part3- 19. Pragmatics and Argument ...

part3- 19. Pragmatics and Argument Structure, Materiały naukowe, The Hanbook of Pragmatics

[ Pobierz całość w formacie PDF ]
19. Pragmatics and Argument Structure : The Handbook of Pragmatics : Blackwell Re...
Page 1 of 11
19. Pragmatics and Argument Structure
ADELE E. GOLDBERG
Subject
Theoretical Linguistics
Ç
Pragmatics
Key
-Topics
Topics
structure
DOI:
10.1111/b.9780631225485.2005.00021.x
1 What is Argument
1 What is Argument Structure?
Structure?
A
RGUMENT
STRUCTURE
has been used to refer to various things in the literature. In the logical tradition,
argument structure refers to the number and type of arguments that are associated with a predicate
(e.g. a verb). The argument structure of give is a three-place predicate, requiring an agent, a theme,
and a recipient argument. On this view, then, one and the same argument structure is expressed by
the ditransitive and dative patterns, as in (1) and (2):
(1) She gave him an apple. Ditransitive
(2) She gave an apple to him. Dative
In recent syntactic theories, on the other hand, argument structure is often taken to refer to a level of
purely formal abstraction, devoid of any semantics. On this view, (1) and (2) may be understood to
represent two different argument structures, or only one if the first is assumed to be syntactically
derived from the second. My use of the term in what follows is a hybrid of these approaches, in which
the argument structure of a clause is defined as the surface syntactic form together with the overall
event-interpretation of a clause. The examples in (1) and (2), therefore, illustrate two different
argument structures insofar as they differ in form. As described below, they differ in their semantics
as well. Examples of argument structure patterns include the transitive, the ditransitive, the
resultative, the sentential clause complement construction, etc.
Most verbs readily appear in more than one argument structure pattern. A question that has been
gaining attention, and that we focus on here, is: What determines which argument structure pattern
will actually be used? A related question is: Why do languages provide alternative ways to express
similar meanings? A great deal of work has noted semantic differences between rough paraphrases
(e.g. Partee 1965, Fillmore 1968, Anderson 1971b, Borkin 1974, Levin 1993, Goldberg 1995). For
example, the ditransitive or double object construction requires that its goal be animate, whereas the
dative construction does not:
Passages (1a-b) provide evidence for such a coherence constraint.
(3) a. Chris sent them a package. Ditransitive
É b. Chris sent that place a package. É
28.12.2007
19. Pragmatics and Argument Structure : The Handbook of Pragmatics : Blackwell Re...
Page 2 of 11
(4) Chris sent a package to them/to that place. Dative
Slight differences in meaning such as this are clearly one factor that distinguishes between alternate
argument structure patterns. They allow speakers to choose which pattern to use on the basis of
differing semantics, thereby offering speakers of a language more expressive power. Less studied,
however, is the role of pragmatics in differentiating among argument structure possibilities.
2 What is
2 What is Pragmatics?
Pragmatics?
For present purposes we distinguish two types of pragmatics - non-conventional and conventional.
N
ON
-
CONVENTIONAL
PRAGMATICS
involves the effects of the comprehension or production of sentences
in particular contexts of use by actual language users having the type of processing and cognitive
abilities and preferences that humans do. These effects are expected to be universal, given that
languages are products of human beings. C
ONVERSATIONAL
PRAGMATICS
(Grice 1967, Horn 1984a) is
perhaps the best known example of non-conventional pragmatics, and the one that is focused on
here.
C
ONVENTIONAL
PRAGMATICS
is the conventional association of certain formal properties of language with
certain constraints on pragmatic contexts. Effects of conventional pragmatics are non-necessary
effects, and so we would expect to find some degree of language variation, at least in degree of
conventionalization.
1
As pertains to clause structure, conventional pragmatics largely corresponds to
ways in which languages choose to package INFORMATION STRUCTURE (Halliday 1967, Chafe 1976,
Lambrecht 1994, Ward and Birner this volume).
