The artillery shot (fired) many shells”
(b) Die Artillerie beschoss das Dorf
“The artillery shot up (poured fire onto) the village”

Such additive meanings are, of course, very characteristic of the whole phenomenon of
Transitivity in the extended sense. They contribute towards the view that in German
effective actions are aligned with the less transitive side of the transitivity spectrum,
in accordance with the general tendency.

3.5 Hindi Compound Verbs.

The data from Hindi cited here are taken from Peter Hook's monograph on the
Compound Verb in Hindi (Hook 1974). The following sentences illustrate transitive
sentences in which the main verb is combined with one of the auxiliaries (lenaa) or
(denaa):

(58) mAI ne das baje aap ko fon kar liyaa

I AGT 10 o'clock you: DAT phone make AUX
“I telephoned you at 10 o'clock.” (166)

(59) mAI ne use paise de diye

I AGT him: DAT money give AUX
“I gave him the money” (167)

In (58) the auxiliary lenaa forms a compound verb with the main verb kar “make”. In
(59) the auxiliary is denaa, and the main verb, meaning “to give”, is also denaa. There
is much evidence that the additive meaning imparted to the verb phrase by these
auxiliaries is one of Transitivity, very much in the expanded sense of this word
developed by Thompson and Hopper (1980). For example, in the two sentences cited,
there is agreement among speakers that the action is successfully and fully completed.
In (58), the implication is that the telephone call was actually made, and a
continuation that the speaker attempted the call but could not get through, for
example, would not be possible. In (59), native speakers of Hindi were fully agreed
about its grammaticality provided all the money was given, but their judgments were
divided if the context supplied suggested that some but not all of the money was given.
It is interesting to find that Effective sentences sound strange when used with
compound verbs of the V-lenaa or V-denaa type. Consider, for example:

(60) rabindranaath ne janagaNamana likhaa

(name) AGT (name of poem) wrote
"Rabindranath wrote ‘JanagaNamana’“ (241-2)

JanagaNamana is the name of a poem which later became the Indian National
Anthem. Confronted with versions of this sentence containing a compound verb,
speakers’ reactions became uncertain:

(61) rabindranaath ne JanagaNamana likh liyaa

(name) AGT (name of poem) write AUX

Some accept this sentence and others do not. The reason for the uncertainty appears
to lie in the question of whether or not the performer of the utterance knew in advance
about the writing of JanagaNamana. (61) is acceptable if the utterer is credited with
knowing that Rabindranath was working on a poem called JanagaNamana, and is
reporting that the writing of this poem is now completed. The next example shows
that when the contrary assumption is forced, the use of a compound verb becomes
impossible:

(62) rabindranaath ne bhaartiiy raaSTr giit likhaa

(name) AGT Indian Nat. Anth. wrote
“Rabindranath wrote the Indian National Anthem.”(241)

Here the compound verb likh liyaa would be impossible, since the utterer could not, at
the time that he was reporting the completion of the work, have known that the poem
would become the Indian National Anthem.
Prior knowledge or intention with regard to the goal of the verb, then, that is to say
referentiality, appears to be an important criterion favoring the use of the compound
verb (see discussion by Hook, op. cit. 240-242). This again points to the compound
verb’s being higher in transitivity than the simple verb, and suggests that the
impossibility or awkwardness of the compound verb to express an Effective action is
linked to the lower transitivity of E-clauses.
The behavior of the verbal expression meaning “to discover”, khoj karnaa, is
especially instructive, because here too the use of the compound verb implies some
prior awareness of the discovered item. Cook points out that Columbus “discovered”
America by accident; consequently the compound verb is out of the question:

(63) kalambas ne amriikaa kii khoj kii

(name) AGT (name) of discovery make
“Columbus discovered America.”(240)

However, the “discovery” of the maritime route to India by Vasco da Gama can be
construed slightly differently, since there was known to be such a route in advance of
its actual exploration. Hence:

(64) bahut saal0 see koSiS karne ke baad waasko daa gaamaa ne
many years of attempt making of after (name) AGT
hindustaan jaane ke lie ek naii raah kii khooj kii/khooj kar lii
India going of for a new way of discovery made
“After trying for many years, Vasco da Gama discovered a new way of
going to India.”(241)

Here, Hook’s jury tended to accept the possibility of the compound verb kar lii,
although the simple verb kii was preferred. The critical case of “discover” shows that
E-verbs like “write” are distinctive not because they create their objects (a semantic
property), but because, for whatever reason, the object is presented as non-existent
prior to the time of the verb (a discourse factor).

