The asymmetry principle


Material Information

The asymmetry principle a functional investigation of transitivity and topic-comment structuring in English
Physical Description:
viii, 153 leaves : ill. ; 29 cm.
Schafer, Michelle Suzanne
Publication Date:


Subjects / Keywords:
Linguistics thesis, Ph.D   ( lcsh )
Dissertations, Academic -- Linguistics -- UF   ( lcsh )
bibliography   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph.D.)--University of Florida, 1998.
Includes bibliographical references (leaves 146-151).
Statement of Responsibility:
by Michelle Suzanne Schafer.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 029546630
oclc - 40154149
System ID:

This item is only available as the following downloads:

Full Text







This project is dedicated to Dr. Gerald Eugene Merwin,
My Father,
who is not alive to see its completion
yet has everything to do with it being done.


There are many people without whom finishing this dissertation would have been difficult, if

not impossible. First, I would like to thank my committee members, each of whom contributed

something unique: Dr. Chauncey Chu, for his support, leadership, faith, and ability to give me

a swift kick when it was time to get going; Dr. Michel Achard, for the inspiration and challenges

he offered; Dr. Galia Hatav, for her faith in my abilities and the opportunity to assist with her

own work which lead to my first real "breakthrough"; Dr. William Sullivan, for his support,

guidance through bureaucracy, and good questions; and Dr. Douglas Dankel, for the

opportunity to think about my ideas in a different context which was instructive and refreshing.

I would also like to thank Cindy Powell and the staff of Anderson 112 for their patience in

answering the same questions year after year; Gordon Tapper and Lucy Pickering, of

Academic Spoken English, for many years of conversation and support; Kathy Kidder and

Anne Wyatt-Brown, two of the finest "bosses" I've ever had, both of whom allowed me plenty

of room for the creative application of linguistics; and to my students over the years, all of

whom have enjoyed (or suffered) my experimentation in classroom application, usually without

being aware that they were the guinea pigs in my latest test run.

Finally, these acknowledgments would not be complete without special thanks to my family,

especially my mother and husband. My mother has always believed in my abilities, supported

my goals, and certainly helped finance the trip from there to here. My husband, who came

along a couple of years into the process, gave me fantastic support and lots of room to think

and do whatever needed to be done. Most importantly, he gave to me extraordinary love and

inspired in me courage that I didn't know I had.



ACKNOWLEDGMENTS.............. ...................................................... iii

LIS T O F F IG U R E S ............................................................................ vi

A B S T R A C T ........................................................................................ ............. v ii


1.1 Introduction................................................... ........................... 1
1.2 M .A .K H alliday..................................................... .................. 3
1.3 Approaches to Transitivity........................................................ 5
1.3.1 Verb-Oriented Theories.................................................. 6
1.3.2 Clause-O oriented Theories............................................... 8 Propositional models...................................... 8 Hopper and Thompson, 1980........................... 11 Cognitive models............................................ 12
1.3.3 Extra-Clause Oriented Theories.................................. 15
1.3.4 Transitivity and Text............................ ................ 19
1.4 Overview of Project................... .. ... ............... 20

2.1 Introduction to the Problem.......................................................... 23
2.2 Syntax................ ................... ........... ............ 25
2.2.1 Grammatical Behavior.............................................. 26
2.3 The Semantics of Transitivity...................................................... 27
2.3.1 The Active, Affirmative, Declarative Prototype.............. 29
2.3.2 The Transitive Prototype: One or Many?........................ 30 Berman and Slobin revisited--event construal 40 Degrees of transitivity.................................... 42
2.3.3 Event Structure and Transitivity............................. ....... 43
2.3.4 "Agent," "Patient," and Other Labels............................ 46
2.3.5 Verbal Modality...................................... .................. .. 50 Fast-paced.................................................. 50 Completed................................................ 51 Perceptually-cognitively salient..................... 52
2.4 Transitivity and Tim e................................................................. 55
2.5 The Asymmetry Principle in Cognition and Communication....... 60
2.6 Concluding Remarks..................................... ..................... 62

IN ENG LISH ........................................................... ............................. 65
3.1 Opening Remarks........................................ .................. .......... 65
3.1.1 Thematic Coherence............................... ....... ............ 66
3.1.2 Perspective.......... ........ .... ........ ........................... 67
3.1.3 Management of Perspective......................................... 68

3.1.4 The M odel in Brief................................... ................ 69
3.2 D iscourse Topicality.................................................................. 72
3.3 Perspectival Distance....................................................... 79
3.4 The Four Clause Types of Transitivity.................................... 83
3.4.1 Telic Transitives ............................................................ 84
3.4.2 Atelic Transitives......... ........... ............ ................. 88
3.4.3 Intransitives/Durative Transitives................................. 95
3.4.4 Dispersed Transitives................................................... 96
3.5 Deep and Shallow Transitivity................................................... 99
3.6 Quantitative Patterns........................... ...... ............. 103
3.7 Concluding Remarks ..................................... 105

4 TRANSITIVITY IN EXPOSITORY TEXT......................... ......... ....... 107
4.1 O opening Rem arks ............................................. ......... ..... 107
4.2 Towards a Linguistic Definition of Expository Text..................... 108
4.2.1 Narrative versus Expository Text................................ 108 Differences....... ..... ................................ 108 Them e............. .......................... 109 Linkage................... .................... ... 111 Sim ilarities.................................. ........ .. 115 Argument Line................................... 115 Illocutionary Force............................. 117
4.3 Topic-Comment Structuring in Expository Text.......................... 119
4.3.1 The Role of Adverbial Expressions: Continuation Spans. 120
4.3.2 The Role of Transitivity in Expository Text: Transition
Spans ........................................................................ 126 Transitivity and topic-switching.......................... 126 Micro-texts and bridge spans........................... 127
4.3.3 Bridge Spans and Transitivity........................................ 131
4.3.4 Passives in Expository Text......................................... 134
4.4 Concluding Remarks: Topic-Comment Structuring and Asymmetry 137

5 TRANSITIVITY: CREATIVITY AND CONSTRAINT.......................... 139
5.1 Summary of Results ..................................................................... 139
5.2 Final Remarks ........................................ .............................. 143

B IB LIO G R A P H Y .................................................................. ............................ 146

BIOGRAPHICAL SKETCH ............................ ............ .... 153



Figure 1.1 Halliday's Six Categories of Clause Types......................... 4

Figure 2.1 Statement of Transitivity................................ 24

Figure 2.2 Mandler's Image Schemas of Motion.................................. 32

Figure 2.3 Mandler's Image Schema of Caused Motion...................... 32

Figure 2.4 Mandler's Image Schema of Agency.................................. 33

Figure 2.5 Croft's Event Views....................................................... 36

Figure 2.6 Kemmer's Event Schema.................................................. 37

Figure 2.7 The Four Clause Types of Transitivity........................... 63

Figure 3.1 Referential Specificity, Information Value, and Grammatical
R elations.......................... ................. ......... .... 79

Figure 3.2 Quantitative Analysis of Clause Types with Percentages of
O ccurrence........... ... ..................................... ..... 103

Figure 3.3 Raw Count and Percentages Clause Types in Adult
Narrative................................................... .......... 103

Figure 3.4 Time Line and Clause Types--raw score out of total/percentage
in text............................................................... ... 105

Figure 4.1 Topic/Subtopic .................. ..... ............................ 123

Figure 4.2 Topic/Subtopics Specified........................... ... ..... 124

Figure 4.3 Complete Topic/Subtopics of excerpt "Speech Acts".......... 125

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Doctor of Philosophy Degree



Michelle Suzanne Schafer

August, 1998

Chairman: Dr. Chauncey Chu
Major Department: Linguistics

Current models of transitivity have been based on the examination of isolated sentences

or text taken almost exclusively from the genre of narrative. The result has been the confusion

of transitivity with linguistic features particular to narrative. This confusion has found its way into

both prototype models of transitivity and discourse-functional approaches. To ameliorate this

situation, this project extended the investigation of transitivity into the genre of expository text.

Narrative texts were first examined to determine what might properly be called transitivityy" and

what was actually a linguistic property of narrative. Then, texts were examined to determine the

function of transitivity in narrative. Finally, expository texts were examined to determine if first,

the model proposed to account for the data in the investigation of narrative would extend to

expository text and second, if transitivity functioned differently in expository text. A model of

transitivity is proposed which better accounts for the surface realization of transitivity by limiting

prototype effects to four sentence types found in English: the telic transitive, the atelic transitive,

the dispersed transitive, and the intransitive or durative transitive. Second, a functional model is

proposed which shows that in narrative, transitivity is a continuation device which contributes to

the management of perspective; in expository text, transitivity is a transition device which

creates bridges from one span of text to another. Overall, this study demonstrated that while

transitivity is indeed a grammatical system for the expression of events, its use as such is only

exploited in those discourse types which turn on events. In non-event-driven discourse,

transitivity gives way to alternative forms of grammatical realization. The principle of asymmetry

guiding grammatical cognition motivates this difference. In discourse types skewed towards

linear organization, linear structures such as event-driven transitivity dominate. In discourse

types favoring non-linear organization, non-linear structures which realize situations and states

dominate. This tendency seems well-encoded by the differences between transitivity and topic-

comment structuring in English.


"Many words are subject to a distinction which is designated by different names and
therefore not perceived as essentially the same wherever found, namely that of a word
complete in itself (or used for the moment as such) and one completed by some addition,
generally of a restrictive nature. Thus we have the complete verb in he sings, he plays, he
begins: and the same verb followed by a complement in he sings a song, he plays the piano,
he begins work. In this case it is usual to call the verb intransitive in one case and transitive
in the other, while the complement is termed its object"
Jesperson, 1924. 88

1.1. Introduction

A fundamental assumption of functional grammarians is that the primary purpose of

language is communication; therefore, the ultimate goal of studying language is the investigation

into this relationship: "The theme unifying the various functional approaches is the belief that

language must be studied in relation to its role in human communication. Language is thus

viewed as a system of human communication..." (Foley and Van Valin, 1984:7).

One question which emerges from this perspective is: what is it that human beings use

language to communicate? The functional response is "If one is concerned with the role of

language in social interaction, then aspects of linguistic structure which serve to signal social as

opposed to purely referential meaning share center stage with purely referential elements..."

(Foley and Van Valin, 1984:9). Lambrecht puts it another way: "...certain formal properties of

sentences cannot be fully understood without looking at the linguistic and extralinguistic contexts

in which sentences have these properties are embedded." (Van Valin, 1993:2). Perhaps the

finest statement for why functional insights are necessary to the study of language comes from

Leonard Talmy, a cognitive scientist. Talmy (1988) states that one of the principal functions of

structure is to provide conceptual coherence. For language this means grammar, which is the

way of "...unifying contentful material within a single conceptual system and rendering it

manipulable--i.e., amenable to transmission, storage, and processing- and that its absence

would render content an intractable agglomeration" (Talmy, 1988:196).

"Unifying contentful material" at the sentence level, though, is insufficient; the result

would be an intractable agglomeration of individual grammatical constructions. In order to

achieve coherence through time (a factor critical to successful communication), the grammatical

structures themselves must be organized. This is one function of discourse structure: the

structural organization of grammatical constructions for the purpose of managing information

flow through text and time. This study, too, will examine a structural system, transitivity, with

respect not only to its referential meaning at the sentence level but to its communicative and

contextual functions through time and text.

Transitivity has enjoyed a resurgence of interest in response to functional inquiry. There

have been at least two major functions proposed for transitivity; first, the sentence-level coding

of events (Langacker, 1990; Rice, 1987; Giv6n, 1993:vol.1) in which transitivity is described in

terms of coding participants and action/event relationships; second, the discourse-level coding of

grounding (Hopper and Thompson, 1980; DuBois, 1987; Giv6n, 1995) in which transitivity codes

foregrounded information and manages information flow in connected texts. Both viewpoints

use much of the same theoretical apparatus (energy flow, attention, framing, agency, etc.)

though neither has been set vis-A-vis the other in a clearly explicated manner. Transitivity

provides fertile ground for such an examination as it appears to extend to both "levels" of

linguistic exploration: the local, particular grammatical instantiation of any referent scene and

the connection of that instantiation with others of its kind. The explicit purpose of this study is

to distill from the various claims made about transitivity what this relationship might be.

Theoretical approaches to transitivity can be roughly classified along three dimensions:

verb-oriented, clause-oriented, and extra-clause oriented. In the first, transitivity is understood

mainly as a property of verbs. In the second, it is a property of logical propositions or events. In

the third, transitivity is a matter of perspective or speaker construal (though in practice, it is often

difficult to separate semantics and pragmatics of the various approaches). Within each

dimension, theorists employ a variety of devices to account for transitivity, and while

grammarians are obliged to give transitivity some mention because of its intimate link to clause

structure, there is considerable variation in the degree of centrality accorded it.

1.2 M.A.K. Halliday

Before discussing each of the approaches above, some mention of M.A.K. Halliday must

be made. Halliday (1967) started the serious discussion of transitivity in a three part work titled

"Notes on Transitivity and Theme in English." While it is not within the scope of this study to

review the whole work, Halliday's basic claims are adhered to in subsequent work (though he is

not always explicitly cited).

For Halliday, transitivity was part of three systems whose point of origin is the matrix

clause, that clause which contains the major predication of the sentence. (The centrality of

predication as the starting point for linguistic discussion is still valid for many linguists:

"Regardless of the type of discourse under consideration, the clauses which constitute the

discourse are constructed around predications consisting of a predicate and its argument" (Foley

and Van Valin, 1984:27).) In particular, the transitivity systems "...are concerned with the type of

process expressed in the clause, with the participants in this process, animate and inanimate,

and with various attributes and circumstances of the process and participants" (38). For

Halliday, transitivity is the system which provides sets of options relating to "cognitive content,

the linguistic representation of extralinguistic experience, whether...of the external world or of

feelings, thoughts, perceptions" (Halliday, 1967b:198). As we will see, transitivity as a cognitive

phenomenon is still a critical, definitive notion.

The basic types of processes Halliday proposes are directed and non-directed action,

meaning activities which are directed or not directed towards producing a specific effect on a

participant. The relationships among participants and processes are thus very important. The

first relationship Halliday describes is that of the grammatical subject of the clause (for English,

every clause must have a subject). The subject has one of two semantic roles, the actor (the

one performing the action) and initiator (one who is energy source of the action). The

grammatical object has one principal role, the goal of the action. Subjects and objects

participate in one of three types of clauses: operative, middle, and receptive. Six relationships

emerge from crossing processes, participants, and clause types.

operative middle receptive



Figure 1.1: Halliday's Six Categories of Clause Types

Halliday locates transitivity within the clause and in doing so retums to features of the

clause as the organizing rubric for his discussion. Thus "subject as actor" or "subject as initiator"

are unnecessary labels as they fall out naturally from directed and non-directed clause types:

"Treatment in terms of clause types enables us to generalize by saying that there are in fact

three distinct types of subject, or subject functions, determined by the transitivity systems; these

could be labeled 'ergative', 'nominative' and 'accusative'" (Halliday, 1967a:46). Halliday

distinguishes clause features, subject type, and participant role as follows.

Clause Feature Subject Type Participant Role
operative ergative actor (not goal) in directed;
initiator (not actor) in non-directed
middle nominative actor/goal in directed;
initiator/actor in non-directed
receptive accusative goal (not actor) in directed;
actor (not initiator) in non-directed

The point so far critical to this study is the emergence of clause types in terms of

transitivity resulting from the interactions between processes and participants. Halliday comes

back to this point again and again throughout all three sections culminating in an argument for

the linguistic uselessness of the binary opposition "transitive" versus "intransitive." In particular,

Halliday points out the difficulty in ascribing verbs this characteristic (a "dictionary analysis")

since so many verbs in English participate in clauses with and without goal participants. That is,

S actor S actor/goal S goal
she washed the she washed the clothes were
clothes washed

s initiator S initiatoractor 5 actor
he marched the the prisoners the psoners were
prisoners marched marched

the presence of a goal participant is often not fully predictable from the meaning of the verb

itself: "The potential distinction, in other words, between verbs which are inherently goal-directed

and verbs which are not, is less useful as a generalization than the actual distinction between

clauses which contain a goal, or rather (an important difference) a feature of goal-directedness,

and those which do nor (Halliday, 1967c:182; emphasis added). This being the case, the

opposition of transitive and intransitive is not useful for linguistic analysis. Instead, we should

employ a full range of transitivity distinctions as associated with the process-participant relations

within the clause.

Halliday's notion that there must be a scale of transitivity distinctions has emerged as

one of the crucial arguments in current functional and discourse grammars. However, in

unlikely contrast, the labels "transitive" and "intransitive" have also remained, confusing the

study of transitivity considerably. One of the principal tasks of this study is to clarify such

confusion, and this point will be taken to task at length in Chapter Two.

The second insight of Halliday's which can be found throughout the literature and which

is also critical to this study is the notion "goal-directedness." Evidence for goal-directedness as

the primary semantic relationship in transitivity is found in both discourse studies (such as

DuBois, 1987) and cognitive studies (Rice, 1987). As such, it enjoys considerable elaboration in

Chapter Two.

Finally, Halliday's contention that the clause is the proper domain of transitivity has also

emerged as fundamental. This notion is taken up briefly in the following discussion and is

elaborated upon in Chapter Two.

1.3 Approaches to Transitivity

There are three primary theoretical approaches to transitivity: verb-oriented approaches

in which the verb controls argument structure; clause-oriented studies in which transitivity is seen

as a property of events realized at the clause level; and extra-clause oriented studies which offer

pragmatic and discourse explanation of transitivity.

1.3.1 Verb-Oriented Theories of Transitivity

Verb-centered theories of transitivity are the most traditional and well-known. These

construe transitivity as a matter of predicate argument structure identifying particularly the role of

the verb and direct object: a verb is transitive if it requires a direct object or if there is a direct

object present. Early theories didn't discriminate among types of verbs nor classify according to

grammatical alternations; any verb which required some kind of object was transitive. Thus, a

representative definition would be "a transitive verb denotes an action which passes over from

the doer of the action to the object of it" but "an intransitive verb denotes a state or simple action

without any reference to an object" (Curme, 1947: 22-23; emphasis added). Curme included

reflexive verbs with the transitive, stating simply that the receiver and doer happen to be the

same person. Other examples include We make fudge, I remember Tony well, or She has no

brothers and sisters (Long,1961 as cited in Rice,1987). More recent theorists, though, do classify

verbs according to type and recognize that different types correlate with different grammatical

behaviors. Thus, we have action verbs, state verbs, process verbs, psychological verbs,

reflexive verbs, verbs with cognate objects, etc., or the aktionsart classes: state, activity,

achievement, and accomplishment (Vendler,1967; Dowty,1979; Van Valin,1993). Moreover, the

definition of transitivity vis-a-vis these verbs has developed so that only those that passivize are

recognized as "fully" transitive, or in some cases, verbs are categorized as transitive and sub-

categorized as also passivizable.

Nevertheless, the syntactic approach to transitivity is not limited to traditional

grammarians. Indeed, this approach finds representation in both current formal and functional

approaches. Configurational and structural approaches championed by Chomsky (1981) and

most recently by Hale and Keyser (1985) define transitivity as a derivative of the syntactic or

semantic configuration of the sentence. In the Aspects model, a verb is transitive because it

subcategorizes for a noun phrase. A verb which has both transitive and intransitive instantiations

would be marked exceptional. In the Govemment-Binding model, a verb is transitive because it

binds its direct object via government relations. Although Hale and Keyser do allow extra-

propositional elements to influence the clause, the mapping between the components which link

perspectives on an event (Lexical Conceptual Structure) with clausal instantiation (Lexical

Structure) is unclear. And still, in all of these cases, the verb is the element responsible for

transitivity; no appeal is made to other elements in the clause which may contribute. Yet

semantic research as early as 1979 (Dowty) and still continuing (Hinrichs, 1985; Verkuyl, 1993)

cites the importance of the NP in determining the syntactic behavior of the clause. A sentence

with an NP object which is not specifically quantified does not perform the same as one which is:

Judith ate sandwiches/Judith ate the sandwiches --> ?Sandwiches were eaten by Judith/The

sandwiches were eaten by Judith. As the passive sentences show, when the NP is not

quantified, passivizability suffers.

Van Valin (1993) also locates transitivity as a property of the verb, though he defines it

as "semantic transitivity." In Role and Reference Grammar (hereafter, RRG), there are two

types of transitivity recognized, macrorole transitivity and syntactic transitivity. The latter more

or less follows the traditional definition above, whereas the former is specifically related to RRG.

Macroroles are universal semantic structures which create the syntactic and semantic behavior

of participants in discourse. Though most familiar when expressed as "agent" and "patient," Van

Valin terms these roles "actor" and "undergoer" which subsume the more specific roles of agent

and patient. There are at most two macroroles represented in any sentence, and as little as

none. Further, it is the number of macroroles that has the most influence on syntactic behavior

rather than the number of "syntactic" arguments. Consequently, transitivity "is understood in

RRG as semantic transitivity, and is defined in terms of the number of macroroles a verb

takes: 2=transitive, 1=intransitive, and 0= atransitive" (Van Valin, 1993:48; emphasis added).

Thus, from the perspective of this study, regardless of the fact that Van Valin defines transitivity

in terms of universal semantic roles, he still defines it as a property of the verb.

Even the brief examination above yields two major criticisms of strictly verb-oriented, or

syntactic, approaches to transitivity. First, these approaches assume that transitivity is

structurally derived from the verb, and do not recognize the contribution of the object NP.