3.6 De-nominal Derivation of E-Verbs: Plains Cree.

A rather typical pattern is for E-objects to be lexically converted into verbs by means
of a derivational affix, or in extreme instances by zero affix (e.g., in English to flower,
to smoke), so that N is converted into a verb meaning “to produce N”. The resulting
verb is usually intransitive, even when a cognate object is added. Of the numerous
examples which could be adduced here, an Algonquian one, Plains Cree (Wolfart
1973), was selected because in these languages the distinction of transitivity in the
verb stem is morphologically overt.
In Plains Cree a suffix -ihke- can be added to nouns to form verbs having the
meaning “make, gather, produce such-and-such” (Wolfart 1973: 72). This suffix is
evidently completely productive. The resulting verb is classified as Animate
Intransitive, i.e. no object and an Animate agent:

(65) meenisk, trench: meeniskehkeew “he digs a trench”.

(66) matotisaan sweat-lodge: matotisaanihkeew “he builds a sweat-lodge”.

It seems, not surprisingly, that the E-object in such expressions is a non-referential,

indeed non-individuated one, and it is significant that the mere act of producing
something does not automatically mean that the suffix -ihke- is used. Consider, e.g.:

(67) namooya miina pakahkam aya kiikway oskaaya

not also I-don’t think anyone any new
waaskahikana eewiiosiihtaahk...
houses will-build
“I don’t think they are going to build any new houses.” (p. 36)

Evidently in a pair like

(68) meenisk, trench: meeniskehkeew he digs a trench.

the noun meenisk is in a sense incorporated into the intransitive verbal suffix ihke. As
is typical of such incorporated nouns in general, meenisk cannot refer to any specific
trench, but is non-referential and generic; it is in fact no longer a noun (for a defense
of this position, and its relevance for noun incorporation see Hopper and Thompson
l984: 710-713). In (67), however, the noun is modified by the adjective oskaaya “new”,
and is playing more of a role in the discourse (although it is still not fully referential,
of course). Such nouns retain some of the morphology characteristic of nouns, and
cannot be incorporated.
This discussion anticipates a point which will be made again later in the paper: We
rarely, and perhaps never, find a thorough-going morphosyntactic distinction which
singles out the class of E-objects against the class of A-objects. Rather, this distinction
tends strongly to be cross-cut by the distinction referential / non-referential. Thus in:

(69 (a) He builds houses

(b) He repairs motorbikes
(70) (a) He built that house over there.
(b) He repaired this motorbike.

we do not find that the (a) sentences are grouped grammatically as a class against the
(b) sentences, but rather that the (a) and (b) sentences form a class in each pair. As a
surface realization Fillmore’s “Factitive Case” is not very useful. The following is an
interesting exception.

3.7 The Indonesian Suffix -kan

The suffix -kan in Indonesian has several applications, including the simple one of
making transitive verbs out of adjectives or intransitive verbs, e.g.

(71) (a) besar “big”: besarkan “increase, make big”

 (b) tidur “sleep”: tidurkan “put to sleep”

In addition to this causative/transitive function, -kan may also add a

prepositional/instrumental meaning which permits oblique NP’s to act as direct
objects; thus, for example, with dative objects, the following are both possible:

(72) (a) Dia menjual permata itu kepada teman nya

he sold jewel the to friend his
“He sold the jewel to his friend”
(b) Dia menjual-kan teman nya permata itu
“He sold his friend the jewel”

There can be little doubt that -kan is a transitivizing suffix. Not only does it function
to convert intransitive verbs and predicates into transitive verbs, but it functions to
create personal direct objects out of dative indirect objects, to signal greater intensity
of the action, and it serves a number of other functions which point to a common
transitivizing function.
This being so, we would predict that: If two clauses, one containing a simple verb
and the other a verb suffixed with -kan, differ in meaning as Effective and Affective,
the Effective clause will have the simple verb and the Affective clause the suffixed
verb. This prediction follows from the general hypothesis that E-verbs are less
transitive than A-verbs. John Wolff has pointed out to me the following data:

(73) (a) Dia mencampur sayur-sayuran

he mix vegetable-salad
“He mixed the vegetable salad”
(b) Dia mencampur-kan sayur-sayuran
“He mixed in the vegetable salad”

Here the (a) sentence, without -kan, conveys the sense of “causing a mixture of
vegetables to come into being as a vegetable salad”, while sentence (b), with -kan,
means “add new things to a mixture already in existence, i.e. mix something in with
something else, stir something in.” The (a) sentence, without -kan, has the force of
bringing ingredients together to create something which was not previously there. In
other words, (a) is Effective and (b) is Affective, exactly as predicted by the hypothesis.

3.8 Summing Up.

So far the data suggest that languages may distinguish between E-clauses and A-
clauses in both morphology and syntax. However:
(a) There is never any special, “dedicated” morphosyntax for this distinction. This is
of course not surprising, since the same can be said about any aspect of a language’s
grammar. We do not expect to find one-to-one correlations between function and form.
(b) If a construction distinguishes between A-objects and E-objects, the distinction
will be one which in other contexts shows up as a difference in Transitivity, and the E-
clause will always be less transitive than the A-clause.
(c) Many constructions which differentiate A-objects from E-objects are sensitive to
the distinction between specific, referential Objects and generic or non-referential
Objects. Typically, E-objects are non-referential, and in many constructions the E-
object is simply a type of non-referential object.
These observations demand a study of the discourse functions of E-verbs and their
Objects.

4. The Functions of E-clauses in Discourse: Malay.

In this section I will consider some of the apparent discourse correlates of the
grammatical and distributional phenomena studied in the preceding sections. The
data for the study come from a Malay narrative, the Hikayat Abdullah. Citations are
from the Malaya Publishing House edition.

4.1 Morphological Correlates of Effectivity.

Transitive clauses in the variety of Malay under consideration here appear in a

number of different types. Two of these are especially common:

(i) The Active type, in which the clause has the word order Subject - Verb - Object,
and the verb has the prefix meng-, for example:

(75) ...sebab ia mengetahui bahasa Jawa

because he know language Java
“because he knew the Javanese language” (85)

(ii) The Ergative type, in which the verb has the form of a Passive, and the Agent and
the Object follow the verb, usually in that order. These three examples show (a) the
Agent as an enclitic pronoun on the verb, (b) the Agent as a full noun phrase with the
agentive preposition oleh, and (c) the Object of the verb supplemented with the
“accusative” preposition akan:

(76) (a) . . .di-ambil-nya tiga puloh ringgit

PASS. take AGT thirty dollar
“he took thirty dollars” (92)

(b) maka di - bawa - lah oleh Tuan Milne merika itu sakalian
then PASS take LAH “by” Mr. Milne them all

ka rumah-nya
to house - his
“Then Mr. Milne took them all to his house” (123)

(c) di - pulangkan - lah oleh bapa ku akan segala perkakas nya

PASS hand-over LAH “by” father my AKAN all tools his

dan peti-peti tulis kapada ku

and case writing to me
“My father turned over to me all his materials and writing case.” (31)
 The enclitic morpheme lah is a discourse particle (cf. Hopper 1979: 40-49). I shall be
particularly concerned with the use of the “accusative” preposition akan and the
verbal suffix -kan, with which it is to a large extent in complementary distribution.
We noticed that in Modern Indonesian this suffix is a transitivizing one, in the sense
that it increases the syntactic and semantic valency of the verb to which it is attached.
Furthermore, under some circumstances the suffix is able to distinguish between
Effective actions and Affective actions, and when it does so, the presence of the suffix
will always indicate that a Patient is impinged upon by the action, while its absence
will indicate that the Object is brought about by the verb’s action.
We should therefore find in actual Malay discourse that actions which are presented
as changing something which is already present in the discourse, or is assumed to be
present, may take either the accusative preposition akan or the transitivizing verbal
suffix -kan. On the other hand, actions which are presented as creating or bringing
about their object will never be characterized as “fully transitive” in the sense of
permitting Objects with akan or verbs supplemented with -kan. In the next two
sections I shall present some limited evidence suggesting that this hypothesis is
substantiated, and will suggest some more general discourse reasons why this should
be so.

4.2 BUAT: The Physiology of an Effective Verb.

Since a meaningful discussion of even a handful of E-verbs would bring this paper to
an impossible length, I will single out the most obvious Effective verb, the general
verb corresponding to English “make”. The root form of this verb is buat, and the
following stems are found in the text: buat, buatkan, perbuat, perbuatkan. The voice
prefixes mem- active and di- passive, 3rd person may be prefixed to these stems. The
semantic force of the prefix per- is not clear in the text, and in what follows it will be
assumed that buat and perbuat are equivalent, differing perhaps in some nuance
comparable to English make and construct. This narrows discussion to the presence
versus the absence of the suffix -kan, and the distribution of the Object preposition
akan.
If the hypothesis that the transitivizing suffix -kan will not refer to an E-Object is to
be sustained, a coherent explanation must be found for such pairs of sentences which
occur in the data as the following:

(77) (a) maka tempat itu di-perbuatkan kebun di-tanamkan segala

and place that PASS-make garden PASS-plant all

jenis bunga-bungaan dan pohon-pohon.