Second, because transitivity is a property of the verb, verbs with more than one instantiation

must be marked so in the lexicon, or have more than one representation in the lexicon. If indeed

transitivity is something relevant to the entire clause, then at some point, all clauses must be

included in the lexicon, seriously compromising the validity of maintaining it as a separate

grammatical component (Rice, 1987:34). On the other hand, syntactic approaches to transitivity

offer two valuable and time-honored insights. First, a direct object (of some sort) plays a role in

determining transitivity. Second, syntactic behavior often provide a litmus test for transitivity. A

sentence which passivizes is considered "fully" transitive. Sentences which marginally passivize

are not "fully" transitive, though may be distinguished from intransitive sentences which do not

have a direct object. So critical is this notion that Rice uses passivizability as the primary test of

transitivity. These two points are the contributions made to the study of transitivity by researchers

who employ essentially syntactic, or verb-based, definitions.

1.3.2 Clause-Oriented Theories of Transitivity

Clause-oriented approaches to transitivity are semantic in nature, and may be roughly

divided into formal models which deal primarily with propositions, and cognitive models which

deal with conceptual structures. As stated above, it is sometimes difficult to tease apart

semantic models and discourse models of transitivity particularly since the authors of the former

are often making claims for the latter based on the semantics in question. Propositional models

M6I'COk (1993) discusses transitivity in relation to voice. For him, there are two "levels"

of semantics at work in the clause: first, the "deep" level of semantic structure where

propositional information works; second, the "surface" level of semantic structure where

communicative effects are created. The propositional meaning is the "literal," actual,

communicatively unmotivated representation of the event-- and, as is formally common, is

"logically" read from the perspective of the verb (as a sort of basic-level meaning): V [x,y] (or

whatever representation is used). Each of the NPs is termed an "actant": the subject or actor is

the first actant and the object or undergoer is the second actant. Communicative meaning,

however, is not "literal" or without organizational intent, but is the "level" of meaning

representing the speaker's intent or organization or interpretation of the event (Giv6n's

'discourse-pragmatic" level). It is read, as is also common practice, from the actual or "surface"

order of the lexical units.

M6I'COk defines changes in transitivityy" as affecting communicative meaning but not

propositional meaning since voice is an inflectional category, but transitivation (his term) is NOT

(though sometimes closely resembles it): "Just like voice, the transitivation achieves

[communicative structure] through manipulation of COMMUNICATIVE SALIENCY of the verb's

actants "(M6l'C6k,1993:30; caps in original). Transitivation differs from voice in that voice can

affect any actant, where transitivation only acts upon the second actant under the condition that

its surface realization is as the "direct object". Furthermore, voice operates at the "deep" level

of semantic structure while transitivation only acts at the surface level. Ml6'COk follows Keenan

and Comrie's 1977 hierarchy concluding that "all things being equal, the higher the syntactic rank

of a sentence element, the higher its communicative salience...It is this trait that is exploited by

the transitivation: it allows the speaker to modify, according to his communicative needs, the

syntactic rank of the phrase whose communicative salience interests him" (30-1). Specifically,

then, Mi6'COk defines transitivation with the modification of the second actant (main object)

without affecting propositional content. Detransitivation works the same way, but with a two-

actant verb.

Giv6n (1993) also maintains different semantic levels in order to account for the

communicative effects of different sentences. As with M61'COk, he also calls them "deep" and

"surface" structure, though ostensibly these are syntactic levels. The deep syntactic structure

most closely corresponds to the semantic structure-- or propositional meaning of the sentence.

In simple sentences (main, declarative, affirmative, active), the relationship between

propositional meaning and surface syntactic structure is straightforward, hence semantically

transparent. In most complex clauses, though, the surface structure is at odds with the

propositional meaning, so much so that they are most easily described independently. In this

case, the semantic structure of the sentence is opaque.

Giv6n justifies these two claims pragmatically. The reason why syntactic complexity

results in propositional opacity is that a complex surface structure communicates "discourse-

pragmatic functions...discourse-pragmatics and propositional semantics are here in direct

competition for coding resources" (1993:30). Thus, the overall structure of complex sentences

is a communicative compromise between the distinct goals of communicating propositional

information and discourse-pragmatic information.

Giv6n goes one step further and assigns each of the "parts" of a sentence to particular

semantic functions. The word level encodes lexical meaning. The clause level encodes

propositional information. The discourse level encodes textual coherence. As a result,

transitivity acts across all three levels at once as instantiated at the level of the clause.

Syntactically, verbs and clauses that have a direct object are transitive. Thus, at a word level,

verbs which require direct objects are transitive, or clauses which have an overt direct object are

transitive. Semantically, transitive clauses encode prototypical semantic features, each of which

focuses on the subject of the verb (agentivity), object of the verb affectednesss), or construal of

event itself (perfectivity). Prototypically, the subject should be a deliberately acting agent; the

object should be concrete and visibly affected; the event should be bounded, terminated, fast-

changing, and in real-time. Since there is a good deal of variation possible with these features,

transitivity is potentially scalar in nature. The final dimension of transitivity is the discourse-

pragmatic: "By 'pragmatic' one means here that the very same semantically-defined transitive

event, coded by the very same combination of verb, agent and patient, may be rendered from

more than one perspective. By 'discourse' we mean the discourse context within which the

semantically-transitive clause is embedded" (1993: 46).

For Giv6n, as for Mdl'COk, transitivity acts on the "communicative" or discourse-

pragmatic level of language. The "semantic" level which communicates propositional structure

remains the same, but some rearrangement on the surface results in a change in perspective.

Giv6n goes one step further by adding the notion of prototypical semantic features.

Semantically, it is not the mere presence of subject and object which are sufficient, but a

particular kind of subject and object together with a particular kind of event. The prototypicality

of transitive events has been excellently investigated by Sally Ann Rice in her 1987 dissertation.

Before discussing Rice, though, a brief look at Hopper and Thompson (1980) is in order. It is

with this paper that transitivity as a complex and crucial grammatical structure re-entered

linguistic theory in full force. Hopper and Thompson (1980)

Most studies of transitivity are intended to explain either transitive/intransitive

classifications or active/passive alternations. A scalar approach, introduced in 1980 by Hopper

and Thompson, paved the way for an alternative understanding based on degrees of transitivity

rather than binary distinctions. In their approach, a clause could be high or low in transitivity

depending on the number and type (high/low) of prototypical semantic features it possessed.

They proposed more than just the three features distilled by Giv6n. In short, there are several

features pertinent to actions and objects and one feature critical to subjects. Each has a high

and low value. As the low value is simply the inverse of the high, I will only list the high values.

Of course, the first feature is that the ideal transitive clause should have two participants. The

ideal subject is high in potency. The ideal action is kinetic, telic, punctual, volitional, affirmative,

and real. The ideal object is totally affected and highly individuated. Individuation has

properties as well: proper, human/animate, concrete, singular, countable, and

referential/definite. Hopper and Thompson understand these features to be the component parts

which make up the "traditional" semantic notion of transitivity: "...[the] carrying over or

transferring [of] an action from one participant to another" (1980:253).

Hopper and Thompson also specify that languages vary according to which features (or

clusters of features) are critical to changes in transitivity. Further, they propose a relationship

between transitivity and grounding, claiming that clauses high in transitivity are also

foregrounded events in narrative. Several language specific studies emerged (see Hopper and

Thompson, Syntax and Semantics, 1982) which corroborated their claims. However, two

difficulties with their work have been pointed out. First, transitivity and grounding do not have an

absolute relationship and it may be necessary to refine which aspects of narrative are actually

related to transitivity, supposing that such a relationship indeed exists (Kalmar, 1982). Second,

the semantic features proposed do not necessarily covary consistently or correlate with

transitivity alternations such as passive, nor must they be morphological in nature (Rice, 1987).

There is also language-specific evidence that morphological indicators of transitivity may not

correspond to particular features, but simply indicate overall changes in transitivization (Kibrik,

1993). Nevertheless, Hopper and Thompson opened the door to a more sophisticated notion of

transitivity, and firmly established the relevance of both semantic and pragmatic factors to its

study. Cognitive models

For Rice, transitivity is ultimately "...a linguistic device optionally employed by a speaker

to conceptualize and organize the actions of entities in the world in order to convey a certain

attitude about an overall event to someone else" (1987:5). As with most theorists, she finds

transitivity a semantic matter with discourse-pragmatic functions. However, she firmly places

herself within a cognitive school of linguistics which takes the work on prototype theory by Rosch

quite seriously (see Lakoff, 1987):

In what is increasingly becoming the accepted view, categories are assumed
to be organized around prototypical or canonical instances. Less canonical
instances extend or radiate outwards from this prototypical center. In fact,
categories may have several centers or may be able to sustain several
prototypical instances. Furthermore, the creation of categories is understood
to be a dynamic process and one that grows out of contact and interaction by
a human with individual members of a category. Naturally, in the developing
infant, physical objects will be the first to be encountered, recognized, and
categorized. Over time, however, abstract entities like linguistic forms
and even events will be experienced, assessed, and categorized in
relation to some prototype or prototypes, which might enjoy either
experiential primacy, superior frequency or perceptual salience, a clear
utility, or higher resemblance relative to other members of the category
(Rice, 1987:4).

Rice is an "extreme" cognitivist, insofar as she understands the conceptual categories she

investigates as primarily non-linguistic, though they may be manifested in a linguistic form. My

view is somewhat more moderate. While I agree that conceptual categories do underlie much of

language acquisition (the research into conceptual categories throughout language is quite

convincing; see Vandeloise, 1984; Herskovitz, 1989; Johnson, 1987, etc.) I also hold that at

some point in the acquisition process, critical mass attains and language becomes a system

acting on its own. It is a cognitive system, yes, but also has its own rules, structures, purposes,

functions, and forms which are very much linguistic in nature. Nevertheless, Rice's line of

reasoning has led to significant insights into the nature of transitivity.

The prototypical transitive event, according to Rice, has many of the same

characteristics as mentioned above: the activity is unilateral, thus the entities are in an

asymmetrical relationship; contact between the entities is important; the second entity is affected

by the first; and the entities are distinct from one another, their locale or setting, and from the

speaker/conceptualizer. However, transitivity is as much a function of the describer's

interpretation of the event as it is the content of the event being described. It is this observation

that accounts for metaphoric extension from the realm of the physical as well as variability in

speakers' acceptance of passives. It is the first point which I find particularly significant.

In Rice's cognitive model, transitivity is prototypically a dynamic event in which the

participants are in an asymmetrical relationship and some change to the "object" is due to an

action/effect which traveled along some kind of directional path from the first participant to the

second. Thus, conceptually, several factors are crucial. First, the two participants are in an

asymmetrical relationship; in other words, one participant must have the means to affect the

other or one participant must have the capacity to be affected. Second, the action must be

kinetic and/or forceful. Third, there must be some kind of path construed between the

participants, even if the activity itself takes place instantaneously. Fourth, the receiver of the

action must be affected in some way, though this can cover a wide range of possibilities,

including change of state, change of location, etc. Of course, the change itself may be

construed from experience. In the well-worn example, John hit Bill, the change to Bill is not

made obvious in the sentence but construed from our experience with the world.

In the cognitivist model, reality is constructed, paving the way for important links

between language and cognition. One of these links is metaphoric extension from "bodily

images" (Johnson, 1987) to non-physical domains. Rice brilliantly makes this point for

transitivity in mental domains and communicative or interpersonal domains. That is, for each of

the primary domains in which we act -- physical, mental, social -- there are prototypical transitive

events which make use of the above semantic components (though inequally). Her examples

serve to demonstrate the most typical aspects of each event.

In the physical world, John deliberately kicked the sleeping poodle, represents two

maximally distinct participants in a single event in which one participant is active and powerful

and the other is passive and defenseless. There is forceful contact which travels along a path

designated by the physics and geometry of the body in one direction. The dog is presumably

both awakened and injured, inducing both an internal and external change of state.

In the mental domain, Rice uses the following examples: Eileen heard the gunfire, Steve

despises Christmas, John solved the equation. In each, there is at least one active, animate

participant, but no restrictions on the type of passive participants. Further, each of the active

participants acts on an entity external to itself. The contact is inferred as a function of the active

participant and the event in question. However, the change that takes place is internal to the

active participant rather than the passive participant; the nature of gunfire, Christmas, and

equations does not change. Thus, in mental domains, the path is understood to "circle back" to

the active participant. This is really not as odd a relationship as it first seems if we consider how

we ascribe information to participants. In Eileen heard gunfire/Gunfire was heard by Eileen, the

state of our knowledge about Eileen has changed, just as the state of our knowledge about the

poodle changed in the example above. In mental domains, there is still change and still two

distinct participants, even though the relationships among them differ somewhat.

In the social domain, Rice cites The policeman questioned Bill. Again, there are two

participants, and while both are human and animate, their distinctiveness rises out of authority

relations. The action is presumably goal-directed; the questions travel a "path" from the

policeman to Bill, not the other way around. A number of possible emotional or mental changes

can be inferred for Bill, from fear and nervousness, to sweating, to silence. Rice also makes the

point that contact and directedness are important semantic properties of social events, and her

investigation into verb and preposition combinations resulting in transitive events contains a high

number of social events.

The strengths of the semantic theories of transitivity include a commitment to finding

coherence underlying apparently disparate examples, and a careful attention to details that

systematically distinguish different types of meaning. While I don't necessarily accept the

syntactic distinction of deep and surface levels, theoretically there is a need to distinguish some

kind of propositional meaning from discourse-pragmatic meaning. Nor do I accept prototype

theory outright, though it is convenient to our discussion to use it as a possible model. This

model will be taken up with greater diligence in Chapter Two. Nevertheless, Rice's careful

explication of transitivity in English using both canonical and marginal examples is compelling.

The primary strength of Giv6n and Rice is their commitment to a view of transitivity as a single

dynamic entity with a complex of components versus a pick-and-choose list approach. The

notion that structurally, transitivity has both semantic and discourse-pragmatic components, and

functionally, it serves to manipulate perspectives on a scene, has profoundly influenced my own

understanding and is foundational to this paper.

1.3.3 Extra-Clause Oriented Theories of Transitivity

Extra-clause oriented theories of transitivity focus on the level of discourse. These

theories of transitivity begin with the notion of construal of an event and end with the notion of

information structuring in a text. There are two levels at which an event can be construed. First,

the clause-level, in which the structure of the individual clause encodes the speaker's construal

of a particular event. Second, the discourse-level, where the structure of an individual clause is

construed relative to other events in the text. Rice deals primarily with the first of these in her

dissertation: the particular ways a speaker represents a particular event vis-a-vis the event and

the conceptual structures the speaker has access to. Giv6n makes use of both in his model.

First, there is the propositional structure of the event which does not change. Second, there is

the particular construal of the event relative to the speaker's intentions and contextual

environment. The second notion is also what Hopper and Thompson were discussing when they

claimed that highly transitive clauses encode foregrounded events in narrative. (In fact, this is

not strictly the case. Rather, highly transitive events typically correlate with temporally-grounded

events in a narrative; those actions which drive the story forward through time and require both

an agent and patient of some kind.)

The pragmatic side of transitivity is the least explored, as the relative paucity of

explanation given by Giv6n attests. However, there are some tantalizing possibilities offered in

various accounts of the relationships between discourse and grammar. DuBois (1987) discusses

the statistical correlations between clause-type (transitive/ intransitive with lexically specified

arguments) and information structuring/flow (particularly the introduction of new participants) as

they relate to ergative patterns in grammar. There was an overwhelming tendency in his

transcripts of narrative discourse for any clause, intransitive or transitive, to have only a single

lexical argument fully specified (only 2.8% of transitive clauses had both arguments lexically

specified, representing just five sentences in the data). His conclusion was that ergative

patterning in discourse forms the basis of ergative patterning of grammatical phenomenon.

DuBois works within Dixon's (1972) semantic division of participant roles, intended to

allow cross-theoretical discussion of ergative/absolutive and nominative/accusative languages.

Dixon discriminates three basic semantic roles as they relate to the predicate: "A" designating

the subject/actor of a transitive (two participant) clause; "0" designating the object/goal of a

transitive (two participant) clause; and "S" designating the single participant of an intransitive

(one participant) clause. Thus, in nominative/accusative languages, "A/S" pattern together,

meaning that the actor/subject of a transitive clause and single participant of an intransitive

clause take the same verbal morphology (nominative). It is the object/goal relationship which

takes accusative marking. On the other hand, in ergative/absolutive languages, it is the single

participant of an intransitive clause and object/goal of a transitive which pattern together and

take absolutive morphology, while the actor/subject of a transitive clause takes ergative marking.

DuBois maintains that part of discourse management is a restriction against using more

than one lexical specification per clause. When broken down according to ergative/absolutive

cases, only 6.1% of 619 lexically specified arguments occurred in the "A" position-the vast

majority of cases were "S" and "O". Thus information flow was managed in terms of a constraint

against using lexical transitive subjects. DuBois interprets this pragmatically as meaning only

one new argument can be introduced at a time (this notion is also found in Giv6n, 1990a/b and

Chafe, 1994, though both authors suggested as much in earlier publications). Further, the new

argument is preferably an intransitive subject or transitive object. In terms of the discourse basis

of ergativity, DuBois concludes: "It appears, then, that speakers often select an intransitive verb,

not necessarily for its conceptual or semantic one-placeness, but for its compatibility with

constraints on information flow" (1987:831). DuBois also found evidence for the same type of

patterning in nominative/accusative languages; my own work in Cree suggests that statistically,

clause type and full lexical specification of NPs correlates with function in narrative discourse.

Yet another discourse perspective on transitivity is presented by Berman and Slobin

(1994). They conducted a cross-generational, cross-linguistic study of the acquisition of

narrative competence. Briefly, Berman and Slobin propose that in a skillful narrative, events are

not presented in a simple linear chain of successive occurrences in time and space. Rather,

events are "packaged" into hierarchical constructions. Part of that packaging arises from a

filtering process whereby speakers construe events over four dimensions: topic, locus of control

and effect, event view (Cause, Become, State), and degree of agency (1994:519).

Topic represents the speaker's choice of protagonist, the participant from whose

viewpoint the narrative is constructed or who the listener is intended to have "empathy" for (in

Kuno's sense of the term). Loci of control and effect represents the energy source of the event

which may or may not be the same as the topic. In a passive sentence, for instance, with an

overt "by-phrase," the affected participant is the topic but the energy source for the event is also


The selection of event view is the speaker's means of representing the semantic nature

of the event-as-process. A Cause-View represents one participant causing a change of state in

another. A Become-View represents a change of state without mention of a causative force. A

State-View simply represents a state itself. These three views are strikingly similar to Halliday's

three clause types, particularly as they interact with locus of control and effect.

Degree of Agency is the means by which the speaker represents the overall dynamism

of the event: "...its dynamic and motivational loading, the extent of its consequences, etc."

(Berman and Slobin, 1994:519). Agency can be strengthened or weakened via choice of verb,

choice of adverbs, and use of subordinate clauses. Berman and Slobin propose three levels of

agency: high, mid, and low. Any particular level comes about via the other three aspects of

filtering as well as voice and lexical options. In their treatment, though, the choices made when

"filtering" an event for expression go beyond mere representation of the experiential event (as is

the case with Halliday and Rice) and have encoding consequences beyond the level of the

clause. For example, construing the following participants and event with different combinations

of topic, locus, event view, and agency impacts the content of the next clause.

Participants: dog, cat event: fast movement of one participant from the other
(1) Topic: dog (2) Topic: dog
Locus of control and effect: dog Locus of control and effect: cat
Event View: Cause-View Event View: Become-View
Agency: high Agency: mid

Discourse One: The dog attacked the cat, barking furiously his right to sit on that bit of
sidewalk next to his house. The cat swiped his nose once with her paw, and that was the
end of that day's battle.

Discourse Two: The dog fled, nose scratched and bleeding. He had to reconsider battle
strategy before challenging the feline next door again.

The impact on the second sentence of each is significant. When the dog is both topic

and loci of control and effect, it left the cat as not only the affected participant but also the new

information. The response of the cat is the next "logical" step in the discourse as it is this

participant the listener has least information about and must know in order to make pragmatic

sense of the following information. In Discourse Two, however, the cat is the controlling force

but is not overtly mentioned, leaving the dog as a possible topic for the next predication.

We are thus brought full circle -- from Rice's compelling arguments for the conceptual

basis for transitivity across physical, mental, and social domains, DuBois' evidence for the

effects of discourse management on clause structure and Berman and Slobin's insights into the

interplay from clause to clause -- to the original hypothesis that transitivity is not a unitary

phenomenon but one which straddles the three principle levels of language structure--syntax,

semantics, and discourse--for the purposes of communication. What remains to be done is a

thorough-going investigation into these interactions. While each of the above researchers has

admirably investigated the particular side of the issue he or she is interested in, no one

investigator has attempted an in-depth study of these at once, particularly as they attain beyond

the genre of narrative. That is the purpose of this project: to investigate the various definitions

of transitivity and particularly its use with the aim of determining what might properly be called

transitivityy" and its functions. This work should be viewed with respect to emerging notions

about discourse such as those laid out by Hopper and Thompson (1980), DuBois (1987), Berman

and Slobin (1994) and Chafe (1994) which hold that the grammatical system itself is closely

linked to discourse forces such as focus and topicality, and the structure of clauses is similarly

linked to the needs of information structuring, cognitive processing, and communication. Thus

the play of any clause is not only a matter of semantics and construal, but also constrained by

the needs and forces of discourse and cognition.

1.3.4 Transitivity and Text

Because transitivity has implications across the semantic, syntactic, and discourse-

pragmatic levels of language, it has been the subject of several discourse studies, as the section

entitled "Extra-clause oriented" demonstrates. The overwhelming majority of these studies have

taken place within the genre of narrative, both spoken and written. Initially, this seems like a

good strategy. Narrative is an intuitively well-understood discourse type, easily recognized and

produced by speakers, with a few clear linguistic structures associated with it; namely, narrative

is past tense, action oriented, and proceeds from beginning to end chronologically.