kinds flowers and trees
“And that place was made into a garden [and was] planted with all kinds of flowers
and trees.” (169)

(b) datang lah berpuloh orang China membelah-belah batu itu

came LAH dozens men Chinese break-up stone that

di-buat nya rumah

PASS make they house
“Dozens of Chinese came along and broke up the stone to make houses.” (167)
The difference between the verb perbuatkan in (a) and buat in (b) is that in the (a)
sentence the verb has a Patient, namely the place, which is “taken” and made into a
garden. In fact wherever buat or its derivative is suffixed with -kan, its meaning is
predictably not “to make such and such”, but “to make or transform such and such into
something else.” In such clauses the Effected object is a sleeper, which may not have
any of the syntactic properties of grammatical patients, and the material or initial
entity which is its source always has syntactic priority over it. Moreover, in the (a)
sentence, where the verb has the -kan suffix, there is a sense that the ‘place’ was
converted (completely) into a “garden”, whereas in the (b) sentence it is not claimed
that the stone was converted into houses, but that (some of) the stone “went into” the
making of the houses. A similar explanation holds for the absence of -kan in the
following:

(78)...chita Eropah di - buat nya basahan, dan sakhlat

chintz PASS make they bathing-suit and wool
di - buat nya seluar
PASS make they trousers
“They made bathing-suits out of chintz, and trousers out of wool” (163)

In these two clauses the verbs have as their Patients the material nouns chita
Eropah and sakhlat, and the E-objects are basahan and seluar. But this time the
verbs, that is in each case buat, are not suffixed. Here a second and crucially
important restriction comes into play: The object indexed in the verb by -kan must be
anaphoric, that is, referential. The generic materials have not been previously
mentioned in the discourse. It is this fact which makes example (78) crucially different
from (77) (a), in which the patient-topic tempat itu “that place” of the transitive verb
buatkan is already in the discourse.
The overtly transitive verb buatkan, then, takes a direct object which denotes the
source of the thing made, and which is referential with respect to the discourse, i.e.
has been previously mentioned. If the same verb appears without the suffix, its only
Object is the thing made. In this case, we might expect that a high proportion of the
objects of buat will be non-referential. In order to test this hypothesis, I have
examined the nature of the E-object for all occurrences of the verb buat/perbuat in a
20,000 word sample of the text, and have compared the referentiality of the object for
all occurrences of this verb with the same data for some other common transitive
verbs. Objects were classified quite simply as referential if they were pronouns or were
accompanied by a definite determiner; otherwise they were classified as non-
referential. Consider first the figures for unsuffixed forms of the verbs buat, ambil “to
take” , and bawa “to carry, bring”:

OBJECT

REFERENTIAL NON-REFERENTIAL

buat/perbuat “make” 2 25
ambil “take” 12 4
bawa “bring, carry” 22 3
 Figure l: Object-referentiality for Three Transitive Verbs

There can thus be no question that the prototypical E-verb in Malay has a discourse
affinity for non-referential objects. At the same time it appears that verbs which one
would judge to be close to the prototype for A-verbs have the reverse affinity for
referential Objects. These facts are, of course, consistent with the de-transitivized
nature of E-verbs which we noted across several languages.
A second observation involving the discourse distribution of E-verbs concerns that
difference between verbs which are prefixed with the Active verb prefix meng-/men-
/mem- and those which are either unprefixed or are prefixed with the 3rd person
Passive/Ergative agent agreement morpheme di-. The figures from the same sample
for the three prefixed verbs membuat “make”, mengambil “take”, and membawa
“bring, carry” are as follows:

OBJECT

REFERENTIAL NON-REFERENTIAL

membuat “make” 8 21
mengambil “take” 4 3
membawa “bring, carry” 26 25

Figure 2: Object Referentiality for Three Verbs with meng-

These figures indicate that whereas the presence of the meng- prefix is for most
transitive verbs associated with a decrease in the referentiality of the direct object, the
verb (per)buat escapes this tendency and in fact actually slightly increases the average
referentiality of its object when prefixed with mem-. The reasons for the increase are
not clear, but perhaps have something to do with the grammaticization of the use of
the meng- form of the verb in certain subordinate clauses, in which there is a greater
likelihood of the Object being presupposed. The significant fact is that in all its forms
the prototype for E-verbs overwhelmingly favors Objects which are of the
generic/indefinite type in discourse in which transitive verbs of the Affective type
show the opposite tendency.
The ultimate reason for this affinity between the E-verb and the non-referential
Object is clear. Typically an E-verb does not report an action carried out on one of the
on-the-scene discourse participants. Instead, E-verbs typically function in one of two
ways. They are like presentatives: they actually introduce new participants into the
discourse, things which are “produced” through an action. Or they report a unified
event in which verb and object are not conceptually separable, but in which the object
is incorporated into the verb. It is therefore not surprising that the predominant
strategies for differentiating A-clauses from E-clauses frequently overlap with
strategies for presenting new participants, such as indefiniteness marking, and for
incorporating Objects into verb stems (such as lexifying). When one excludes
phenomena which really have these latter functions, what is left is a set of ragged
ends on the far periphery of the grammar.
5. Conclusion.