Nevertheless, while linguistics has long been cautioned by practitioners such as

Longacre (1977, 1983) to be mindful of the connections between notional type and surface

structure, this has rarely been the case. While linguists do take pains to mention the notional

type they are culling data from, there is left implicit the assumption that the language itself-- the

grammatical or semantic structures in question-- are wholly distinct from the environment within

which they are occurring. Even discourse linguists who devote their time to studying these

connections seem to carry this same hidden belief: while genre has an effect, it is more of a

loose framework around the language within; it has no real impact on structure itself. In other

words, a given grammatical or semantic construction has structure and function which is

defining, and the structure and function hold regardless of the text type in which they are


Of course, not every linguist holds this assumption. DuBois' (1987) study discussed in

the previous section is an excellent counter-example. His work on the discourse basis of

ergativity specifically comes to the conclusion that there are discourse preferences for syntactic

structures. These preferences act independently of the semantics of syntactic structures by

exerting selectional pressures for the purposes of cognitive processing. In other words, there are

"top down" linguistic forces at work which are every bit as powerful as the "bottom up" pattems

we are accustomed to assuming for linguistic structure.

Thus, a careful distinction must be drawn between the structure and function of any

grammatical construction and the notional type in which that structure occurs. Further, attention

must be paid to teasing apart the linguistic features characterizing a discourse type which may

be having an effect on surface structure. This has not been done for transitivity and narrative.

Many of the proposals for the structure and function of transitivity have been made without

sufficient regard to the effects of narrative. This difficulty will be addressed in three ways in this

project. First, a "cumulative" structure/function definition of transitivity will be proposed followed

by a point-by-point investigation of the validity of each of the claims. Second, the linguistic

features of narrative will be investigated and subtracted from the definition of transitivity where

that seems appropriate. Third, the resulting "definition" of transitivity will be taken into both

narrative and expository texts for refinement, and to investigate the functions) of transitivity in

these two contrasting text types.

1.4 Overview of Proiect

A couple of assumptions central to this project should be expressed at this point. Of the

various linguistic meta-approaches to language, I fall in the camp of the interactionistt." This is

my own designation and includes perspectives such as those expressed by DuBois (1985a:363)

that "grammars code best what speakers do most" and by Elizabeth Bates (pc) that language as

a system is characterized by the mapping of a "hyper rich dimensional system" (meaning) onto a

"low dimension communicative apparatus" (grammar) resulting in a dynamic system wherein the

vast, fluid world of meaning must be relayed via a comparatively rigid means of communication.

Talmy (1994) as mentioned earlier says basically the same thing: language serves the needs of

communication but is constrained by the cognitive requirements of comprehensibility.

For these reasons, I reject models of grammar which propose various sorts of endlessly

embedded continue, imposed upon one another, and resulting via this imposition in a coherent

set of constructions with surface realizations of clauses and phrases. This is the "feel" of many

purely cognitive or purely discourse models of grammar. Instead, I side with likes of Bates,

Talmy, Giv6n, and Slobin who envision grammar as the dynamic system which emerges from

the tension of two distinct forces in linguistic communication: the richness of meaning as

expressed through the constraints of cognition.

Thus, my model of transitivity by and large rejects the prototype models of transitivity

currently in favor (Hopper and Thompson, 1980; Rice, 1987; Giv6n, 1995). Instead, I propose a

model of transitivity which has its base in the semantic-cognitive system but is realized

syntactically through four different clause types: telic transitive, atelic transitive, dispersed

transitive, and intransitive (or durative transitive). These clause types are particular to English in

keeping with Slobin's proposition that different languages select different "aspects of the mental

image" to be realized grammatically (1996:72). In Slobin's words (1996:76):

.The expression of experience in linguistic terms constitutes thinking for
speaking a special form of thought that is mobilized for communication.
Whatever effects grammar may or may not have outside the act of speaking,
the sort of mental activity that goes on while formulating utterances is not trivial
or obvious, and deserves our attention We encounter the contents of the mind
in a special way when they are being accessed for use. That is, the activity of
thinking takes on a particular quality when it is employed in the activity of
speaking In the evanescent time frame of constructing utterances in discourse
one fits one's thoughts into available linguistic frames 'Thinking for speaking'
involves picking those characteristics of objects and events that (a) fit some
conceptualization of the event, and (b) are readily encodable in the [native]

This perspective does not contradict Rice's claim that "...a verb/proposition/event is not so much

transitive as it considered transitive by the speaker/conceptualizer"; rather, it constrains this

claim by proposing that transitivity has language-specific surface realizations according to the

needs of communicative coherence. The model of transitivity will be taken up primarily in

Chapter Two.

Functionally, the two principle claims made for transitivity is that it manages perspective

on participants and events and signals foregrounded information. My own research confirms the

first conclusion and disconfirms the second. Transitivity does indeed manage participants and

events in narrative, but the second claim is more an accident of the interaction of transitivity and

narrative than a quality of transitivity itself. These issues will be introduced in Chapter Two and

developed in Chapter Three.

When transitivity is explored in expository text, our assumptions about the centrality of

function as part of a definition of a structure is challenged. Since expository text does not

employ participants and events as its main notional structures, transitivity plays a different role

than in narrative. In particular, transitivity is employed functionally at textual points of transition,

signaling movement from one major part of the text to another as well as textual material which

is off the main argument line. This subject is the task of Chapter Four.

The final chapter concludes the project by examining the role of transitivity in narrative

and expository texts and hypothesizing how these results either support or challenge traditional

assumptions about the interaction of notional type, surface structures, and grammar as a system.


2.1 Introduction to the Problem

This project assumes that grammar is the system which emerges to mediate the rich

world of internal representation with the particular cognitive and production constraints of

communication. The purpose of this chapter is to propose a working--and workable-definition of

transitivity. Presently, there is a wide range of possibilities in the literature which need to be

evaluated, amended, salvaged, or discarded. Major claims frequently made for transitivity which

need particular attention include: first, transitivity is or has a structural prototype; second, there

are degrees of transitivity; third, the passive construction is the only legitimate test for transitivity;

fourth, transitivity is conceptual, not linguistic; fifth, because transitivity has structure, that

structure is universal in nature; sixth, transitivity functions to foreground information in texts. A

more complete compilation of the extensive claims made for transitivity can be found in the

"Statement of Transitivity' on the following page. This statement will provide an organizational

blueprint for the initial discussion.

A second issue must be dealt with as well. Most linguistic research which takes context

into account has been done within the genre of narrative. Transitivity is no exception; it has

been investigated exclusively through written or oral narrative. This has had some unfortunate

consequences for understanding the function of transitivity since narrative itself is a genre

marked by unique configurations of linguistic structure (see Fleischmann, 1990). In order to

understand what is happening with transitivity at a discourse level, the linguistic structure of

narrrative--in particular, tense and aspect--must be separated from whatever it is that transitivity

is doing. Once this has been accomplished, it will be much simpler to investigate the function of

transitivity in both narrative and non-narrative texts. The bulk of this task is taken up in Chapter

Three, but the work is begun in this chapter.

We begin with the "Statement of Transitivity" as the starting point of our investigation

into what properly belongs to transitivity and what properly belongs to narrative. Excerpts will be

used where appropriate to validate points although the bulk of Chapter Two is dedicated to

transitivity at the level of the clause, i.e., the syntactic and semantic claims. The discourse and

functional nature of transitivity will be taken up primarily in Chapter Three and will include more

extensive use of texts.

Figure 2.1: Statement of Transitivity
Structure-Function Definition: Transitivity is a complex phenomenon involving syntactic,
semantic, and pragmatic components which functions to conceptualize and organize events and
participants in the world so that their relationships to one another may be communicated
according to the attitudes and intentions of the speaker in a manner which is cognitively coherent
and manageable, and which is grammatically verified by the availability of the passive form.
The above are understood to be (possibly) prototypical in nature, thus scalar and capable of
metaphoric extension as well.

Structural Components
Syntactic- There is one syntactic component to transitivity and a necessary property
of that component, syntactic manipulation is the principal grammatical
test for the fully transitive statement:
(i) there must be at least two participants structurally bound by the same
clause; the two participants must be distinct entities "in the world"
(ii) the availability of the passive alternation is the strongest grammatical
test for transitivity
Semantic- There are four semantic relationships identified for transitivity:
(i) Asymmetry the participants must be in an asymmetrical relationship
generally due to the unidirectional nature of the action, but
sometimes falling out from the inherent nature of the participants
themselves or their relationship
(ii) Agentivity one of the participants is a deliberately acting agent,
capable of instigating the action or is the source of the action
(iI) Affectedness one of the participants must undergo some kind of
change or experience some kind of effect due to the action; this change
or effect may be understood from world knowledge or inherent to the
the meaning of the action itself
(iv) Perfectivity the action should be bounded, terminated, fast-changing,
and take place in real time
Discourse-Pragmatic- There is one pnncipal discourse component to transitivity:
(i) Constraint of Information there must be a limited number of participants
and forms of expression of those participants

Functional Claims
There are four proposed functions of transitivity:
(1) Propositional meaning the semantic components of the transitive clause serve to create the basic,
propositional, non-communicatively influenced meaning of the expression
(2) Perspective transitivity permits the language user to express a given situation,
action, or event from more than one perspective in order to create the desired communicative effect
(3) Management of Information Flow transitivity provides the site over which participant relationships
are arrayed in order to constrain the introduction of new information during communication, in part
maintaining and managing communicative coherence
(4) Grounding high or prototypical transitive clauses correlate with foregrounded information

2.2 Syntax

The claim for syntax is that there must be at least two participants who are distinct in the

world (in other words, are not co-extensive) and bound structurally within the same clause. The

necessity of two participants is a claim made by nearly all linguists regarding transitivity, even

those who propose a scalar model. Hopper and Thompson (1980) do make the strongest

attempt at a scalar model, but to a great extent still regard High transitivity with two participants

as the prototypical and basic level against which less transitive statements should be measured.

If transitivity is to be a useful designation, then the simple binary opposition of transitive/

intransitive where the former is transitivityy" (in which two participants is a necessary but not

sufficient condition) and the latter is not transitivityy" (because it is not transitive) needs

amending. This is not simply a matter of eliminating terminological confusion but of creating

theoretical clarity. There are three options possible: first, keep the term transitivityy" for the

theoretical construct and use other terms for thebinary opposition itself, such as valency (if,

indeed, the binary opposition holds true); second, use some other term for the theoretical

construct, perhaps "argument structure" and.keep the terms transitive and intransitive (same

caveat as above); third, keep the term transitivijy" for the theoretical construct and reform our

understanding of that construct so that it is useful for linguistic analysis. Option three is the

choice of this study.

Initially, I will follow Rice, Hopper and Thompson, Givon, Halliday, and Berman and

Slobin in proposing that transitivity is a scalar phenomenon. However, I will take this to its

logical extreme by avoiding the traditional notion that transitive statements must have two

participants, claiming rather that if there are degrees of transitivity, then the number and nature

of participants is simply one component. In order to substantiate this claim, first the "nature" of

the participants must be made clear (as it turns out, "co-extensive" is more a matter of construal

than strict pragmatic or logical "in the worldness") as well as the "nature" of the activity (to be

discussed further in the semantics section). Second, what is meant by "scalar" or "degrees" must

also be made clear and useful.

2.2.1 Grammatical Behavior

In order for "degree of transitivity" to be more than intuitively convenient, evidence must

be found in the grammatical behavior of languages themselves. For an issue such as transitivity

and the broad claims made for it, the grammatical behavior should be found not only at the level

of the clause, but also at the level of discourse. The first question is what counts as the clause-

level behavior of transitivity?

In English, the attraction of the active-passive alternation is the clarity and economy of

their connection. The passive is so obviously a structure which "re-expresses" an event by

shifting the topic-focus relationships (a discourse phenomenon) while maintaining role

relationships (a semantic phenomenon). The popularity of the passive as a teaching tool and

starting place is the simplicity and elegance of the transformation. Are there other structures

equally instructive? For example, what distinguishes the passive from the middle voice? Simply

speaking, there is a single "fact" with grammatical consequences: in the middle, the force or

effector role is not available syntactically which opens the door for alternative expressions not

allowed by the passive.

(1) Olaf opened the door/The door was opened (by Olaf)
(2) The door opened/ *The door opened by Olaf
(3) ?Olaf opened the door with a creak/?The door was creaked open (*by Olaf)
(4) The door opened with a creak/The door creaked open
(5) Olaf, opened his, door/His, door was opened (*by Olaf,)
(6) His door opened/Olafs door opened
(7) His door creaked open/Olafs door creaked open

In these sentences, the middle construction makes it possible to incorporate manner into

the verb (a productive process in English), whereas the passive partially or completely blocks

this movement. That is, in a passive sentence, although the affected participant is the subject, it

is still not the source or effector of the activity (the original roles are maintained). Manner cannot

be attributed to the verb when the locus of force is either implied or stated in the adjunct. In

other words, the conceptual iconicity in an active "transitive" sentence is violated in the passive

but preserved in the middle. Only when the single overt participant is both the topic of discourse

and the locus of force of the verb can manner be incorporated. In the middle voice, this is the

case: the single participant is the only participant, and is the discourse topic, grammatical

subject, and locus of force. In many languages, the middle is marked by a distinctive case, such

as absolutive, further distinguishing it from "normal" subject-verb-object relationships.

The possessive also has a powerful effect grammatically. Although there appear to be

two distinct entities in the world (certainly Olaf and the door are not one and the same thing),

they are not construed as distinct, as evidenced by the non-availability of the passive. In the

middle voice, though, there is a single effector/affected, and the possessive does not interfere

with the predication options available to it.

Thus, it can be argued that in English, there are both syntactic and discourse

consequences when the relationship between participants and event is differently

construed. Simply shifting topic-focus relationships does not induce a change in participant-

event relationships. Rather, there are grammatical constructions which instantiate changes in

these relationships which in turn have consequences for other levels of grammar. Therefore, it

is theoretically palatable to label participant-event relationships differently, and this "label" is

called transitivityy."

As the examples above show, it is not merely the number of participants bound by the

predicate, but the nature of that relationship which affects grammar. On the other hand, the

passive alternation is only available to clauses with two distinctly construed participants. Hence

the number of participants is not irrelevant, either. If the statement bolded above is reasonable,

then it's necessary not to limit the discussion of transitivity to only those sentences with two

participants. An alternative must be found which accounts for participant-event relationships,

includes a place for two-participant constructions, but is not limited to them. Immediately, as we

shall see, an appeal has to be made to semantics.

2.3 The Semantics of Transitivity

As the brief discussion of transitivity and syntax shows, "syntax" is not the crux of the

problem. Semantics, on the other hand, is quite the opposite. There is a bewildering array of

semantic explanations of transitivity traversing a number of theoretical approaches. For some

theories, such as Role and Reference Grammar (Van Valin, 1993), transitivity is a straight-

forward matter of semantic macro-roles and not that complicated a matter. For the cognitive

camps, however, transitivity is--or is related to-conceptual structure and is quite complex.

Currently, cognitive models of transitivity are producing considerable results so it is to these

models our investigation turns.

There are, by and large, two principle processes currently competing for attention in

cognitive theory: prototype theory and implicature. Prototype theory (Rosch 1973, 1975, 1977,

1978 and 1981; Lakoff, 1987) proposes that categories are organized around prototypical or

canonical instances. In some cases, the canonical example itself is the prototypical image

against which categorization is made. In other cases, characteristics are abstracted from

canonical members and it is these characteristics which form the prototype. Implicature (Grice,

1975) is a different process altogether. Implicature is not concerned with the organization of

categories but with the processing of information. This is "linguistic cognition" (Giv6n, 1995)

which takes place between overt linguistic information (such as the sentence) and covert

information (such as the knowledge brought to that sentence by speaker and hearer).

Implicature may be defined as "stable linguistic inferencing," a linguistic-cognitive process which

is part of normal linguistic functioning.

Prototype theory has enjoyed quite a bit of attention and many theorists subscribe to

prototype analyses for an enormous range of linguistic behavior. Implicature is less used, yet is

probably the more appropriate choice for much of what passes for discourse-level and pragmatic

processes. A closer look needs to be aimed at some of the uses of prototype theory. For

example, does it really make sense to say there is a prototypical sentence type? A basic model

against which all sentences are created and judged? The arguments for such are based on

cognitive models, such as Lakoffs "idealized cognitive models," and these kinds of highly

abstract grammatical schemas are generally supported by those who study the development of

the human conceptual system (Mandler, 1991). However, there are equally compelling reasons

to believe that at a certain point of complexity--i.e. discourse-- prototype theory is an insufficient

or an inappropriate account for linguistic behavior; it is not that prototype theory is wrong, but

rather that it is applied with too broad a brush stroke.

2.3.1 The Active, Affirmative, Declarative Prototype

Giv6n (1995) proposes quite forcefully that there are three dimensions essential to the

semantics of transitivity corresponding to the "prototypical transitive event" (76):

Agent: The prototypical transitive clause involves a volitional,
controlling, actively-initiating agent who is responsible for the event, thus its
salient cause.
Patient: The prototypical transitive event involves a non-volitional,
inactive non-controlling patient who registers the event's changes-of-state,
thus its most salient effect.
Verbal Modality: The verb of the prototypical transitive clause
codes an event that is perfective (non-durative), sequential (non-perfect) and
realis (non-hypothetical). The prototypical transitive event is thus fast-paced,
completed, real, and perceptually-cognitively salient.

Giv6n explicitly justifies his definition by appealing to methodology: "... one studies the

pragmatics of voice while holding the transitive semantic frame more or less constant...For this

reason, I will exclude from the discussion here a group of de-transitive voice constructions that,

in one way or another, tamper with the semantic core of the transitive event" (76, emphasis

added). There are several problems with the above statements which will first be noted, then

taken up one by one. The discussion below intends to investigate more closely the claims made

for a semantic-transitive core, one which Giv6n gives the most clear and concise expression to

and to which most researchers in this area appear to subscribe.

The first difficulty with this approach is not the methodology itself but the foundation

upon which the method is based: that there is such a thing as a single, unitary transitive

prototype against which ALL clauses not exhibiting the semantic prototype should be assessed.

It is, in essence, a non-enlightening definition, claiming that all sentences which are not

"perfectly" transitive are not so because they do not have the defining components of the

transitive event which make for prototypical transitivity. At issue here is not whether voice

alternations exist, or whether they can be compared, but whether all should be related back to

the same semantic core.

The second difficulty is in the labeling of "agent" and "patient" for most salient cause

and effect. Cognitively, the most salient cause and effect may be quite distant from the

semantics normally associated with "agent" and "patient," namely that of a human agent and a

non-human patient. As we will see, there is interesting evidence showing that "agent" is too

strong a role to be at the core and "patient" has a dine of its own which affects grammatical

behavior. Further, there is also evidence that the non-human patient, at least in English,

participates in a unique structural pattern which should be included in the array of options

accorded transitivity.

The third difficulty is actually comprised of many sub-difficulties relating to the notion of

verbal modality. For the most part, these stem from the association of transitivity and narrative,

even among researchers who simply wish to investigate the clause. That is, there is a close

historical connection between the study of transitivity, the study of voice and modality, and the

study of narrative. The result has been some blurring of the lines separating each. For

example, there is relatively little evidence when sentences are studied in isolation that

perfectivity is a necessary component to semantic transitivity. That is, relatively durative

activities still enjoy many of the de-transitivizations that Giv6n discusses. However, in context,

many such activities are grounds against which more punctual events occur. Second, the notion

of "sequential" is also problematic since frequently the inherent semantics of a given event do

not suggest definition as a time-bounded entity. Rather, sequentiality is inferred as a matter of

temporal juncture between two events, suggesting instead that boundedness by time is an

implicature. Finally, there is an array of data showing that the perceptual saliency of the

transitive event is a function of two participants interacting, a relatively limited occurrence

textually. In fact, the salience of any particular event is figured not by its number of participants

but by the event itself relative to other events in context. This necessarily includes many events

which are not transitive in the sense of having two participants.

2.3.2 The Transitive Prototype: One or Many?

Any discussion of a transitivity prototype necessarily begins with the distinction between

event structure and transitivity. Event structure is a pre-linguistic cognitive structure

representing a conceptual stage of processing whereby portions of the stream of experience are

excerpted or "stamped out" into a particular sub-unit with particular characteristics. Event

structure is eventually expressed via the grammatical unit, the clause. But what characterizes

event structure? What evidence indicates that it is pre-linguistic? And how does it relate to


In part, the notion of event structure arises from the need to account for language

acquisition. That is, at some point in the child's development, the non-linguistic world inside the

head gets "mapped onto" a linguistic world through which it is expressed. Developmental

linguists, such as Slobin (1972, 1982, 1994, 1996, 1997), have begun to address what kinds of

concepts must be represented at the onset of language acquisition with regard to what actually is

expressed in the early stages of speech. Mandler (1991) argues "Language is unlikely to be

mapped directly onto sensorimotor schemas. There is a missing link: a conceptual system that

has already done some of the work required for mapping to take place" (414). Specifically, she

argues that it would be "extraordinarily difficult" for the child to make a leap from the continuous

stream of physical data to the symbolic representational world of language. Some mediating

system intervenes to parse "...the movement involved in picking up one object and placing it into

another to the statement, 'The marble is put into the cup.' Some kind of conceptual summary of

what is happening in this situation is needed; it is this conceptual summary that is mapped,

rather than the sensorimotor schemas themselves" (414).

Mandler notes that the abstract schemas proposed by cognitive linguists such as Lakoff,

Langacker, and Talmy represent the same kind of material that psychologists investigating the

foundations of the human conceptual system also propose. First, she cites evidence that infants

are able to parse the world into coherent bounded objects as early as 3 months old, if not sooner

(415) and that they are able to encode the causal relationships into which objects enter as early

as 4-6 months (416). She is careful to state that while the extent to which these accomplish-

ments are perceptual versus conceptual is unclear, it is clear that the information is available to

be operated on.

Mandler shows that the kinds of concepts which seem to be formed in infancy are well-

represented by image-schema notions. Together with evidence from Slobin inter alia it is also

clear that many of these notions apply directly to event schemas and their linguistic expression.

First, infants are able to distinguish objects that engage in "biological motion" from those that do
not. They are also able to distinguish caused and non-caused motion. Mandler cites the work of
Bertenthal, Proffitt, Kramer and Spetner (1987) and Leslie (1984), respectively for each of these
points. Thus there is a sensorimotor capability of distinguishing motion which is self-caused from
motion which is other-caused. She further argues that while the image-schema of PATH is
indeed simple and basic, paths do not happen without starting points; hence one kind of image
schema emerges for the notion 'self-caused motion" designating the simple idea that an
Endpoint does so from a point of origin outwards along a path. A second image-schema for
animate motion characterizes the trajectory itself--when an Endpoint moves, it does so in a
particular manner. The two schemas simply distinguish an Endpoint-oriented image of motion
from a trajectory-oriented image of motion.


Figure 2.2: Mandler's Image Schemas of Motion

Finally, she (420) also notes that infants are "inordinately attracted by all moving objects,
and of course, many of those are inanimate... Inanimate objects are those that begin motion only
through the interaction of other objects; specifically, they are caused to move" [emphasis added].

Thus a third image-schema emerges outlining caused motion:


[o--> -1

Figure 2.3: Mandler's Schema of Caused Motion

Mandler (420) introduces one more leap which brings us a step closer to what happens

at the level of linguistic representation. She combines the schemas of animate motion and

caused motion to produce a schema which she says accounts for the notion agency:


Figure 2.4: Mandler's Image Schema of Agency

This schema combines all the above aspects together. An animate Endpoint causes itself to

move and in doing so, moves an second Endpoint. The result is a complex image-schema

sensitive to both Endpoints of an event as well as the trajectory.

While Mandler admits that we do not know when such a conceptualization arises in

infancy, she cites Leslie (1982,1984) which shows that infants between 4 and 7 months old show

surprise responses when animate objects move inanimate objects without contact between the

two while also showing no surprise when inanimate objects do so. In other words, by 7 months

there appears to be evidence for some basic notion of agency as it relates to animate-inanimate


Linguistically, there is evidence that as children acquire language, they are sensitive to

the above distinctions. Developmentally, the various parts of motion are among the first to be

expressed linguistically. Slobin (1982) cites an example with the ergative language, Kaluli. In

Kaluli, there is an ergative noun suffix which attaches only to agents of actions, that is actions

where an agent acts upon another human or non-human and causes a change of state or

location. The ergative suffix is attached to sentences such as Mother is cooking but not to

sentences such as Mother is sleeping. Two remarkable facts occur developmentally. First, as

early as 26 months old, the child consistently marks the ergative suffix in two word utterances,

particularly with highly dynamic verbs such as give, take, grab, and hit. Second, the ergative

suffix is never over-extended to the subjects of sentences such as Mother is sleeping.

On the other side of the grammatical fence, additional evidence is found in accusative

languages. In Russian, for example, the accusative suffix attaches to any object in an

affirmative sentence regardless of the type of action involved. Thus, "book" in I read the book, I

saw the book, I threw the book all receive the same accusative inflection. If the inflection simply

relates to the grammatical notion "object of activity", then it should be applied equally across all

instances as it is being acquired. However, Slobin cites a study by Gvozdev (1949) which shows

this is not the case. Children did not acquire the inflection equally, but applied it first to a

particular subset of events, namely those involved in direct, physical action on things: give,

carry, put, throw.

Slobin concludes that the Russian underextension makes the same point as the lack of

Kaluli overextension: the child seems to be encoding aspects of a "prototypical event of object


It seems that the child does not begin with categories, such as 'actor,' or 'agent',
looking for the linguistic expressions of such notions in his or her native language.
What the child may begin with is much more limited and childlike ways of
conceiving basic events and situations, at first matching grammatical expression to
primary or basic event schemata [such as 'causal agent']...It is such prototypical
events, rather than case categories, that seen to provide the initial conceptual
framework for grammatical marking.

Thus there is linguistic evidence for the conceptual underpinnings of event structure and

its subsequent linguistic expression. What I would like to point out, leading to the following point

on the possible nature of these conceptual image-schemas, is that developmentally, both in

terms of the infant studies and early language acquisition studies, it is not necessarily the whole

of the event which is attended to, but one part of it. That is, the infants studies cited by Mandler

and the acquisition studies analyzed by Slobin show sensitivity to the starting and endpoints of

an event, as well as the path itself, not just to the event as a whole. Mandler even suggests that

her schema of caused motion (the "prototypical transitive event" Slobin defines) is slightly more

complex and therefore a slightly later conceptual creation. Developmentally, concepts of object

permanence precede concepts of motion which precede concepts of agency. Thus it makes

sense that humans have very early schemas for objects, then for objects in motion (both self-

and other-caused as Mandler suggests), then for objects which cause motion in other objects.

For the notion of a unitary transitive prototype, it is critical to understand that the image

schema proposed for the "prototypical transitive event", including a probably human agent with a

probably non-human patient causing a change of state or location, is a slightly later acquisition

COMBINING the earlier schemas of independent, animate motion/ change of state and non-

independent, non-animate motion/change of state. This conforms remarkably well to the three

"levels" of transitivity most commonly proposed: high transitive (two participants engaged in a

unilateral activity with a clearly causal agent and clearly affected patient); "mid"-transitive (one

participant engaged in a change of state or location without external cause); and "low" transitive

(one participant engaged in a durative state involving little to no change-these latter two are

often referred to as "intransitive"). The evidence above suggests the more appropriate

explanation would include three prototypes (or three typical instantiations of a single event

complex), rather than just one. This is precisely what Croft (1994), Kemmer (1994), and Bakker

(1994) variously propose.

Croft (1994:92) presents seven characteristics typifying a "simple event," two of which

are of particular interest here:

(f) simple event structure consists of the three-segment causal chain:
(g) simple event structure is endpoint-oriented: possible verbs consist of
the last segment (stative), the second and last segments inchoativee),
or the whole three segments (causative).

Basically, he is claiming that verbs have an underlying Idealized Cognitive Model (hereafter ICM,

following Lakoff, 1987) which represents a "self-contained event...isolated from the causal

network and individualized for various purposes...Subjects and objects represent the starting

point and end point respectively" (92). However, these ICMs do not create verb classes or event

classes, rather they correspond to the unmarked event view associated with the verb: causative,

inchoative, and stative. Thus in an interesting twist on the problem of transitivity, Croft argues

for an underlying event conceptualization (see figure below) which can have all or any part of it

"contained" for the purposes of representing an event. Further, the events denoted by verbs are

typically associated with one part (or all) of the idealized event; these are the event views. The

astericks and parentheses mark the point at which the "typical association" ends.

Causative. The rock broke the window

rock window (window) (window)
"-------> ----(*).-
cause become state

Inchoative* The window broke

window (window) (window)
""-------- (*)"----('*
become broken

Stative: The window is broken

window (window)

Figure 2.5: Croft's Event Views

Croft then argues that verbal markedness patterns found in various languages (English, French,

Japanese, Korean) represent the most "natural" or "typical" event view associated with the verb


This construal is based on human experience of the event, and is not surprising. In
general, unmarked statives are found with verbs denoting events that in human
experience most commonly are inherent properties, not implying any intemal
cause.... Unmarked causatives are found in verbs denoting events that frequently
or almost always occur in human experience with an external cause.... Unmarked
intransitives...are found in verbs denoting events that are commonly not permanent
(i.e. they can change) and most frequently occur-or are construed to occur-
without an external cause.

Basic and derived forms can be found with each of the basic event views:

Causative->lnchoative->Stative on Causative Event View
(1) Olaf opened the door.
(2) The door opened.
(3) The door is open.

Inchoative->Stative on Inchoative Event View
(4) Sven fishes.
(5) Sven's a good fisherman.

Stative->Caused Stative on Stative Event View
(6) pure sugar
(7) purified sugar

Croft then ties together various linguistic perspectives on events by showing that causative-

inchoative-stative, transitive-intransitive-adjective, and active-middle-passive are semantically

different manifestations of the basic tripartite ICM of events as figured above (102):

Verb Form Event Views
Derived causative inchoative stative(resultative)
Simple transitive intransitive adjective(stative)
Basic Voice active middle passive/resultative

Hence, in the end, Croft is arguing for a model with three "basic" forms, each of them

representing a portion of the ICM of events, but each of them also forming a natural "locus"

against which events construed as corresponding to that center are assessed thus producing

marked and unmarked forms for EACH of the event views.

Kemmer (1994) adds to the discussion the notion "relative elaboration of events" which

is not unlike Croft's event views, though she introduces the specific aspects of events which are

manipulated to produce different event perspectives. She explains the "relative elaboration of

events" as "the degree to which different schematic aspects of a situation are separated out and

viewed distinct by the speaker. The speaker in effect can choose to tum up' or turn down' the

resolution with which a particular event is viewed in order to highlight its internal structure to a

greater or lesser extent" (211).

Two important "schematic aspects" are the distinguishability of participants and

distinguishability of sub-components of the event itself. Kemmer also uses the by-now-familiar

event schema to represent what she simply terms a "2 participant Event Schema":


Figure 2.6: Kemmer's Event Schema

She uses the terms "Initiator" and "Endpoint" to capture both the schematic placement and

semantic characterization of the participants themselves. She is specifically claiming that the

distinguishability of participants is one semantic parameter along which the dine from "transitive"

to "intransitive" differ (209):


+ -

Kemmer observes that two-participant events encode maximal distinguishability of

participants, reflexives encode an event where essentially the Initiator and Endpoint are the

same participant (though linguistically marked with "two" participants through reflexive pronouns,

clitics, etc.), middles encode a similar situation but without the linguistic marking of two

participants, and one-participant events code a single participant who may be either Initiator or

Endpoint. Striking in her analysis is the notion that the semantic characteristics of the

participants and the number of participants are two separate issues; events may be construed as

having two participants with fully distinguished semantic characters, as having one participant

with either semantic character, or some blend within (the particular combinations limited, of

course, by individual languages). This separation allows a much neater and satisfying

explanation for constructions like the English get-passive which combines the semantic

characters Initiator/Endpoint with a construction traditionally associated with a topicalized

Endpoint, thereby preserving the pragmatic impact of the passive voice with the semantic import

of a reflexive and explaining the constraint against any mention of an extemal cause. It also

provides an explanation for why intransitive constructions are generally of two basic types:

intransitive activity (a single participant activity with an Initiator) versus an intransitive stative (a

single participant activity with an Endpoint).

Bakker (1994) is also interested in explaining the linguistic behavior of middles,

particularly in Ancient Greek. His anaylsis, though, offers some provocative insights into the

relationship between events and the semantic parameters informing them. He locates the crux

of transitivity in Ancient Greek in the subject/agent-related parameters given by Hopper and

Thompson (1980): Volitionality, Agency, andi adding Causation. Volitionality, Agency, and

Causation are separated basically by telicity, where Volitionality is simply free will of the subject,

Agency is goal-directed volitional action (to self or other participant), and Causation is other-

participant goal-directed volitional action producing an effect. Transitivity, according to Bakker,

is on a bi-direction dine in relation to these elements (25):



Thus, a causative event with two participants is high in transitivity while a volitional event with a

single participant is low in transitivity. When the parameter of affectedness is added, it can be

found in either end of the subject/agent spectrum: the subject/agent can be affected or can do

the affecting. In the former case, the middle arises (either as a "natural" situation type as in

growing up or as a construed property expressed by the middle form).

Again, what is noteworthy about Bakker with regards to this discussion is the separation

of transitivity" into natural types according to the semantic characteristics of the participants and

inherent semantics of events themselves. Thus a picture emerges of "prototypical" transitivity

which is not adequately captured by proposing a single two-participant event with a complex of

features, but calls for a range of at least three "prototypes" each crossed by the dine of

participant characterization from responsible Initiator/Effector/Causal Endpoint to non-

responsible Affected/Non-Causal Endpoint. Recalling Mandler's basic motion schemata and

Slobin's early language patterns provides additional support for the notion that event structure is

best represented by the three "stages" of a causative event: first, a single participant capable of

volitional activity (animate, biologically caused motion and early acquisition of ergative

morphology); second, a single participant undergoing some change (inanimate, non-biologically

caused motion and early acquisition of accusative morphology); third, two participants interacting

where one causes a change of state or location in the other (animate, biologically caused motion

acting to effect inanimate, non-biologically caused motion and the relatively early acquisition of

SVO word order patterns versus the relatively later acquisition of fully passive forms, "OV-ed by


These basic event structures form the conceptual basis which the linguistic system of

transitivity expresses. Specifically, transitivity expresses the event view (Croft) or event

perspective (Slobin, Berman and Slobin) of cognitive/conceptual event structures (this will be

discussed in greater detail below). There appear to be three basic event views expressed by the

transitivity system, as the discussion above proposes. There is, though, the puzzling absence of

a fourth, the purely stative. That is, one would guess that the first, and most basic perception of

an infant is "existence." Since children first learn the names of objects, there is even some

developmental support for this notion. Croft suggests as much with his "stative" event view

which is often grammaticized by adjectives. However, it seems to me that pure states such as

identification and ascription-while predications--are not really events at all. Cross-linguistically,

there is better evidence via the number of languages which create states through adjectival

constructions that pure states should be considered a conceptual model of its own, separate from

the event structures relating to transitivity. I will assume as much for this project. Berman and Slobin revisited--event construal

As discussed briefly in Chapter One, Berman and Slobin identify four dimensions of

event construal (selection of topic, loci of control, event view, and degree of agency).

Specifically, they make the claim that these dimensions are part of the preverbal message:

"Before one can say anything, one must decide what one wants to say. As Levelt has put it:

'The construction of a preverbal message is a first step in the generation of speech' (1989, p.

107). We propose that, in constructing a preverbal message, the speaker must make decisions

on at least four dimensions of event construal..." (517). Once the dimensions are decided

upon, various grammatical options are entailed, of which transitivity is one system. Transitivity,

then, is part of the grammatical system but not the underlying cognitive system; that is, I accept

for this project that Berman and Slobin's preverbal dimensions are (part of) the general cognitive

orientations available, probably throughout the cognitive system, but linguistically, transitivity is

one grammatical system through which these dimensions are realized. This represents a

departure from Rice (1987) who seems to claim that transitivity is the "underlying system" which

is realized verbally: "...transitivity [can be] viewed as a continuous phenomenon and as a

phenomenon (or as a family of phenomena ranging across different cognitive domains) that can

be explained in an essentially non-linguistic way" (7). My claim is that Event construal and its

dimensions are the underlying, preverbal system, and transitivity is one of the linguistic systems

instantiating them. Other options, as discussed by Berman and Slobin, include, for instance, the

lexical system which provides means for expressing degree of agency, as with adverbs such as

"unintentionally" and "deliberately" which modify the interpretation of the event through non-

grammatical means.

Specifically, it is my claim that transitivity realizes the selection of Event view. Event

view refers to the "cognitive perspectives on events, rather than to verbs" (Berman and Slobin,

519). Berman and Slobin identify three choices for event view: Cause, Become, and State. In

Cause view, the event is represented with an actor who causes a change of state in the

undergoer (Berman and Slobin make use of Van Valin's macrorole terms, actor and undergoer,

as mentioned in Chapter One). In the Become view, a change of state takes place without

mention of external force. For the State view, reference is made only to the continuing state of

the participant. While Berman and Slobin draw the obvious connections between Cause-

Become-State to the verbal systems proposed by Vendler, Dowty, and Van Valin, I would

hazard these three Events views are realized at the level of the clause as the three prototypes of

transitivity rather than lexically as verbal categories. In summary, if the overall function of

grammar is to mediate internal, mental representation with communicative needs and intents,

then transitivity specifically mediates the internal (preverbal) representation of Event view and

the communicative intent of perspective (discussed in detail in Chapter Three). Transitivity as a

system is limited to the expression of Event view only-- the whole of event construal and its

communication employs a vaster amount of linguistic (and probably non-linguistic) material. Degrees of transitivity

Given the argument above, the question of whether there are "degrees" of transitivity

becomes an even more interesting issue. Traditionally, there have been two alternatives

proposed to this problem. The first is a continuum from "High" to "Low" (Hopper and Thompson,

1980; Rice, 1984); the second offers specific levels of high, mid, and low (Halliday, 1967). While

to some extent both claim the same thing, Halliday's specific levels seem to find greater validity

grammatically. That is, grammar in general does not tend to operate along smooth dines

between endpoints, but to manifest particular forms according to particular points between two

endpoints. The lexicon compensates for grammar's relative intransigence by offering the

possibility for the kind of subtle and continuous changes implied by the notion "continuum."

This interpretation fits well with the notion of three basic event schematas which are

realized through various transitivity options. It is something of a misnomer to claim "high," "mid,"

and "low" levels of transitivity when each of the clause structures typically associated with these

labels is actually a prototype in and of itself. Rather, there are degrees of transitivity within each

of the basic types, probably with their own, separate semantics. For example, "really high"

transitivity may be distinguished from merely high transitivity through the agency/patient

implicatures discussed below. "Very low" transitivity can be distinguished from simple "low"

transitivity by the distinction of durative action versus state. It is the "middle" transitives that are

the most interesting, encompassing as they do varied combinations of participant semantics and

clausal expression. In the end, rather than conceptualizing three degrees of transitivity, we

instead should conceptualize three event structures realized via transitivity, each with its own

possible dines and opportunities for grammatical expression.

This argument is essentially what Croft, Kemmer, and Bakker are claiming: there are a

variety of basic event types possible--Mandler provides independent evidence in her

speculations on the physical origins of event structure: "self-caused" animate, biologically

inherent motion and "other-caused" non-animate, non-biologically inherent motion. I would add

to that a third category, also readily available to the child: non-caused motion which are changes

of state and/or location without external cause but also without integral "control" such as

sleeping, breathing, living, etc. These three also correspond well to ALL of the sides of an event:

before, during, and after which are all grammaticized: -->A/ A---->B/ B--->.

For example, what makes for a sentence which passivizes well need not be something

prototypical, but a particular collocation of semantic factors in the event structure. In particular,

one in which minimally, affectedness can be deduced (either by change of location or change of

state) and (ideally) where the trajectory of the event and the endpoint of the event are one and

the same. Interestingly enough, in English, following Slobin (1994), Talmy (1991), and Hopper

(1991), this is NOT necessarily the norm. In English, very often predication is spread out over

several members of the clause, most often over a combination of the verb (incorporating

manner) and a prepositional element (indicating path). Statistically, this is a quite common

phenomenon, so much so that it seems rather absurd to claim that the active, declarative, with

human subject and non-human affected patient is THE prototypical eventive sentence in English.

Rather, it is the sentence on the far end of the affectedness and trajectory scale-highest in

eventive dynamism, perhaps. Thus, what is most typical is a combination of endpoints and

trajectory: either an endpoint which is capable of self-trajectory (mid-transitive, "become" or

"act," single-argument clause); an endpoint not capable of self-trajectory (a low transitive, "state",

single-argument clause); and two endpoints sharing a trajectory (a high transitive, "cause", two-

argument clause OR a mid-high transitive with two endpoints sharing a trajectory spead out over

a clause + phrase combination). Within each "prototype," there are a range of better members.

2.3.3. Event Structure and Transitivity

It is the contention of this project that Event Structure is the cognitive system and

transitivity the linguistic system. In fact, given Mandler's hypotheses about the possible origin's

for the expression of events, it seems likely that transitivity is to some extent "inferred" by the

child as a later, or second, step in this acquisition process. According to Mandler, the child first

distinguishes biological movement from non-biological movement, and "recognizes" or "infers"

the second is caused motion (see also Bruner, 1989). Another way of viewing the matter is that

the infant begins to recognize which endpoint of motion or change is responsible for the activity

and which is not. This is the possible beginning of perspective. The notion that affectedness is

also interpretable as "the party not responsible for the change" is linguistically verifiable by

grammatical behavior of components which are construed as both affected and responsible in

some manner, ie. the get-passive in English, reflexive constructions, and middle constructions.

Thus, in the structure of events, several components are basic which find themselves coded by a

variety of linguistic expressions (none of which should rightfully be called more basic or

prototypical than the other):

(1)-- The Trajectory Component

This component is part of two based in the notion of caused, other-caused, or non-

caused motion begun in the perception of biologically and non-biologically rooted motion

(Mandler, 1991). It is where the notion "unilateral activity" in transitivity originates.

(2)-- The EndPoint/s Component

This is the second part of the two bases in the various notions of caused motion.

Basically, the motion is seen in its relationship to where it begins and where it ends (hereafter

distinguished as Starting point and Ending point, respectively). This is ultimately where the

conceptual notions of causation, intention, and responsibility originate since these are the

various semantic/pragmatic perspectives ascribable to the endpoints involved. Thus,

linguistically, this is also where such notions as "agent" and "patient" evolve from.