I conclude, then, that although the distinction of “Affective”and “Effective”, or

“Factitive”, or “Resultative” objects has often been made, and phenomena which
involve it have been pointed out from time to time, the importance of the distinction to
grammar is minimal. Frequently, and perhaps always, when we encounter the
distinction it turns out to be a one of referentiality in the verb’s object which is really
being marked, not the (for languages) more trivial question of the semantic nature of
the verb-object relationship.
This observation raises an issue which is, when placed in a more general context, a
truly central one for linguistics. Are any functional aspects of the motivation for
grammars understandable in terms of purely “local” semantic contrasts like
Affective/Effective? Or will such “local” semantic phenomena always turn out in the
end to be epiphenomenal and derivative of the role the grammatical constructions
concerned typically play in discourse?
Consider, for example, the question of person, definiteness, and animacy which has
often been mooted as an explanatory factor in the marking of transitivity. We might
view this feature of nouns and pronouns as a hierarchy which is fixed in speakers’
minds as a way of ordering the world of things about them. In this view, the many
linguistic phenomena which appear to be responding to such a segmentation of the
real world are explicable at the sentence level by simple classification of the
arguments of the sentence, and no wider discourse context is of the slightest help in
understanding the grammatical phenomenon. On the other hand, we might find that
when discourses as a whole are studied in detail, what emerges as an explanatory
parameter is a set of broader, more pragmatic organizational principles such as
anaphoricity, topicality, foregrounding, and so on. From this perspective it might now
seem that we were mistaken in thinking of the categories of person, animacy, and
definiteness as fixed, semantic fields, and that instead these are imprecise “local”
correlates of the broader principles which at the sentence level are seen only obscurely
and indirectly.

REFERENCES
Bennett, Charles E., 1966[1914]. The Syntax of Early Latin. Vol. II. Hildesheim: Olms.
Chao, Yuen-Ren. 1968. A Grammar of Spoken Chinese. Berkeley: University of
California Press.
Fillmore, Charles E. 1968. “The Case for Case” in E. Bach and R. Harms, eds.,
Universals in Linguistic Theory. New York: Holt, Rinehart and Winston, 1-90.
Hook, Peter. E, 1974. The Compound Verb in Hindi. Ann Arbor: University of
Michigan Center for South and Southeast Asian Studies.
Hopper, Paul J., 1979. “Some observations on the typology of focus and aspect in
narrative language.” Studies in Language 3.1, 37-64.
Hopper, Paul, and Sandra Thompson, 1980. “Transitivity in grammar and discourse”
Language 56.251-99.
Hopper, Paul, and Sandra Thompson, eds., 1982. Studies in Transitivity. New York:
Academic Press. (Syntax and Semantics 15).
Jespersen, Otto, 1933. Essentials of English Grammar. New York: Holt. Lakoff,
George, 1977. "Linguistic Gestalts.” CLS 13.
Lord, Carol, 1982. “The development of object markers in serial verb languages.” In
Hopper and Thompson, eds. 277-300.
Lyons, John. 1977. Semantics, vol. II. Cambridge University Press.
Lyons, John. 1968. Introduction to Theoretical Linguistics. London: Cambridge
University Press.
Munro, Pamela, 1982. “On the transitivity of ‘say’ verbs”. In Hopper and Thompson,
eds., 301-318.
Quirk, Randolph and Sidney Greenbaum, 1973. A Concise Grammar of Contemporary
English. New York: Harcourt Brace Jovanovich.
Wolfart, C., 1973. Plains Cree: A Grammatical Study. Philadelphia: Transactions of
the American Philosophical Society.

1
Full reference: Hopper, Paul J. 1986. “Causes and Affects.” In W.H. Eilfort, P.
Kroeber, and K. L. Peterson, eds., Causatives and Agentivity. Papers of the
Parasession of the 21st Annual Meeting of the Chicago Linguistic Society, pp. 67-88.
Chicago: CLS.