(1) + (2) = (3)-- The EndPoint and Trajectory Complex

These two together act as the basis of Event Structure (Croft, Langacker, Lakoff,

Kemmer, Rice, Mandler, Talmy). However, contrary to the claims of Rice and Giv6n, the

particular combination of two endpoints plus a unilateral trajectory is not more basic than any of

the other combinations. Thus, I am following Croft and Kemmer who propose that there are a

range of possible prototypes which in turn are coded by a variety of linguistic expressions each

with a probable most "typical" or "canonical" form. Berman and Slobin (1994), Slobin (1994),

Talmy (1991), and Hopper (1991) among others imply as much with their particular

interpretations. Slobin (inter alia) recognizes three basic event types to Event Construal: Cause,

Become, and State. Talmy shows that typologically, different languages convey endpoint +

trajectory complexes in typical ways with different linguistic parsings of the whole. Hopper shows

that English often chooses to array two enpoints + trajectory complexes over several elements in

the sentence, rather than in just the traditional "transitive" construction. Luchjenbroers (1996)

shows how the complex constructions Hopper investigates have real and verifiable discourse

effects in terms of information packaging. Bruner (1986) summarizes the various infant studies

suggesting that we are sensitive to the roles of causation and intention at very young ages.

Giv6n and Lang (1994), without really intending to, give additional support in their diachronic

treatment of the get-passive in English by demonstrating how this unique construction developed

as a means of expressing both affectedness and responsibility simultaneously. Hargreaves

(1996) also demonstrates inflectional categories which are directly sensitive to the knowability of

intention in the morphology of control verbs in Newari.

The event complex (here really a simpler version compared to others), then, is a basic

cognitive, conceptual structure which gets coded through a variety of linguistic devices, including

participant semantics and transitivity. Transitivity, though, is not the cognitively basic structure;

it is one linguistic/grammatical system expressing a certain part of the event complex, that part

being the particular array of endpoint and trajectory relationships. At the level of syntax, this

results in clause structures as stamped out by circumscribing endpoint + trajectory relationships

marked by language-specific renderings of the "responsibility" of the endpoints (as in accusative

versus absolutive languages). At the level of semantics, this results in the various sentence

types and alternations allowed and disallowed as expressions of energy flow and participant

relationships (influenced, I would contend, by pragmatics in the guise of cultural and "real world"

knowledge). At the level of discourse, this means participants and "event" relationships PLUS

perspective-- a discourse-semantic interaction, construed variously in languages according to

what the transitivity system allows together with language-specific constraints on coherence.

Since transitivityy" as a term is fast-losing its significance due to overextension, we

confine it to the grammatical level of expressing EndPoint-Trajectory relationships (hereafter P/T

relationships). Each language encodes these relationships differently, and in a manner I believe

is well-described by Talmy's "verb-framed" and "satellite-framed" typological distinction (1985).

Slobin (1996) confirms as much in his comparison of texts in English and Spanish, showing that

while event structure conceptually encompasses the whole of P/T relationships, there are regular

patterns instantiated in various languages. In English, very often P/T relationships are not coded

within a single clause but in clause-phrase combinations (see Hopper and Luchjenbroers cited

above), as in Cora swung his blade and cut down the opposing swordswoman. Spanish,

however, tends to code both the "manner" of the activity and its trajectory in a single verb. This

difference results in both different lexicalizations of verbs and also different pattern of discourse.

For English, transitivity actually has FOUR forms in which it is expressed, each

instantiating particular P/T relationships and clause structure. These are the four clause-types of

transitivity in English (please see Figure at chapter's end for a schematic representation and

textual examples).

Intransitive(Durative transitive)- Endpoint but little Trajectory, bound by the same clause
Atelic Transitive- single Endpoint plus Trajectory in either direction, bound by the same clause
Telic Transitive- two Endpoints plus Trajectory, bound by the same clause
Dispersed Transitive- two Endpoints plus Trajectory specified across a clause and a phrase

These forms are descriptive in nature, although they have theoretical and typological bases.

They will also be used as working terms throughout the project. Ultimately, I believe the above

terms are more accurate than the concept of degrees since, as shown above, there are not

degrees of transitivity from a single prototype, but (at least) three different "prototypes" each with

its own "degrees." Languages are more properly investigated for how they express

Point/Trajectory complexes than for their "degrees" of transitivity. At the level of text, participant

semantics, lexical semantics, and pragmatics also come into play to create the entire implicature

normally labeled transitivityy." Throughout this project, I will use the terms Starting point, Ending

point, Trajectory, and P/T complex when referring to the purely schematic aspects of transitivity.

2.3.4 "Agent," "Patient" and Other Labels

Having dealt with the extension of prototype to transitivity and discarded the notion of a

single prototype in favor of at least three prototypes expressed via four clause types, we now

move to a different level of meaning--participant semantics. Recall that Giv6n, Hopper and

Thompson, and Rice among others accept the roles "agent" and "patient" as core relationships

within transitivity. An alternative approach exemplified by Croft and Kemmer is to separate the

semantic labels attached to participants from their schematic roles. Taken together, there is

convincing evidence that event participants code at least two different sets of parameters (at the

level of the clause): first, the purely schematic, abstract position in event structure, either as

starting point or ending point; second, a semantic character expressing the relationship of

participant to event, generally as the initiating, responsible party or the affected party undergoing

a change of state. The most common labels for the semantic roles has been "agent" and

"patient," respectively.

Rice (1987) argues that in order for there to be a transfer of effect from one participant to

another, one participant must possess the means to affect the second in an activity which is

forceful and flows in only one direction. The means possessed by the first participant may be

inherent to the activity, the first participant, or the relationship between participants and may be

understood from real world knowledge as well as linguistic knowledge. Hopper and Thompson

make similar claims, stating there must be two participants for High transitivity, and that one

participant is an Agent high in potency and volition. The other participant, which is highly

individuated, contrasts by being totally affected by the kinetic action.

Van Valin and Wilkins (1996) argue against "agent" as a prototype. They base their

conclusion on the work of Holisky (1987), who demonstrated that agency is actually an

implicature at the "...intersection of the semantics of the clause (the semantics of both the actor

NP and the predicate) and general principle of conversation (cf. Grice, 1975)" (309). Holisky

designates a pragmatic principle in the process, designated by her as Pragmatic Principle [8]:

You may interpret effectors and effector-themes which are human as agents (in the absence of

evidence to the contrary). Where Holisky based her assertions on evidence in Tsova-Tush, Van

Valin and Wilkins offer straightforward examples from English.

(8) Larry killed the deer.
(9) Larry intentionally killed the deer.
(10) Larry accidently killed the deer.
(11) The explosion killed the deer.

In (8), there is no explicit mark for whether the actor is an agent or not (though by Holisky's

principle [8] above, we interpret it to be so), while (9) confirms this as a potential interpretation.

On the other hand, (10) provides "information to the contrary," disconfirming the agentive nature

of "Larry" and leaving him merely as a causal participant. Sentence (11) shows that the actor

need not be animate at all, and therefore no appeal is made to Holisky's pragmatic principle.

It is not the case that NO verbs have an inherent agent. Obviously, murder does. But

inherent agency is relatively rare. Van Valin and Wilkins argue:

This has important implications for the notion 'agent prototype' and the idea that
relations like force and instrument are just 'less good' members of the category of
which agent is the central member. The features in question do not in fact define
a prototype, but rather constitute the relevant information on the basis of which
the agent implicature is made Moreover, force and instrument are not variants on
agent at all but rather are variants of the effector role which underlies agent as
well. All three roles are doers which differ in certain crucial ways along the
dimensions of the lexical properties of the NP and their position in the causal
sequence defined by the verb. (320) [emphasis added]

While the entirety of their proof is too extensive to go into here, basically, the NP

properties they are referring to are "animate/non-animate" and "instigator/non-instigator."

Animacy distinguishes effectors capable of feeling, self-expression, and self-movement from

those incapable of the same. Instigation refers to whether the effector is "self-energetic", that is,

has an inherent energy source, as with natural forces. In combination, the three primary starting

point/effector roles emerge: the agent who is animate, instigator and preferably human (though

animacy can also be ascribed to culture-specific entities); the force which is non-animate but

instigator; the instrument which is non-animate, non-instigator, but still the effector.

There are two striking consequences of the above analysis. First, a more constrained

and thoughtful approach is taken to the ubiquitous notion of prototype. Second, a viable

alternative linguistic process--implicature--is offered to balance the picture. As we shall see in

the subsequent chapters, implicature is an important factor to processing at the level of


Just as "agent" is not necessarily a prototypical participant, "patient" is equally dubious.

As Bakker demonstrates, the "prototypical" patient actually reduces transitivity in Ancient Greek

by lowering the volitionality, hence the overall measure of transitivity for the event (1994, 43).

This should not be surprising since something similar happens in English when the whole of

event structure is spread out over the verb phrase plus a prepositional phrase: Cora swung his

blade and cut down the opposing swordswoman. While the first predication, Cora swung his

blade is "transitive" in the sense of having a syntactic direct object, the entire event in question

has two parts: the initiating event (swinging the blade) and the ending result (cutting down the

opponent). Recall that Kemmer's elaboration of events allows various levels of focus on the

event so that a whole structure such as might be expressed in a basic transitive clause, i.e. Cora

killed the opposing swordswoman, can be expressed via attention to any of its parts.

Nevertheless, "the blade" in question, while a prototypical patient (it is affected by change of

location, non-human) is really but an instrument in the overall event; it is not the instigating

force, and is not the locus of effect.

Just as "agent" is an implicature from a more basic role, Effector, "patient" is likewise an

implicature based on lexical, semantic, and pragmatic factors. The appeal of agency and

patienthood as necessary qualities derives from their antonymical quality; they are not merely

distinct, they are maximally distinct. Yet they are not essential to high transitivity. Instead, they

mark the far endpoints, the points at which transitivity as a system is easiest to see. Indeed, the

semantic characterization at work here seems to be "asymmetry." In "really high" transitivity,

there are indeed both highly agentitive subjects and highly affected participants, e.g. John kicked

the sleeping poodle/Brutus smashed the antique vase. On the other hand, for merely "high"

transitivity, (as defined by the possibility of passive and/or middle alterations), only one of the

participants need be capable of effecting or of being affected: Olafjumped the fence; Bertha

danced Swan Lake beautifully; The moon orbits the earth; Sven broke the glass; Gertrude drove

the car; The wind blew the leaves gently.

Participant asymmetry, "skewing" as it is termed here, has actual structural effects. That

which is not skewed towards the affected participant (or which is not neutral with regards to either

end) has limitations. Thus, the difference between Brutus shattered the glass and Brutus

smashed the glass is that the first is skewed towards the Affected and the second towards the

Effector. As a result, the middle construction is available to the first, but not to the second: The

glass shattered/* The glass smashed. There are discourse consequences as well. Whichever

participant end is skewed towards is also more likely to continue as topic in subsequent

discourse. It should be remembered, though, that P/T complexes alone do not have this quality.

The P/T complex is schematic, conceptual, but lacks the semantic richness of participant


For this study, the term "patient" will be avoided, and instead the term "Affected" will be

used for the participant which is the locus of effect. In the same vein, "Effector" will be the term

used for the semantic character of the participant which initiates or causes the activity in an

event. For participants, then, "Starting point" and "Ending point" will be used when the reference

is to event structure, while "Effector" and "Affected" will be used when the reference is to

participant semantics.

2.3.5 Verbal Modality--Teasing Apart Semantics and Pragmatics

There are still a few issues which need to be settled. Recall that Giv6n claims that the

prototypical verb in a transitive clause is "fast-paced, completed, real, and perceptually-

cognitively salient." We shall take each of these one by one and see what a closer examination

yields for transitivity. Fast-paced

The notion "fast-paced" is characterized as transfer of effect from source to goal which is

so fast as to be instantaneous. Yet even for telic transitive clauses, there seems to be little

support for this: Sven kicked the dog/The dog was kicked by Sven; Sven cooked the dinner/The

dinner was cooked by Sven;/Sven studied the questions/The questions were studied by Sven;

Sven studied chemistry/?Chemistry was studied by Sven. In the examples above, there is a

clear decrease in the instantaneous nature of the event, yet all but one passivize; only in the

case where the goal is not quantified does passivization suffer. The instantaneous quality of the

action seems to have less to do with transitivity than is supposed. All the sentences above are

telic transitive clauses despite the range from "fast-changing" to "slow changing" to durative

event. The lack of importance of instantaneous effect is not surprising given the components of

transitivity proposed thus far; it is not how quick the verb is but the particular set of asymmetries

instantiated in the clause. Completed

A completed event is linguistically defined as one which is bounded. Boundedness,

according to Giv6n, is a verbal quality meaning "non-lingering" (p. 46, vol. II). "Non-lingereing"

as a terms seems to imply time; in fact, it is more accurately understood as "completion of the

event." Yet another way of looking at it may be that when the activity is over, all of the goal is

affected, not just part of it. Verkuyl (1990) argues persuasively for the contribution of the object

to the "inner aspect" of the predicate. When the object is of a "specified quantity," the phrase

yields a bounded event because the quantity of the object is exhausted by the activity. However,

it is not necessarily the case that the goal is not affected in non-bounded situations, but that the

effect is not complete.

Completedness is easiest to reckon within telic transitives. In these clauses, the end of

the activity is signaled by overt mention--the affected object itself. In essense, the event finishes

itself by completing the change of state or location of the Affected participant.

(12) Olaf ate the fish.
(13) Olaf cooked the fish.
(14) Olaf put the fish in the refrigerator.

Thus it is easy to see how telic transitives include a strong element of completion. Nevertheless,

atelic transitives also have a sense of completion except that it is generally inherent to the verb


(15) Svenjumped.
(16) Gertrude sneezed.
(17) Brutus grinned wickedly.
(18) Brunhilda chortled in joy.

Each of these is a "completed" activity--the activity does not linger (for example, although we

don't normally conceptualize a grin as punctual, to make it non-punctual additional information

must be added to the sentence: Brutus kept grinning wickedly). Thus it is clear that the activity

is not bound by a change-of-state of a second participant but by the inherent semantics of the

verb itself as it refers to the one engaged in the activity.

Obviously, it is the intransitive clause which has the least sense of completion.

However, as these sentences show, it is not completion which is the semantic factor, but the

durativity of the events themselves.

(19) Bertha grew up in Minnesota.
(20) Sven knew all the answers.
(21) Elvira lived in New York for ten years.

One other factor influences our sense of verbal completedness: tense. Any activity in

the past tense carries with it a sense of "doneness." Thus, if (12) were re-written Olaf eats the

fish/fish, our sense of boundedness is completely changed. The question is--has the transitivity

of the event changed as well? According to the explanation above, transitivity has not changed.

There is still a P/T complex with an Starting Point, an Ending Point, and a Trajectory joining

them. There is still and Effector and Affected. What has changed is not the transitivity of the

clause, but the speaker-hearer relationship to the event. The pragmatics are different, not the

semantics. Completedness is more properly understood as a pragmatic implication related to

tense than as a semantic element of transitivity. Perceptuallv-cognitively salient

Many authors have argued that dynamism--the degree of "activity inherent to the verb"-

is an important aspect of transitivity. Intuitively this seems to be the case, and this is no doubt

where the attribute "fast-changing" originated. Yet there is no quantifiable or explanatory

definition by which dynamism can be assessed. I would like to offer one in terms of transitivity

as hypothesized thus far. This discussion springs from the observation of many linguists, but

primarily those from functional and cognitive schools, and may serve as an explanatory device

for dynamism.

At first glance, it appears simple enough to appeal to the Affectedness relationship in

measuring dynamism. Intuitively, break and shatter are in a degree relationship in which the

goal of shatter is more affected than the goal of break. This observation holds true for other

verbs as well: drink-gulp-quaff, nibble-eat-gorge; throw-hurl, etc. On the other hand, there are

lexical sets which change the effectiveness of the source instead: kill-murder; clean-scrub-scour,

hit-punch. There are still other alternations, skewed so far in the direction of source or goal that

different lexical forms are used: kill-die; sit-set; teach-learn; buy-sell; etc.

Notably, the difference in these pairs of terms is not just the number of arguments the

verb can take but the semantic roles of the arguments themselves. That is, for many actions,

there are both lexical and grammatical options for selecting which perspective on a situation is

opted for in discourse. In the sentences below, (25) and (26) demonstrate that English allows

both the pragmatically understood force and instrument to be the "source" of the action.

Example (27) shows that both of the above can passivize. In (28)-(35) various grammatical

forms topicalize the affected participant whereas (28), (29), (30) and (35) show that lexical

choice also contributes to the clausal combinations permitted by the grammar of English.

(22) John broke the window.
(23) The rock broke the window.
(24) The window was broken by John/the rock.
(25) The window broke/*by John.
(26) The window broke easily/with difficulty/*difficultly.
(27) *The windows break easily/Windows break easily.
(28) The windows break easily, but the wooden frames do not.
(29) Windows broke easily; wood was but little harder.
(30) (John threw the rock) The window broke/shattered.
(31) John broke/destroyed/crushed the window with a rock.
(32) *The window destroyed/crushed.

Both the Effector and Affected roles get lexicalized in English, and certain constructions such as

the middle are also Affected-oriented.

Weirzbicka (1988) offers an intriguing analysis of causation and lexicalization. English,

she proposes, is filled with many level of causative, instantiated in lexical causatives as well as

periphrastic causal constructions. The list of lexical causatives (252-3) which "pays attention to

different strategies of human causation" rather neatly dichotomizes between Effector-oriented

and Affected-oriented lexical items: order, command, demand, require, suggest, advocate,

instruct, persuade, decree, ordain, authorize, commission; beg, implore, beseech, entreat, plead,

intercede, apply, appeal, enjoin. There were a few items difficult to classify either way, which

seem to form the "basic" category against which the others are organized: ask, request, tell.

This list further substantiates the idea that given an event, the whole of the event is conceptually

available but certain highly relevant, highly predictable relationships are the ones that actually

become lexicalized items.

Is there anything in our discussion so far which might explain or provide an

organizational umbrella for the behavior described above? I believe the cognitive relationship of

asymmetry, as discussed by Rice but made linguistically relevant by Slobin, may be the answer.

Slobin (1973) makes strong arguments for the cognitive impact of asymmetry as a basic

organizing principle in human beings. Infants very quickly show a preference for asymmetrical

relationships such as handedness. Other infant studies show surprise responses can be

provoked when certain basic asymmetrical relationships are violated. Further, one of the

hallmarks of linguistic sophistication in terms of narrative is the degree to which speakers can

differently construe cause-effect relationships among events (Berman and Slobin, 1994). Less

mature narrators tend to string events together using a simple conjunctive "and." As speakers

mature, they form hierarchical relationships in the narrative, beginning with simpler causal chains

(and then he did x because...), gradually incorporating the sophisticated tense/aspect and lexical

"complexes" we associate with adult narration.

Loosely speaking, this language development is understood by these authors as a

manifestation of our cognitive predisposition towards asymmetry. Lexically, asymmetry

manifests itself in the sort of relationships discussed by Bybee, Langacker, and Weirzbecka.

Lexical choice has an effect on transitivity insofar as any verb is "skewed" towards one end of

the Effector-Affected continuum. Those verbs which include both the source and goal

semantically, generally speaking, participate in telic transitive clauses. Those skewed strongly to

either Effector or Affected result in atelic transitives. Thus "perceptual-cognitive salience" of a

verb may be partially defined as the degree to which the lexical item is skewed towards the

affected participant in an action. Halliday (1985) suggests as much himself when he claims the

transitivity alternations in English are less a matter of inherent meaning than whether the event

was viewed from the perspective of the source or the goal.

There is not to my knowledge any kind of list composed according to these parameters

(skewed towards source, goal, or neither; but see Levin, 1993), and it is outside the scope of this

study to do so. Nevertheless, I will appeal to this notion with the hopes that someone may find it

useful for further work.

2.4 Transitivity and Time

The following excerpts will provide examples for the semantic character of transitivity as it is

delimited by discourse. Actually, most of this discussion is directly germaine to Chapter Three

and will be explored in greater detail at that point.

Uttering a curse in his well-practiced falsetto, Cora swung his blade and cut down
the opposing swordswoman. His contoured breastplate emphasized features
which were not truly present.
From "Lady of Steel", R. Zelazny

There was still a company of archers that held their ground among the burning
houses. Their captain was Bard, whose friends had accused him of prophesying
flood and poisoned fish, though they know his worth and courage. He was a
descendant in the line of Girion, Lord of Dale, whose wife and child had escaped
down the Running River from the ruin long ago. Now he used a great yew bow,
loosing arrows at the dragon until all of them but one were spent. The flames were
near him. His companions were leaving him. He bent his bow for the last time.

From The Hobbit, J.R.R. Tolkien

Each of the above excerpts is the initial paragraph of the episodes in question. The first is the

first paragraph of a (very) short story. The second is the beginning to a particular episode in The

Hobbit, marked by a change of scene and the eventual closure to this string of events. Of

particular interest here are the last two sentences of each (which in the Zelazny text includes the

whole of the paragraph).

The two sentences, His contoured breastplate emphasized features which were not truly

present and His companions were leaving him, both have many of the elements required of

"high" transitivity. There are two participants, a grammatical subject and object, and a path of

movement (the first perceptual, the second physical). Both sentences passivize, and textually,

could passivize with little or no distortion to the story-line (again, this is a discourse constraint on

the perspective-providing function of transitivity and will be explained in greater depth in the

following chapter).

Uttering a curse in his well-practiced falsetto, Cora swung his blade and cut down
the opposing swordswoman. Features were emphasized on his contoured
breastplate which were not truly present.
From "Lady of Steel", R. Zelazny

There was still a company of archers that held their ground among the buying
houses. Their captain was Bard, whose friends had accused him of prophesying
flood and poisoned fish, though they know his worth and courage. He was a
descendant in the line of Girion, Lord of Dale, whose wife and child had escaped
down the Running River from the ruin long ago. Now he used a great yew bow,
loosing arrows at the dragon until all of them but one were spent. The flames were
near him. He was being left by his companions. He bent his bow for the last

From The Hobbit, J.R.R. Tolkien

Nevertheless, it is intuitively obvious that these sentences, while subscribing to the

structural dictates most frequently associated with transitivity, don't "feel" particularly transitive.

Given the propositions above--namely, that there is more than one Event Structure prototype

and that the semantic character of Starting points and Ending points in Event Structure are

derived by implicature rather prototype--we can see these sentences conform to a structural

prototype while differing in participant and verbal semantics, much as Kemmer suggests for the

distinction between reflexives and middles.

For the sentence, His contoured breastplate emphasized features which were not truly

present, two dimensions of participant semantics are played with: the Effector/Affected

dimension and predicate causality. First, his contoured breastplate is in the subject slot, hence

the Effector role, but according to Van Valin and Wilkin's features, has neither animacy or

instigator qualities. The NP is not capable of self-motion nor is it self-energetic. Thus, it is more

of an instrument than either an agent or force. But it is an instrument of what? We will return to

this question shortly.

The object, features, is also not an easily identified endpoint because it is poorly

distinguished from the initiator in the event. In fact, ultimately, the features and the breastplate

are one and the same item, separated from one another by an act of description. Finally, the

predication itself is hardly "eventive" in the sense that it points back to itself; the verb

"emphasize" means to make more perceptually salient some part of an existing state.

Despite the above, these not-very-distinct participants and this not-very-dynamic event

are cast together in a clause structure prototypically associated with just the opposite. This

results in an "act of description" or an "event of description." In narrative theory (Fleischmann,

1990; Labov, 1972), descriptive events are labeled "evaluative" in terms of story function and

often provide the ground against which a "complicating action" (an event which moves the time-

line of the narrative forward) is figured. Descriptive events are also used to draw the reader's

attention to some salient fact which provides essential plot-line information. This is precisely the

function of this sentence. His contoured breastplate is the instrument used to ground the

information that the state being predicated is somehow false. It presages an account for the

gender mis-match between the name, Cora, and the male pronoun references. For this story (in

the appendix), this information is the key knowledge against which the plot unfolds.

The linguistic signalling of this information comes in the particular combination of

features. I would like to reiterate that this effect is not accomplished by prototype violation-there

is no violation of the structural prototype as the passivization attests. Rather, there is a

systematic combination of features along various parameters "signalling" to the reader that an

inference must be made in order to complete the meaning (for an excellent discussion of the

difference between inferencing as a process and linguistic cognition as a process, please see

Giv6n, 1995, chapter 8). For discourse, implicatures are created which are then 'tested" against

subsequent information. The various parameters involved include, at least:

Event Structure-prototypical dimension of clause meaning based on
early-acquired conceptual schemas of motion
Participant Semantics- derived by implicature according to lexical
semantics, pragmatics from underspecified semantic roles
Discourse Coherence--implicature-driven processing involving clausal
semantics, pragmatics, topic chains, etc.

As we shall see in Chapters Three and Four, the latter two play a crucial role in licensing the

appropriate use of alternative realizations of transitivity, such as the passive and middle.

In the second example sentence, His companions were leaving him, a different set of

features are played with, almost exclusively confined to the level of discourse. A momentary

diversion needs to be made at this point to the linguistic structure of narrative. The features

mentioned above have to do with the nature of narrative as a past-time, event-driven construct

(Longacre, 1983; Fleischmann, 1990; Schiffrin, 1981; Hatav, 1996). Specifically, they have to

do with the contention that preterite as a tense has no inherent sequential meaning. This has

been a somewhat controversial point since the perfective nature of the simple past is "...ideally

suited to reporting experience that has been cognitively packaged into synthetic units amenable

to representation as points along a time line" (Fleischmann, 24). Nevertheless, as Comrie has

argued, a weak point of linguistic analysis has been the failure to differentiate context-

independent, inherent meaning from context-dependent, derived meaning (Fleischmann citing

Comrie, 1985). Fleischmann summarizes the argument thusly (24-25):

Several investigators...have therefore argued that the basic function of this PAST is
to mark the foreground of discourse, while the background is marked by its IPFV
counterparts...[yet] neither the figural quality of the PRET nor its sequential quality
is an intrinsic part of its meaning; rather, these properties emerge from the
interaction of its basic meaning (PAST time + PFV aspect) with a specific
discourse context--narrative. In other words, the foregrounding ability of the PRET
and the backgrounding ability of the IMP are contextual implicatures derivable
from the synthetic and analytic visions of their respective aspects. [emphasis

Early investigations into the function of transitivity (most notably, the well-cited Hopper

and Thompson, 1980, and Hopper and Thompson, ed., 1982) made precisely the claim that the

purpose of "high" transitive constructions in discourse was to foreground information; most

particularly, to place a clause and the events it was depicting on the time-line. Yet, our example

sentence easily and clearly passivizes; another feature of "high" transitivity. Given the

discussion above, it is simple enough to distinguish the inherent meaning of transitivity (which is

to provide the three event perspectives introduced above) and its discourse function (which is, I

will contend in the following chapter, to provide the grammatical base for the manipulation of

perspective according to the constraints of discourse coherence) from the textual function of

tense-aspect configurations. In other words, transitivity in and of itself has little or nothing to do

with grounding or figuring of time-line information. This is the primary task of tense-aspect in

discourse. Thus our example sentence is "high" in transitivity insofar as it depicts an event with

distinct Starting and Ending points, a unilateral trajectory, and a change in the circumstances of

the affected participant. However, it is not on the time-line because it is cast in the imperfective,

providing the ground for the subsequent action.

On the other end of the spectrum are so-called intransitive events which are indeed on

the time-line. These are events featuring either of the single-participant event structures: a

participant causing a change of state with no clause-bound mention of the endpoint or a

participant undergoing a change of state with no clause-bound mention of an extemal cause.

One brief excerpt from "Lady of Steel" will suffice to demonstrate.

Simultaneous then, attacks came from the right and the left. Beginning his battle-
song, he parried to the left, cut to the right, parried left again, cut through that warrior,
parried right, and thrust. Both attackers fell.

In this paragraph (the second full paragraph of the story), it is obvious that each of the chained

clauses represents a new point on the time-line. The implicature is made by the use of the

simple past in perfective form, the use of prepositional phrases specifying change of direction

("dynamic location"), and our knowledge that sword fights are fought by independent series of

strokes through time (versus a bomb, which kills en masse with a single discharge). Additionally,

the zero marking of the subject/topic clinches the reader's knowledge that the same participant is

being talked about. The use of the comma to separate events instead of a period is a signal

given by the writer that these events, while sequential, nevertheless happen in very short time

intervals. It is a kind of "close focus" camera effect, a metaphor used by a number of linguists

(Kemmer, Hopper, Talmy, etc.) and one which will be developed rather extensively in Chapter


Finally, not only do we find single-participant clauses on the time-line, but single

participant clauses from both perspectives on the event: Starting and Ending points. First, we

have as Initiator, Cora, who is the subject/topic of the entire chained sequence of events.

Second, we have as Ending points the attackers who are finally defeated in the final clause. It is

important to note here that pragmatically, that is, by inference, the reader understands perfectly

well that what is happening here is a highly transitive series of events: Cora is attacked--> Cora

fights back--> the attackers are killed. What the writer has done is gone inside the events, so to

speak, and presented them from a close-focus perspective, the view of the participants.

Grammatically, this is expressed via individual, single participant clauses. Thus even while

there is a kind of "macro-Event Structure" implied, the expression of that structure is not

constrained to just one view. It is, however, constrained to the view of the participants.

Nevertheless, there is a cognitive basis for the confusion of time-lines, plot-lines, and

transitivity. The two-participant transitive sentence (in English) captures two aspects of events,

force dynamics and passage of time, and construes them in a parallel fashion to how they are

understood. That is, in a sentence such as Bill kicked the sleeping poodle both the "line" of force

and the "line" of time as the event happens are iconic with the grammatical construal and order

of participants: from Bill to the dog both in terms of trajectory and time. It's a rather natural leap

to claim that the iconcity operates on the discourse level as well, especially since that is

sometimes the case. Part of the challenge of this study is to take this notion of transitivity into

realms other than narrative and determine its validity and usefulness in contexts where

information flow is separate from time flow.

2.5 The Asymmetry Principle in Cognition and Communication

Thus far in the discussion on semantics, the goal has been to reduce the proliferation of

components by searching for commonalities that may unite them on a more abstract level. The

particular relationship singled out has been "asymmetry." The discussion was not meant to

discredit or detract from the semantic work done so far; indeed, Rice's excellent semantic/

cognitive explanation of transitivity has been formative in my own work. Rather, the question is

not whether there are cognitive or semantic aspects to transitivity--certainly, there are-but what

it means that it is these aspects which are grammatically specified. Why should there be

conceptual endpoints in event structure? Why should skewing towards one end of the

Effector/Affected relationship be significant? A preliminary answer lies in the statement above:

asymmetry. Grammatically, where grammar is understood to be the system mediating intemal

representation and communication by providing "...a set of options for schematizing experience

for the purposes of verbal expression" (Slobin, 1996:75), asymmetry provides one means by

which this goal is accomplished. Specifically, asymmetry organizes grammar by presenting sets

of salience relationships that can be systematically attended to through time. That is, one way of

organizing information through time so that it is both processable and communicates what the

speaker intends is through marking salience so that the hearer knows which information is to be

attended to most closely. Regarding transitivity, salience applies to event semantics, participant

semantics, and participant-event relationships as instantiated in the clause.

The "Asymmetry Principle" is simply a label for a set of relationships the linguistic

community has long noted: heads and tails, verb and satellite, topic and comment, source and

goal, binary branching nodes, etc. The only "leap" being made is to say that this is not merely

coincidence, but in fact the cognitive principle along which grammar is largely organized. Note

the claim is being made for grammar only. If, as Talmy (1994) claims, grammar is the cognitive

system mediating what's in the brain and what is communicated, then asymmetry is the (or one

of the) cognitive principles) doing the organizing. One way of doing this is by systematically

creating salience relationships so that at any one time, both speaker and hearer are aware of and

can manage attention to the various elements in the discourse. In the case of human beings, the

central elements appear to be the participants--this is our locus of concern. And our way of

creating and maintaining order is through sets of asymmetrical relationships which manifest

salience relationships so that we know where our attention is to be directed.

If asymmetry as an explanatory grammatical principle is to be taken seriously-as a

reasonable explanation for the patterns evident in transitivity--then it must also be asked why the

set of distinctions clustering in transitivity should do so? Why this particular group of features? I

would propose that the set of features distinguishing transitivity as a system do so because they

are event-oriented--transitivity is a grammatical structuring of events, a kind of event gestalt.

The characterizing element of an event is its linearity, both of motion and of time in an iconic

relationship. Transitivity realizes the most perceptually salient parts of event structure- the

endpoints and the trajectory connecting them. But as Talmy (1994) has pointed out, languages

homologize not only time, but space as well. If transitivity is the grammatical structuring of

events, of "linear" information, then it may be reasonably asked if language does not also

grammaticize "non-linear" information? Does language have a grammatical structure for

information arranged not through time, but in space? Asked in another way, is information ever

conveyed in structures that are not event-driven? As will become apparent, this is not an empty

line of inquiry, but essential to understanding the functional and discourse nature of transitivity.

The role of the assymetry principle will be taken up again in Chapters Three and Four as it

relates to the discourse structuring of narrative and expository texts.

2.6 Concluding Remarks

Transitivity, as we now see, is a well-constrained system expressing certain aspects of

event structure-Starting points, Endpoints, and Trajectories. It is most properly viewed as one

component in the ecology of the clause, specifically that component concerned with the

semantics and pragmatics of Point-Trajectory relationships. At the level of discourse, though,

the effects to which transitivity is put are largely constrained by coherence-and not simply

coherence as a cognitive mechanism, but coherence as it is arrayed by different genres in

different contexts. In fact, at the "level" of discourse, much of what is crucially important for the

formation of clauses is not that critical anymore, and becomes a new and larger "chunk" which is

put to use for even more abstract informational purposes. Like a symphony, if the clause is the

individual player--and the sentence the entire brass or woodwind section- then discourse is

ultimately the conductor.

Figure 2.7: The Four Clause Types of Transitivity

Legend: non-bold graphics--optional elements (probably language specific)
T-- "Little Turtle's Big Adventure"--David Lee Harrison
C-- "Lady of Steel"--Roger Zelazny
H--The Defeat of Smaug, excerpt from The Hobbit--J.R.R. Tolkien
A/B-different participants

P/T Complex: two Points, directed Trajectory
Syntax: bound in same clause
Semantics: two participants, Effector/Affected
Discourse: Basic Perspectival Distance



One morning, a rumble like thunder woke the
little turtle. (T)
Finally, warm rain melted the snow. (T)
Then one day, a boy saw the turtle. (T)
Now he used a great yew bow. (H)
His last throes splintered it to sparks and
...she bathed the wound. (C)
He heard her gasp. (C)

1) Full on the town he fell. (H)
Both attackers fell. (C)
His arms ached by the time he had dealt with
the second. (C)
He started, but it was only an old thrush. (H)
...Smaug shot spouting into the air, turned over...(H).
His battle song broke as he... (C)

2) Birds scolded./Autumn came. /Snow fell. (T)
Everyday, the little turtle walked. (T).
The great bow twanged. (H)
An axe flashed.. (C)

P/T Complex: one Point, directed Trajectory
Syntax: bound in same clause
Semantics: one participant, Effector OR Affected
Discourse: Constrained Perspectival Distance


P/T Complex: one Point, undirected Trajectory
Syntax: bound in one clause
Semantics: one participant, Affected
Discourse: Unconstrained Perspectival Distance

He lay there, clutching his thigh... (C)
..and all seemed anxious to claim the glory...(C)
But the wound extended higher. (C)

(complex constructions)
Marvelling, he found he could understand its
tongue, for he was of the race of Dale. (H)
He could watch mice scamper. (T)
But the little turtle kept on looking. (T)
He tried living in the forest. (T)

And he picked him up and carried down a
shady path. (T)
Suddenly, she had drawn aside his loincloth
to continue her ministrations. (C)
After a time, Edwina helped him to his feet.

2. Perspiration broke out on his brow...(C)
..and his final assailant's head rolled away
after her departing sisters (C)
SThe black arrow sped straight from the string,
straight for the hollow...(H)
0- 00Early one morning, the little turtle started out.
Unafraid, it perched by his ear... (H)
Men with steam shovels and bulldozers were
DISPERSED TRANSITIVES working in the clover fields. (T)
1-two participants
2-single participant

P/T Complex: 1--two Points, directed Trajectory+path
2--one Point, directed Trajectory+path
Syntax: specified across clause plus phrase
Semantics: 1--Effector/Affected
2--Effector OR Affected
Discourse: Constrained Perspectival Distance


3.1 Opening Remarks

Transitivity has long been labeled a complex phenomenon. Indeed, Chapter Two

demonstrated as much at the clause level. The transitivity system is sensitive to the

conceptualization of events, the dynamics of participant semantics, and the communicative

intent of the speaker.

The discourse functioning of transitivity is no less complex, though this complexity arises

more from the interaction of several elements than through the quality of transitivity alone. That

is, just as identifying the semantic responsibilities of the transitivity system entailed teasing out

that which properly belongs to transitivityy" and that which is part of a different system (tense, for

instance), the discourse functioning of transitivity requires the same discriminating eye. DuBois

makes a similar claim: sometimes choosing an "intransitive" was a matter of meaning,

sometimes a matter of discourse (1987:831-832). The same can be said of the transitivity

system in general. Care must therefore be taken not to become too absolute in the assigning of

a "function" to any particular transitive clause. Sometimes a telic transitive is just a telic

transitive. It is only in combination with other discourse elements that transitivity makes its

contribution to the clause as a whole, and by extension, to the discourse needs of speakers and


Nevertheless, there is a discourse function specific to transitivity: the management of

perspective for the purposes of thematic coherence. Several aspects of this statement

must be considered. First, what is "perspective" and what relationships does it have to other

levels of grammar? Second, what about perspective is being "managed" and to what effect?

Finally, what specific role does perspective management play for the purposes of thematic

coherence? To begin this discussion, we must start with "thematic coherence" since, at the

level of discourse, this is the ultimate purpose transitivity serves.

3.1.1 Thematic Coherence

Coherence is intuitively understood as the degree of relevance the concepts and

relationships in a text have with regard to one another which permits plausible inference to be

made about underlying meaning (Crystal, 1987:119). Hence, the content of any given clause

should support what has gone before, add to it, and help predict what is to come next. A good

deal of the research in discourse on topicality, for example, investigates the structural cues by

which predictability and continuity are created.

Giv6n gives a somewhat different structural definition of coherence (1995:327): "The

paragraph (or clause-chain) boundary in discourse [is]...where all types of discontinuity cluster-

discontinuities of reference, spatiality, temporality, aspectuality, modality, perspective, etc.

Taken together, they are the visible, measurable manifestations of a single epiphenomenon--

thematic coherence."

According to Giv6n's definition, thematic coherence is understood by language users at

the beginning or endpoint of a sequence of utterances. His claim is quite specific-- we process

coherence at the level of discourse through discontinuity as opposed to cohesion. Cohesion is

the clause-to-clause link which tells the hearer that this stretch of discourse still belongs together.

When cohesion breaks--is made discontinuous--the hearer understands that a new thematic

block is beginning. This is a highly testable claim and one which the transitivity system by and

large supports.

The function proposed for transitivity is that it manages perspective for the purpose of

thematic coherence, and by inference, for the purposes of cohesion. It seems, as will be shown

throughout the chapter, that transitivity at the level of discourse provides a site for the

management of perspective and, as a clause-level system, contributes to both cohesion and

coherence. This cohesion is not created through linking the transitivity of one clause to another,

nor by linking even one transitive clause type to another. The transitivity of any given clause

doesn't "predict" the transitivity of the following clause so much as it predicts the perspective

from which the following clause may be viewed. Transitivity as a semantic system is about the

whole of cognitive events; transitivity at the level of discourse is about the endpoints of events.

Thus my conception of how transitivity works in discourse is similar to what DuBois claims for

Preferred Argument Structure: it is not a discourse structure, per se, but a discourse preference

for syntactic structure (1987:823).

3.1.2 Perspective

The notion that transitivity encodes perspective, or something like it, is not new. Rice

(1987) states as much in a variety of ways--"At heart, transitivity is a linguistic device optionally

employed by a speaker to conceptualize and organize the actions of entities in the world in order

to convey a certain attitude about an overall event to someone else" (p.5), transitivityy is...a

function of a speaker's evaluation of an event" (p.36), and "...transitivity is as much a function of

the content of the event being described as it is of the describer's interpretation of that event"

(p.38). Hopper and Thompson (1980) make a similar claim when they associate transitivity and

foregrounding, particularly as their claim relates to "telic predicates"--those that are bounded by

tense and participants into a single, "whole" event.

Berman and Slobin (1984) provide the most detailed model of clause-level discourse

negotiation in narrative with their four dimensions of event construal. Briefly, event construal

includes the selection of topic, selection of loci of control and effect, selection of event view, and

selection of degree of agency. Selection of topic is essentially the selection of who or what the

clause will be about. Selection of loci of control and effect is the selection of who or what is

actor and undergoer. Berman and Slobin point out, much as Kemmer does, that the topic need

not be the actor; these two choices are distinct. The topic can be either actor or undergoer (in

my terms, Effector or Affected). Selection of event view is the choice of the event-participant

relationship (Cause, Become, State). And selection of agency is how much "motivation loading"

the entire scene is given. To some extent, selection of agency corresponds to my discussion of

asymmetry as it pertains to slanting towards the Effector or Affected participant. These four

dimensions act together in the construction of a clause and have a subsequent effect on what

can follow in discourse. If there is a fault in Berman and Slobin's approach, it is their total

reliance on semantic parameters with little to no explanation for the structural elements which

link clauses. My approach attempts to fill in this gap.

I claimed in Chapter Two that transitivity is the system underlying "event view," which

Berman and Slobin designate with three views Cause, Become, and State. Here, I am proposing

to operationalize what is meant by the attitude, interpretation, or selectional affect of transitivity.

Simply put, the particular transitive clause type chosen frees up or constrains the participant

perspective available in the subsequent clause. In other words, perspective here is not a

loosely defined or intuitive notion but is actually constrained to participant points of view. As we

shall see, each of the four clause types of transitivity in English has certain freeing or

constraining effects as it interacts with other discourse dimensions (similar to, but not the same

as, Berman and Slobin's). But these effects are limited to predictions about participants.

3.1.3 Management of Perspective

According to above discussion, what is being "managed" is the availability of perspective

through participants. Availability in this case encompasses both anaphoric and cataphoric

signals. Anaphorically, for a participant to be recognized as thematically continuous, it must

already be established in discourse and marked accordingly. Cataphorically, for a participant to

be expected to continue, it must be cast in a grammatical form that hearers know mark thematic

importance. In either case, the form some participant takes opens up, closes down, or maintains

the hearer's interpretation that this participant is thematically important.

There are two widely recognized means by which availability is managed. First is the

nominal form the participant takes. Whether something is fully specified or pronominalized,

definite or indefinite, specific or generic gives critical information to the hearer as to the thematic

importance of the participant. The second strategy is clause structure. Participants in main

clauses tend to be more thematic than those in subordinate clauses. Subjects tend to be more

thematic than objects. For the management of perspective via transitivity, both of these

strategies work together with the four clause types to form stable implicatures by which the

hearer may infer which participants and/or events are likely to continue. As we shall see below,

intransitives provide the least discourse constraint, atelic and dispersed provide the greatest

constraint, and telic transitives provide a middle ground by establishing a limited yet robust

choice of viable participants.

3.1.4 The Model in Brief

When discussing thematic coherence above, Giv6n offered several structural means by

which thematic coherence is established: aspectuality, spatiality, temporality, referentiality,

perspective, and so on. He also goes on to warn discourse researchers to keep in mind that the

visible elements encountered are but heuristic measures through which the unseen, theoretical-

cognitive entity--thematic coherence--is known. It is in this spirit the following model is offered.

The particular play of transitivity in English is not itself a universal; at this point, I cannot even

claim typological ramifications. Rather, transitivity as a system instantiated through the four

clause types has a function which serves the needs of thematic coherence. Specifically,

transitivity provides a grammatical-semantic site over which the construal of perspective may

take place. In English, it is the primary clause-level system through which participants are

introduced, maintained, or removed from narrative discourse.

Nonetheless, given the nature of transitivity as a semantic system, it is hardly mysterious

that it should fulfill the duty of perspective manager at the level of discourse. After all, it is the

transitivity system which takes care of the number of participants and their event-driven

relationships. Nor does transitivity, as we shall soon see, act alone. At least three other

discourse elements conspire with transitivity for the purposes of thematic coherence. They are

treated as part of a "multi-variable model of transitivity in discourse" because, in use, it is rather

difficult to separate them from one another.

The four discourse elements to be considered are:

(1) the relative discourse topicality of the participants of a given clause;
(2) perspectival distance;
(3) the four clause types of transitivity;
(4) the "depth" of transitivity given discourse context.

The first element, discourse topicality, assesses how hearers know the topical status of a

given participants. Two factors must be taken into account here, referential specificity and

information value. The first is my term for the degree of referentiality accorded a participant.

Traditionally, we understand this as a continuum from full lexicalization to zero anaphor,

indefinite to definite. For discourse, the degree of referentiality specified for a participant relative

to surrounding participants is a crucial indicator of thematic importance. In a telic transitive, for

instance, the referential specificity of each participant vis-a-vis the other tells the hearer which

participant will likely continue as topic, hence whose perspective is available for comment.

Information value is a more global term, encompassing but not limited to referential

specificity. It was introduced by Chu (1996:5) to explain topic chaining, overt and zero anaphor,

and word order in Chinese. Basically, Chu distinguishes topichood as operating across two

information structure tiers, the source tier and the management tier. The source tier "is

concerned with where a piece of information comes from" and functions to signal relative

giveness and newness to discourse. The information management tier " concerned with how

a linguistic form in used in terms of informative value". The management tier incorporates a

scale from High value to Low value. As will be shown, it is the effects of discourse topicality on

the management tier which works with transitivity to manage perspective.

The second discourse element, perspectival distance, is the "ground" against which

transitivity is "figured." That is, since Hopper and Thompson's influential 1980 paper, much has

been said in support of their notion that High transitivity (my telic transitives) communicate

foregrounded information. Most often "foregrounded information" has been interpreted to mean

information on the narrative time-line--information which advances the story forward. A few

researchers, notably Kalmar (1982) have argued otherwise, claiming that where high transitivity

may correlate with time line information, it cannot be said that this is the only foregrounded

information in the text. Some other explanation must be sought.

My answer to this question is "perspectival distance." This is a notion introduced by

Talmy (1993,1996) but used intuitively by many discourse linguists with the reference to the

"camera" effects of clause structure. Perspectival distance refers to the distance from a scene or

event created by a particular linguistic construct. A more distant perspective is created via more

inclusive and indefinite nouns with more stative verbs. A closer distance is created with more

specific nouns and more active verbs. In narrative, the closest distance is direct speech where

participants are most immediately engaged with one another.

As will be shown, perspectival distance provides a better measure of the narrative

structure to which transitivity relates. Each of the clause types suggests a range of perspectival

distance. Cognitively, this range constrains what can happen next by creating boundaries

around the scene thereby limiting the number of potential subsequent actions. At the furthest

point, instantiated by intransitives, perspectival distance is constrained fairly little; it serves more

a framing function, suggesting the general direction participants might go given the schematic

knowledge aroused by the scene. Telic transitives are at a "basic" distance, one encompassing

both participants and the event joining them, but excluding surrounding participants. Atelic and

dispersed correspond with a fairly close perspectival distance, where only one participant "fills

the screen" and time segments are shorter. Thus, it is distance from the event and the discourse

implications therein that transitivity realizes, rather than points on the time-line itself.

The third element of the model, the four clause types, takes into account the previous

two. This section gives a fuller explanation of the four clause types and the kind of perspective-

taking they entail. This section takes both a qualitative and quantitative approach in an attempt

to get a feel for the discourse patterning of the four clause types in English narrative.

The fourth element investigates what transitivity and its alternations mean at the level of

discourse. I am proposing here that discourse puts considerable constraints on the kind of

perspective shifting allowed. Thus, where a telic transitive clause out of context may be fully

capable of passivizing, in context, there are allowed and disallowed instances. When alterations

are permitted, I call this "deep transitivity": perspective-shifting which is allowed from the level of

discourse down through both semantics and syntax. "Shallow transitivity" is when shifting is not

allowed and this has a variety of underlying causes. One final note: in this chapter, examples

which are my own are presented in the same font as the main text. Other authors are presented

in an alternative font-type.

3.2 Discourse Topicality

The question of topicality is widely discussed in the linguistics literature. And while an

intuitive or notional definition is still theoretically unsatisfying, it is not entirely erroneous given

the nature of topicality itself. Like "thematic coherence," topicality is largely an invisible entity, a

"felt" understanding which evolves as the speaker and hearer negotiate text. Certainly, topicality

has its visible, measurable markers (pronouns, zero anaphora, topic-chains, etc.) but these are

the overt, linguistic clues to the unseen cognitive entity. Most importantly, topicality functions as

an anchor within text; only within an understood frame of reference can coherent communication

take place. At the discourse level, topics establish boundaries by eliminating other possible

topics and together with general knowledge arouse sets of expectations on the hearer's part onto

which information is pegged. (For an excellent discussion of approaches to topicality, see

Goutsos, 1997.)

Intuitively speaking, topicality is the quality of being "about" something, wherein the

speaker has established a discourse within which some particular participant, event, or scene is

understood by the hearer to be referred to and/or is expected to continue. Topicality is

established through repeat reference to wholes and to parts of participants, events, or scenes.

Discourse topics are most often established through reference to parts. The hearer has an

expectation of what is going on--a schema or script--and the hearer's understanding of the

speaker's actual discourse topic is created through the negotiation of expectation based on the

schematic knowledge and speaker's actual linguistic cues. Thus, essentially, it is a matter of

inference. In this case, not merely inference, but stable linguistic inference-an expected and

normal part of linguistic competence--hence a matter of implicature.

Topicality has three measurable levels, the clause-level, paragraph or inter-clausal level,

and the discourse level, yet these three levels are not often distinguished in the literature. Most

often, "topic" is defined at the clause-level with the somewhat unenlightening phrase "what the

sentence is about," including descriptors such as the initial element in the sentence, the

proposition about which the speaker is providing more information, or the given elements of the

utterance (Chu, 1983;Gundel, 1988). More satisfying definitions were offered by Chu (1983),

Prince (1981), Chafe (1994), etc. who defined topic in terms of its cognitive status in the

discourse: "Semantically, a topic is definite in the sense that it has to be a particular entity or

event that has already occupied the mind of the speaker and hearer" (Chu, 1983:132). Prince

and Chafe developed models of "activation states" to account for the variety of definite noun

phrases that occur in discourse, including those that the original conceptions of topic could not

account for: I saw a new house today, but I didn't like the kitchen. It was too small/I saw a new

house today, but I didn like the neighborhood. It was too crowded/I wanted to see houses

today, but the realtor wasn't feeling well We have to schedule for next week. In each of these

sentences, the conjunct clause contains a noun phrase which is new to the discourse, occurs in

the site of new information, but is definite and the topic of the next sentence. Prince explains

this through types of inference wherein any nominal excites a field of related terms which are

allowed into the discourse without explicit previous mention. The nominal are all inferrable

from various kinds of pragmatic knowledge. More recently, Prince has proposed discourse-

based degrees of givenness from the perspective of hearer-discourse interaction (1992,

summarized in Chu, 1996):

Hearer Discourse
old old already evoked
old new has not been evoked; speaker believes it's
old to hearer
new new has not been evoked; speaker believes it's
new to hearer
new old has been evoked; speaker believes hearer
is not aware of it

Prince's proposals together have the advantage of covering at least two sources of discourse

information, semantic/pragmatic and discourse-negotiated.

Other distinctions created in discussions of topic include such categories as discourse

topic versus sentence topic. Where Ochs, Keenan, and Scheifflin explicitly identify their topic as

a discourse topic, Lambrecht (1994) claims his discussion is of a sentence topic and is to

explicitly contrast the "focus" of the sentence, which is the new information. But what linguistic

features actually distinguish a "discourse topic" from a sentential topic? Givbn claims the

distinction is quantitative in nature "...the main behavioral manifestation of important topics in

discourse is continuity, as expressed by frequency of occurrence (1984:138). For example,

pronominalization is a direct indication of activation status in discourse; whatever is

pronominalized must be known to both speaker and hearer already. However, this does not

much help in determining topicality in discourses where there are multiple participants. Nor does

it provide a means to distinguish a discourse topic from a sentential topic since that which is

sententially frequent is also important to the discourse.

Chu (1996:5) provides a better explanation of the distinction by proposing instead two

information tiers with different discourse functions, the source tier and the management tier.

"The source tier is concerned with where a piece of information comes from while the

management tier is conceded with how a linguistic form is used in terms of informative value".

For Chu, then, the previous definitions of topic in terms of cognitive activation status are

explanations of the source tier--where the information comes from. Giv6n and Lambrecht's

propositions discuss topic from the perspective of management, what happens to information

once introduced into the discourse.

Another difficulty with Lambrecht's contrast of topic and focus is the conflict arising when a

topical noun receives focal stress as in "Who left the dishes on the coffee table? Your father, of

course"? It would be absurd to claim that "your father" is not a known element. At the same

time, the phrase is clearly the focus element, the "new" information insofar as it is the answer to

the question. Chu (1996:5) also provides an answer to this dilemma by proposing that topic be

understood not simply in terms of given and new information, but functionally as high or low

information value: "A better way to interpret it is whether and to what extent a linguistic form

serves to add information to what has already existed in the repertoire of the hearer/reader".

Chu further proposes that high informative value and low informative value (hereafter, HI-value

and LO-value, respectively) represent a dine of values, not a binary opposition:

LO-value<------------------------------- -----------------------.--. HI-value

subj w/ < subj. w/ < assert. pred. < assert. pred.
given info new info w/ given info w/new info

The discrimination of a source tier and management tier significantly impacts our

understanding of topichood. Much of the work that has been done on sentence-level topics has

dealt with the source tier; and much of the work on that tier has been aimed towards

understanding how various discourse markers such as definiteness are used with "new"

nominal. Chu brings to the table another aspect of discourse structure, the information value of

any given nominal. This concept focuses not on where information comes from, but how it is

used, "managed," in discourse. Further, "information value" allows us to operationalize the

impact of any given nominal by defining "new" information or "focus" as any information which

adds to or amends the state of knowledge of the hearer. Thus, what is "new" in terms of its

source does not have to be high in information value, nor is what is given necessarily low in

information value.

It quickly becomes evident that "source" and "management" work together. In English, it

has been traditionally claimed that given/new distinctions are associated with the subject and

object positions, respectively. When a nominal occurs in the subject position, it is considered to

be the sentential topic, and by definition known to speaker and hearer. By contrast, the nominal

in the object position is the "focus" because it is part of the asserted information. These

correlations hold in default cases; that is, if they are not otherwise challenged. Complications do

arise when texts are investigated.

(I) Uttering a curse in his well-practiced falsetto. Cora swung his blade
and cut down the opposing swordswoman.
"Lady of Steel." Roger Zelazny

(2) She came into my shop with a gash in her thigh and blood seeping
out a wound in her stomach.
"Bra Melting." Janni Lee Simner

(3) "But you don t tax jockstraps!" Mirabel Stonefist glared.
"And the Ladies of the Club," Elizabeth Moon

Each of the texts above not only introduces a new participant, but is the first line of the story. In

fact, it is not unusual to begin a short story in English with the introduction of a new participant

rather than a scene-setting. In these cases, the first appearance of a nominal in the text is also

its first use. Just as givenness alone does not determine information value (information value is

determined by use), neither does the grammatical role determine information value.

In these three excerpts, the new participant is in the subject slot, the site of "given"

information. As stated above, it is also the site of topical information. In other words, one way of

introducing a topical participant into the discourse is to place them in the subject slot, thereby

indicating to the reader the relative importance of the participant. It's as though the participant is

simultaneously being flagged as "new" and "given" at the same time; in fact what is happening is

the participant is being flagged as "new" and topical" at the same time. In terms of source, the

participant is brand-new; in terms of management, the participant is high in information value.

At the same time, not all participants making their first appearance in the object position

of the asserted predicate are HI-value; in fact, many are not. In the Zelazny text, the second

"participant" occurs in the site most frequently associated with High information value: Cora

swung his blade and cut down the opposing swordswoman. According to Chu's management

continuum, this is typically the position of HI-value participants. However, according to Prince's

"source" categories, this participant is contextually inferable. We rather expect as readers that

anyone uttering curses in a well-practiced voice while swinging blades is most likely cutting down

an opponent (as opposed to cutting down trees, for example). That it is a swordswoman is

interesting, but does not hold our attention; though marked with the definite article, the

swordswoman is not mentioned by name, is not specified for the reader. If instead the sentence

read Cora swung his blade and cut down the opposing swordswoman, Hilda we would be

disturbed as readers when the following sentence began with "His contoured breastplate"

referring back to Cora (rather than to Hilda).

We need to explain why participants new to the discourse and subsequently topical may

be used in the subject position when first introduced. And likewise, why participants in the object

slot, such as the opposing swordswoman, are not topical. It's obviously inadequate to label

subject and object positions as merely slots for old and new information respectively. Rather,

there is a collocation of functions for each position which serve to designate information value.

These include referentiality, action schemata, and perspectival distance. (Perspectival distance

will be taken up in greater detail in the following section. Suffice to say here that you should find

low occurrences of "close focus" constructions with new participants unless placed within highly

recognized activities, such as the conversation in (3).)

Giv6n (1995:379) states much the same thing: "The grammatical subject, the clause's

primary topic, code the event participant that is most continuous--both anaphorically and

cataphorically" (emphasis added). Subjects tend to have greater continuity, greater topicality,

head action chains, and have lower information value. Objects tend to have less referential

distance and less topicality but are receivers of action and also tend to have higher information

value. English has worked out a system where participants are referentially coded apart from

grammatical position so that the semantics of an event can play itself out unhampered. I call

this system "referential specificity." Essentially, given two participants, whichever one has a

stronger degree of referentiality is regarded as topical. Degrees of referentiality are assessed on

two scales, the definiteness hierarchy from zero anaphor to generic and the animacy hierarchy.

DEFINITENESS: zero --> I/you --> proper noun --> definite --> indefinite --> generic
ANIMACY: human --> self-instigating animate --> non-self-instigating animate --> inanimate

(4) A headache the size of her healthcare plan--no, better make that the
national deficit--was turning Hillary Rodham Clinton's skull into the
local percussion section.
"Exchange Program." Susan Schwartz

(5) It was shortly after Mrs. Batchett left the planetarium that she saw
the fairy, the elf. and the gnome.
"On the Road of Silver," Mark Bourne

In (4), the two scales work together. A headache, although the Effector, is clearly not the

thematically important participant, for two reasons. First, it is lower in referential specificity

(hereafter, RS) than the second participant-a proper name is higher on the definiteness scale

than an indefinite. Second, it is inanimate and non-human whereas the object is human. Also,

'Hillary Rodham Clinton' is a well-known figure and is occurring in the site typically associated

with High information value. Contrast this sentence with (1), from "Cora." In this opener, the

subject of the sentence is the Effector--the most unmarked of action chains--and is named by a

proper name. The object, while also human and strongly Affected is not named, but referred to

via a definite descriptive noun phrase. Because the Affected participant has a lower RS than the

Effector, the tendency for the object to become the subsequent clausal topic is overruled. Note

that the author could have chosen a combination of atelic transitive and a subordinate clause,

thereby eliminating the coding competition for topic (generally speaking, subordination is a

stronger discourse factor in backgrounding): Uttering a curse in his well-practiced falsetto, Cora

swing his blade forcefully, cutting down the opposing swordswoman. In this sentence, there is no

competition for clause-level topic and no need for RS as a deciding discourse factor.

Example (5) combines an interesting mix of definiteness and clause structure to cue the

reader in on who the topical participants are. Obviously, Mrs. Batchett is the prime candidate for

immediate sentential topic because she is both in the subject position and of higher RS than the

other participants mentioned. On the other hand, the It in the opening phrase It was shortly after

Mrs. Batchett left sets up the reader to expect the situation the pronoun obliquely refers to. As a

result, although the reader knows that Mrs. Batchett is immediately the most important character,

the definite form of the other participants together with the oblique reference at the very

beginning of the story alludes to the later importance of the fairy, the gnome, and the elf. In fact,

these three do play significant roles later in the narrative.

Sentence topics, which can be either events or participants, are also established through

repeat reference to their wholes or parts. Further, it is at the level of the sentence that referential

specificity plays a part in the setting up of hearer's expectations about what is coming next.

Basically, given two participants, whichever has the highest referential specificity is expected to

continue. If both are of equal value, the participant in the object slot is more likely to be topic in

the next sentence because the object is the site of higher information value. However, if both

participants are already established in discourse, then it could be either participant or the event

which gets topicalized for the next sentence.

Finally, the various discourse effects of subject and object, given and new, referential

specificity, and high and low information value all act together to form the "discontinuities" Giv6n

refers to as marking thematic boundaries. When the above factors combine to counter reader

expectations, this is also a signal that a juncture has occurred. The chart below summarizes

these effects. Subject and object are crossed by given/new and higher or lower RS. These

result in information values which then signal various states of thematic continuity and



HI-Value -- "discontinuity" discourse juncture
GIVEN LOWER RS LO-Value "continuity" thematically established

OBJECT NEW HIGHER RS HI-Value "discontinuity" new clausal topic

GIVEN LOWER RS LO-Value "continuity" est. topic continues

Figure 3.1: Referential Specificity, Information Value, and Grammatical Relations

If the subject is new and has a higher RS than the object, then it also has higher information

value and is likely occurring at a discourse juncture (a new paragraph, for instance).

3.3 Perspectival Distance

We've already seen that the narrative time-line is structurally a matter of implicature

rather than a linguistic entity in its own right. We also have shown that events on the time-line

do not have to be telic transitives, but may be atelic and dispersed transitives as well. Yet, there

is certainly evidence indicating that there are textual interactions between transitivity and

narrative. If neither time-line or plot-line are sufficient guides, what else could account for the

regularities observed so far?

Talmy (1994) offers a possibility with his notion of perspectival distance. In essence,

Talmy shows that regularities in certain transformations are most easily explained as a matter of

implied/perceived distance from the event. The further "away" from an event a linguistic

description is cast, the less detail of the event is implied in the description. The "closer" the

linguistic depiction is, the more detail is implied. For example, the following sentences can be

understood not simply as aspectual distinctions, but as a matter of distance from the event.

(6) There was a cat and dog running down the sidewalk.
(7) The cat ran after/chased the dog.
(8) The dog fled.

In (6), the perspectival distance is far away, what I term a "panoramic view." Essentially,

an event is construed as a state; the distance is so far away that the action itself is seen as an

entire picture, a framed Point-Trajectory complex, rather than as movement from one endpoint

to another. From the point of view of discourse, just about anything can follow this kind of

sentence as long as it stays within the very loose constraints imposed by the overall action-in

this scene, that of a neighborhood.

(9) There was a dog and cat running down the sidewalk, children swinging
in backyard playgrounds, and a light playful breeze nipping at the leaves
and swishing little girls' ponytails this way and that.
(10) There was a dog and cat running down the sidewalk. I opened the door
slowly, peeking out again to make sure the coast was clear.
(11) There was a dog and cat running down the sidewalk. Without warning,
a bomb ripped the morning's peace to shreds. No more dog. No more cat.
No more sidewalk.

The perspectival distance in (7) is what I term "basic." Similar to Lakoffs notion of basic

level terms, this is the sort of sentence most often given when someone asks for an example

sentence. There are two endpoints and a trajectory given, but instead of being framed from a

distance, they are framed at the level just enclosing the activity and participants. Often,

additional information is excluded at the level of the clause; if there is additional information

about direction and such, it has a much more "adjunct" interpretation. This is also the level at

which discourse constraints begin to assert themselves. This example includes an Effector and

Affected with equal referential specificity engaged in a purposeful, unilateral activity. It is NOT

the case that just anything may follow this statement. Rather, as will be discussed in the next

section, we expect that the Effector, Affected, or Predication will follow. This is considerably

more constraint than the panoramic view entailed.

(12) The cat chased the dog. The dog scrambled down the sidewalk, yelping
in despair.
(13) The cat chased the dog down the sidewalk. It meowled fiercely as it
closed in on the hapless Rottweiler.
(14) The cat chased the dog down the sidewalk. But it was a humid day
with a baking sun, and the chase ended as quickly as it began.
(15) ? The cat chased the dog. The earth tilted slightly on its axis and
a car backfired in Detroit.

The final perspectival distance I call "close-focus." Essentially, it is as though the

linguistic camera were directly on top of the scene with only a single participant filling the lens.

Time sequence here is very short, a matter of moment-to-moment experience. Discourse

constraints are strong, limited to participants only. Further, only participants set up by discourse

frames or directly entailed by the predicates chosen are allowed. Narratively, close-focus is

often experienced as fast-paced, action-filled, tense, and dramatic sequences of central

importance to the plot. Also, note that with (7), the verb "fled" implies something from which the

dog is fleeing and potentially, something to which the dog flees.

(16) The dog fled. It ran faster and faster until it escaped the cat.
(17) The dog fled. The cat leaped. The dog snarled in fury as
the feline landed on its back. Raising a pad of extended claws,
the cat went to scratch, but too slowly! With a snap, the dog chomped
the proffered paw. And that is how the neighborhood tabby became
known as Tripod Tom.
(18) ? The dog fled. The sun rose. Sally baked chocolate chip cookies for breakfast.

It is obvious at this point that there is a connection between perspectival distance and

transitivity. Panoramic views correlate with intransitives, basic views with telic transitives, and

close-focus with atelic and dispersed transitives. While I am not ready to claim that transitivity is

the encoding of events vis-A-vis linguistic space (as does Langacker, for instance), it is clear that

at the level of discourse, narrative structure can easily be visualized as a matter of distance from

events. The narrative introduction, particularly in longer works, usually begins at a panoramic

distance describing the general scene. Thematic participants and events are introduced through

actions while the narrative proceeds to climactic moments. These are often construed as close-

focus scenes, with all the reader's attention taken up with actions and events portrayed moment-

by-moment. The denouement goes back through the basic level to the panoramic as the scene

or story winds down to a point of resolution. In a novel, this pattern repeats itself again and

again. It is violated to great effect in multi-participant stories where each chapter focuses on a

single thematic character and the middle chapters of the book end in cliffhangers until all the

participants can be brought together in a final resolution.

There is a fourth level of narrative distance I call "immediate." This is the domain of

direct and indirect speech. These two, particularly the former, are the most cognitively involving

aspects of narrative, even though presented in past tense morphology (s/he "said.") Direct

speech has a sense of being close to the present moment, as though the reader where sitting

very near the participants. This may be why in spoken narrative, direct speech so often happens

in the historical present. There is considerable debate about whether direct speech should be

considered a transitive event (see Longacre, 1983). While it is possible to passivize speech

statements (using "was said by him/her" instead of "s/he said"), it is rarely used in practice.

Further, there is better evidence cross-linguistically that direct and indirect speech forms are

formulaic and thus distinct from grammar. I agree that direct and indirect speech are "different"

on several linguistic accounts and probably should not be considered as telic ditransitive events

(passivization implies the rather odd notion that speech is a physical object given to another

participant). As a speech act, direct and indirect speech match up with atelic transitives which

feature a single participant at an extremely close distance. Suffice to say at this point that direct

and indirect speech present a problem this project does not solve. They will not be considered in

the data as events coded by the transitivity system, but as speech acts with cultural expectations

that go along with them. These expectations may interact with transitivity or they may not.

At this level of discourse, it must be remembered that patterns are correlations. In other

words, telic transitives do not mean "basic level perspectival distance." They mean an event

encoded to include both points and the trajectory in a single linguistic unit. The appropriate way

to view this matter is from the perspective of the largest structural unit, the narrative itself.

Narrative structure has a format and induces a set of reader expectations. The author has at his

or her disposal all of the lexical and grammatical options of the language they are writing in.

Certain of these options better fulfill the structural requirements of narrative at its various points.

But as long as the narrative expectations are fulfilled, there is relatively little constraint as to

which option the author chooses. This is, more or less, a linguistic definition of style. It also

helps account for why statistical analyses are useful at the level of discourse. Given a particular

point in a narrative, there is approximately a 70-80% chance of finding structure X. There is a

20-30% likelihood of finding structure Y or Z since both fulfill the structural need in question. It is

simply a matter of conventional expectation and style that determine X, Y, or Z.

Nevertheless, at some point it becomes a "chicken-and-egg question" to ask whether

some given grammatical structure "has" some given discourse function or whether discourse

"makes use" of the semantic/pragmatic potentialities inherent to any grammatical construction.

The whole question is most likely wrong. It is not a matter of what comes first, but a matter of

linguistic ecology. Grammar and discourse have a relationship and to suppose either is more

fundamental is erroneous.

In conclusion, if we separate perspectival distance from time line, and associate

transitivity with the former, we get a more accurate picture of what transitivity accomplishes in

narrative. The time-line is essentially a structural artifact of narrative, created through

implicature by the tense-aspect configurations of successive predications. Perspectival distance

is the linguistic boundaries imposed on a scene at any point in narrative. Transitivity is one

linguistic system available to the speaker for this purpose. The specifics of how transitivity works

to fulfill this function will be taken up in the following section.

3.4 The Four Clause Types of Transitivity

It was claimed in Chapter Two that Endpoint-Trajectory (hereafter P/T) relationships are

expressed by four basic grammatical types in English--intransitive, atelic transitive, telic

transitive, and dispersed transitive (with the caveat that intransitives may be more property

understood as durative, unbounded actions, and states should be considered a

semantic/conceptual case in their own right). This chapter proposes that the discourse function

of P/T relationships is the control of communicative intent through manipulation of perspective.

Simply put, perspective is the participant-event relationship as it operates at the level of

discourse. In terms of grammar, perspective is understood to be the grammatically available

points of view from which a scene/event may be construed. For English, there are three basic

points of view: the Initiating Point, the Ending Point, and the Trajectory itself. Semantically,

these correspond to the Effector, Affected, and Predication, respectively. In terms of clause

structure, the three points of view get played out in the four clause types of transitivity.

For discourse, perspective-taking opens up or closes down participant availability. That

is, given context and transitivity, regular predictions can be made about who or what is coming

next in a text. Each of the clause types of transitivity interact with discourse topicality and

perspectival distance to form the part of the implicatures we recognize as thematic coherence.

3.4.1 Telic Transitives

At the level of discourse, the particular P/T configuration lets the reader know who is

available for perspective shifts or changes. The choice of a configuration either opens up or

limits what can come next. In the broadest sense, given a telic transitive, there are three basic

perspectives that can be appealed to: Effector, Affected, and Predication. In the following

example, the subsequent sentence can topicalize the Effector, Affected, or even the action itself.

Further, for the last option, the sentence can be skewed towards the Effector or Affected

participant or simply comment on the action itself.

(19) John kicked Bill. He punched Bill/him. (Effector)
(20) John kicked Bill. Bill yelped in pain. (Affected)
(21) John kicked Bill. It was a really hard kick. (Predication)
The kick really hurt. (Affected)
The kick was really hard. (Effector)

Even these simple sentences show some interesting things regarding reader

expectation. With sentence (19), where both Effector and Affected have equal referential

specificity (hereafter, RS), it is somewhat awkward to continue the Effector role as topic without

additional maneuverings. Note the sentence reads better if temporal adverbs are added in.

(22) First, John kicked Bill. Next, he punched him.

As stated in the discussion of discourse topicality, if both Effector and Affected are of equal

referential value, the Affected role generally wins the role of topic because of the association of

new information with the object slot. Thus in (19), despite the fact that according to sheer

transitivity both participants and the action are available for topicalization, the likelihood is tipped

away from the Effector role. If the Effector is to continue as topic, then the balance must be re-

shifted somehow. This can be partially done through pronominalization, though with an activity

of such violent energy, pronominalization alone doesn't quite work; the two sentences still don't

"flow." Coherence is better accomplished here through temporal adverbs which make salient the

linear aspect of the activity. In other words, once the reader knows that John "first" did

something, the reader expects to know what John did next. Adverbs impose an overt temporal/

spatial structure on a text, creating strong expectations about what will follow.

The second example, with the Affected taking the topic role, is the most straightforward

and unmarked of the options. Effector and Affected have equal RS and the Affected participant

is the site of highest information value. This role is allowed just about any kind of response,

though; new information is not confined to nominal.

(22) John kicked Bill. Bill yelped in pain.
(23) John kicked Bill. Bill yelped in pain, then kicked him back.
(24) John kicked Bill. Yelping in pain, Bill kicked him back.
(25) John kicked Bill. Bill slugged him.
(26) John kicked Bill. Bill started to cry.
(27) John kicked Bill. Bill stared at him in astonishment.

Obviously, we are confined to our expectations of the scene, the action schemata, a factor

essential to coherence.

(28) ?John kicked Bill. Bill painted flowers.

As a testament to our desire for coherence, many readers of this sentence unconsciously assert

some sort of modifier to make coherence kick in:

(29) John kicked Bill. Bill continued painting flowers.

The addition of "continued" makes the proposition refer back to a previous state which had been

interrupted by John's kick. Even though we don't know what this state is exactly, we are satisfied

that the sentence "makes sense."

The final option that a telic transitive allows is some comment on the Predication itself.

In this case, the term "Predication" includes the whole of what the scene implies. Thus, given

any particular activity, subparts of the activity are available as well as the whole of the activity


John kicked Bill. Part of Activity/Skewing
(30) The kick was really hard. whole/ Effector
(31) It really hurt. whole/ Affected
(32) It was a really hard kick. whole/ Predication
(33) His foot made hard contact with Bill's shin. subpart/ Effector
(34) His boot swung in a long arc toward Bill's leg. subpart/ Effector
(35) His pants nearly ripped with the effort, subpart/ Effector

The sentences (33) -(35) show two things. First, what is available in terms of Predication is

limited by the action itself and what the reader knows must be the case in the scene. But within

that context, anything is fair game from some comment on the activity to the clothes the Effector

is presumably wearing. Second, when referring to a subpart of the activity, it is very difficult to

skew towards the Affected participant without first topicalizing that participant, or at least re-

mentioning it. The energy of the activity flows from Effector to Affected and thus the subparts of

the activity follow the same path. This is not true for the whole of the activity because by

definition the trajectory in a telic transitive has both an Initiating Point and an Ending Point,

hence both points are available semantically.

Narrative examples of the above points are as follows. In the first one, we see an

example of a telic transitive followed by another telic with the Predication as the thematic

participant of the sentence. The second excerpt shows the case of a telic followed by a clause in

which the object has become the subsequent topic

Uttering a battle-cry in his well-practiced falsetto, Cora swung his
blade and 0 cut down the opposing swordswoman. His contoured
breastplate emphasized features which were not truly present.
"Cora," Roger Zelazny

Polyta stepped over to Colleen and 0 rubbed the dog's head for
comfort. Colleen whined softly in response and 0 rubbed her cheek
against the woman s thigh.
Manannan s Isle, L.S. O'Brien

Excerpt three is a good example of telic transitives working with information value to produce a

text in which the anthropomorphized machines are cast as the villains. The third excerpt also

shows the range of exceptions to which text is prone.

One morning a rumble like thunder woke the little turtle. Men with
steam shovels and bulldozers were working in the clover fields. The
machines pushed over trees. They dug up clover. They tore down hills
and 0 filled up holes. The machines made so much noise that the turtle
couldn't hear the birds or the frogs.
"Little Turtle's Big Adventure," David Lee Harrison

Here, the first sentence sports an indefinite subject acting as Effector on the established

protagonist. The object, the little turtle, has low information value compared to the subject,

which is indefinite, Effector, and occupying the site of thematic continuity. This combination

fairly begs explanation and overwhelms RS as a deciding factor in participant topicality.

It is also interesting to note how similar in effect the semantics and pragmatics of this

construction is to inverse marking in proximate-obviative languages. For example, in Cree, an

Algonquian language, a third-person narrative must consistently mark proximate and obviative

participants. Briefly, the proximate participant is the one whom the reader is intended to keep

track of--it is the changes to the proximate participant which are of greatest concern. All other

participant are morphologically tagged as obviative. Literarily, the obviative encompasses roles

as diverse as "companion," "antagonist," and "narrative prop."

Pragmatically, there is also pressure put on the text to conflate "agent" and "proximate

participant." When this is not possible because the proximate participant is being acted upon,

special morphology is used to signal this: inverse markers. The inverse form is statistically rare

and marks those sentences in which the pragmatics of the proximate-obviative distinction

conflicts with the semantics of the event being communicated. Inevitably, these are points of

high drama during which the proximate participant faces some great challenges to be resolved

throughout the rest of the story.

English marks this kind of conflict as well, though not morphologically. Instead, various

semantic and syntactic structures are used to "re-arrange" the reader's perception of who the

"proximate" and "obviative" characters are. Semantically, this contrast is created through the

amounts and kinds of information given about a participant. The more the reader's knows about

the character's motivations, intention, history, and reactions to a given event, the more likely that

character is playing a "proximate" role. This is the literary protagonist; morphologically marked

in Cree and notionally tagged in English. The antagonist or narrative props are those

participants who act in an English narrative, but whose motivations, intentions, and history are

less known. All in all, the "bad guy" is the least known character in an English narrative.

In the "Turtle" text, these relationships are being established in the first sentence. This is

the second paragraph of the story and provides the motivating event for the conflict the

protagonist must resolve. Grammatically, this sentence features a force working upon the

protagonist. Now, it would have equally felicitous to have written the sentence One morning the

little turtle was awakened by a rumble like thunder. There is no grammatical discontinuity

whatsoever and the event of "waking up" is sufficiently familiar to children to ameliorate the

effects of using a passive. But the pragmatic force of the sentence would be lost; the sentence

as written succeeds in being dramatic and focal precisely because it flouts conventional

expectations of participant relationships. Like the inverse construction in Cree, and indefinite

Effector in English provides information on both the semantic and pragmatic levels of


3.4.2 Atelic Transitives

Atelic transitives provide the most profound perspective shifts in that they present only

one side of the story, so to speak. Functionally, atelic transitives are important for two reasons:

(1) the introduction of new participants, especially in verbal narratives; (2) close-focus scenes

when the action is presented in very small time periods.

In verbal narratives, DuBois (1987) demonstrates that one-place predicates occur for at

least two reasons, the introduction of new protagonists and semantic necessity. He offers

quantitative evidence from several studies to show that when speakers bring into a text a new

protagonist--a participant who will continue thematically and from whose perspective some part

of the tale is told--they do so overwhelmingly in one-place predicates. Other uses of one-place

predicates are dictated by semantic necessity, he claims. Speakers use them because they

have the right meaning.

In written narrative, the patterns are somewhat different. First, new participants may be

introduced into the text in any number of ways, though the most common are through telic

transitives (where the second participant is of lower RS) and atelic transitives. The atelic

transitive, though, embraces two possibilities. First, the classic scene-setting introduction which

begins with a description of a place and narrows down to a statement about the central

participant. This sort of introduction is the norm in novels and longer short stories, but is also

found in truly "short" short stories as well. The second use is not exactly atelic--the use of

opening conversation. In this case, as stated previously, it is difficult to know whether direct

speech is telic, atelic, or neither. It is most definitely a case of one-sided perspective which is

a feature of atelic transitives. Direct speech also strongly implies a second participant, someone

who is being addressed. What we find with stories that have conversation as their opening lines

is the same use of RS that we find in telic transitives. If one of the two conversationalists has a

higher RS, that participant is interpreted as the topical participant (in narrative terms, the

protagonist; but this is a limiting term since sometimes stories are about more than one person,

and when this is the case, the story is "about" them both, or even about an unseen discourse

topic). If they have equal RS, more of the story must take place to settle the difference. In

some cases, the reader comes to find that both of the participants are topical, and some third

character or some other event intervenes which confirms their equal role.

It is important to note at this point that I am not claiming that transitivity itself has

anything to do with the setting of a protagonist through direct speech. Rather, the fact that

conversation in narrative follows similar patterns as transitivity demonstrates that it is the

discourse functioning of perspective that is the driving force. In both cases, conversation and

clause structure, aspects of topicality are put to use in order to achieve thematic coherence.

"But you don't tax jockstraps!" Mirabel glared.
"No," said the king. "They're a necessity."
"For you. maybe. How do you expect me to fight without my bronze

"Men can fight without them," the king said. "It's far more
economical to hire men, anyway. Do you have any idea of what the
extra armor for the women in my army costs? I commissioned a military
cost-containment study, and my advisors said women's uniforms were
always running over budget." The king smirked at the queen, on her
throne a few feet away, and she smirked back. "'I've always said the
costs to society are too high if women leave their family and
"We'll see about this." Mirabel said. She would like to have seen
about it then and there, but the king's personal guards--all male this
morning, she noticed--looked too alert.

"And the Ladies of the Club," Elizabeth Moon

The above text is the opening paragraphs of the story. Two things establish Mirabel's

place as topical participant. First, she has highest RS; her proper name is used while both the

king and queen, participants of greater social power than Mirabel's, are named only through

definite reference. Second, Mirabel is grammatically topicalized in the final paragraph through

repeated pronominal reference. The king, on the other hand, though he is the second

conversationalist, never gets pronominalized and is consistently marked as a non-topical

participant because of this.

The next examples are taken from the opening lines of Star Trek: New Frontier, a series

of four novellas. All four books are related to one another not only by use of characters and

setting, but through close chronology. The four smaller books are in essence one long story.

Book One. House of Cards
Falkar regarded the remains of his troops and, as the blazing Xenex sun
beat down upon them. 0 decided to wax philosophical about the situation.
"It is not uncommon to desire killing a teenager," he said. "However, it is
not often that one feels the need to send soldiers to do the job."

Book Two, Into the Void
Elizabeth Paula Shelby, gaped at Admiral Edward Jellicoj. Hej could
not have gotten a more stunned reaction out of heri if he'd suddenly
ripped off hisj own face and revealed himself to be a Gorn wearing an
exceptionally clever disguise.

Book Three. The Two-Front War
"1 want to blow those bastards out of space." 0
The Excahbur had just been rocked by the opening salvo from the
black-and-silver ship that hung 100.000 kilometers to starboard.

Book Four. End Game
The refugees from the Cambon bleated in fear as they were herded into
a large auditorium. Pacing the front of the room was the woman whom
they knew to be Laheeraj...apparently, a high muck-a-muck in the
hierarchy of the world of Nelkar. Shej looked at them angrily, her fury
seeming to radiate from her in such a manner that is measurable by

The first excerpt begins with a telic transitive, though a fairly "shallow" one. It wouldn't

be grammatically impossible for this sentence to passivize, however it would be highly unlikely,

given the overall context. It is the opening sentence, the subject of the sentence is the Effector,

and the object is of lower RS. The reader expects that Falkar is the topical participant and this is

the case.

The second excerpt begins with, according to Rice (1987), a prepositional telic transitive.

Again, the sentence could grammatically passivize (though there is greater disagreement among

native speakers with prepositional telics). Further, should the sentence passivize, the following

sentence could still be used and would still make sense. This is an example of "deep"

transitivity. Nevertheless, the author has chosen to go with a straightforward telic with equal

participant RS, skewing reader expectations towards the object. And this is precisely what

happens. The following sentence features Jellico as sentential topic, albeit it in a somewhat

unusual example of dispersed transitivity-the path aspect is construed across the "mental"

domain which encompasses both intellectual and emotional states. Two features of thematic

coherence are preserved. First, the greater thematic continuity of topic is retained; even though

Shelby is not the subject of the second sentence, the statement is still about her reactions, not

his. Second, intra-clausal cohesion is maintained by following the pattern expected of telic

transitives with equal RS. Jellico does become the subject of the following sentence. In fact, the

author has made very effective use of the competing motivations of the subject and object roles

in English.

The third excerpt appears misleading until it is remembered that this is a series of books

in close chronological succession. Book Two ends on a cliff-hanger; Book Three begins

precisely where Two leaves off. Book Three begins with direct speech from Captain MacKenzie,

who is the central participant of the series; the four books are about him. Thus there is no

reference to him as speaker. Rather, the reader picks up the story just after the ship has been

hit. The passive used in the second sentence places MacKenzie's ship, the Excalibur, as topical

participant, the receiver of the action and also the object of the reader's sympathies. The

narrative continues from there. It is also clear here that scenes need not begin with any

particular kind of transitive clause. Functionally, the scene must begin and narrative exploits a

number of possible ploys to do so.

The fourth excerpt is an example of the atelic introduction. Although the reader knows

all the participants involved, Book Four begins at a new place (unlike Book Three). The

refugees are established as the topical participant with an atelic clause joined to a subordinate in

which they are the subject of a passive. Recall that with a telic transitive, any of the three

possible topicalizations are available: Effector, Affected, Predication. The same holds for the

passive which is simply a telic transitive with a topicalization switch. Thus, aspects of the scene

which are available include who is doing the herding of the refugees. That is what comes next.

The refugees are herded into an auditorium, information which is offered in the same clause.

Thus it is unlikely that the environment the scene is taking place in will be topicalized. The next

participant who is introduced is Laheera, whom the reader knows is the "bad guy" behind the

current troubles. She is still presented from the refugees' point of view; they as yet do not

understand that she is the antagonist. Hence she is consistently spoken of in either the passive

(sustaining the implication that she is the subject of someone else's actions/attention) or with

telic transitives, thereby keeping the refugees as part of the scene.

We have seen in the above texts that the task of introducing participants and managing

perspective takes place in narrative via a number of devices. Among those devises are the telic

and atelic transitive. Atelic transitives also serve a second function. DuBois states that one-

place predicates are also used when they are semantically appropriate. Even so, atelics serve a

Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E1KE1BNW9_R8200Q INGEST_TIME 2013-09-28T01:02:44Z PACKAGE AA00014289_00001