Two notions that play a central role in the packaging of information structure are TOPIC and FOCUS,
which we can define as follows (see Gundel and Fretheim (this volume) for a more in-depth
discussion): a sentence topic can be defined as a Ñmatter of [already established] current interest
which a statement is about and with respect to which a proposition is to be interpreted as
relevantÒ (Lambrecht 1994: 119). On focus, we follow Halliday (1967: 204), who writes: ÑInformation
focus is one kind of emphasis, that whereby the speaker marks out a part (which may be the whole) of
a message block as that which he wishes to be interpreted as informative.Ã’ Similarly Lambrecht (1994:
218) defines the focus relation as relating Ñthe pragmatically non-recoverable to the recoverable
component of a proposition [thereby creating] a new state of information in the mind of the
addressee.Ã’
Other notions that are often used in discussions of information structure describe whether particular
arguments within a discourse have been previously mentioned: that is, whether the arguments are
DISCOURSE-OLD (GIVEN) or DISCOURSE-NEW (Prince 1992). The correlations between focus and topic
on the one hand, and discourse-old/given or -new on the other are complicated, but some rough
generalizations can be made. Continuing topics are given in that they have to have been mentioned in
order to be continuing as topics (e.g. she in (5a) below); even newly established topics tend to be
accessible or anchored in the discourse as opposed to brand new, insofar as they appear with a
definite determiner or are explicitly related to a discourse-old entity by means of a possessive
determiner or relative clause (e.g. her mother in (5b); see Francis et al. 1999). Focal arguments are
often discourse-new (a snake in (5d)); discourse-old elements can serve as foci only if they are
accented (her in (5c)).
(5)
ÉÉÉÉ
Discourse
-old (given)
old (given)
Discourse
-new
new
topic
topic (a)
(a)
(a) She hit a
hit a
hit a pole
pole
pole.
(b) Her mother feared snakes.
focus
focus(c) George said they called HER.(d) She saw
(d) She saw
A SNAKE
SNAKE
SNAKE.
28.12.2007
19. Pragmatics and Argument Structure : The Handbook of Pragmatics : Blackwell Re...
Page 3 of 11
Since all of the cells of the matrix are instantiated, it is clear that the notions of topic and focus
cannot be reduced to the notions of discourse-old and -new, as is sometimes assumed. Still, it is
clear that most commonly, topics are discourse-old (e.g. (5a)), because most topics are continuing
topics. Focal elements tend to be discourse-new, since asserting information most commonly occurs
via the mention of a new entity (e.g. (5d)). These common correlations are important to keep in mind
when we try to relate proposals, such as many of the ones discussed below, that are couched
exclusively either in terms of topic vs. focus or in terms of given/old vs. new.
3 Information Structure and Argument Structure
A simple transitive argument structure pattern can appear in a cleft construction, a left-dislocation
construction, or a topicalization construction:
(6) It was a giraffe that the mouse saw. It-cleft
(7) The giraffe, the mouse saw it.
Left-dislocation
(8) The giraffe, the mouse saw.
Topicalization
It is generally recognized that sentence-level constructions such as those represented in (6)-(8) are
associated with their own information structure properties. For example, in an extensive analysis of
the Switchboard corpus of spoken language, Gregory and Michaelis (to appear) document the
functions of the left-dislocation and topicalization constructions, finding subtle distinctions between
them. The fronted NPs in the left-dislocation construction are not previously mentioned and yet do
persist as topics. The fronted NPs in the topicalization construction display the opposite tendency: the
majority are previously mentioned and do not persist as topics. Thus, the left-dislocation construction
is topic establishing, whereas the topicalization construction tends to be used for moribund topics.
It is not immediately obvious that argument structure, which has to do with the semantic relation
between a verb and its arguments, should have any direct relationship to conventional pragmatics. As
Lambrecht (1994: 159) observes, Ñthe independence of semantic and pragmatic roles is an obvious
consequence of the fact that information structure has to do with the use of sentences, rather than
the meaning of propositions.Ã’ Nonetheless, on the view that different syntactic complement arrays
reflect different argument structures, we will see below that argument structure patterns are indeed
associated with information structure generalizations.
4 Preferred
4 Preferred Argument Structure
Argument Structure
Du Bois (1987) proposed a Preferred Argument Structure for the way argument structures are actually
used in discourse. Assuming Dixon's (1972) system for classifying core arguments, intransitive
clauses have only one core argument, the subject or S, and transitive clauses have two core
arguments: the actor or A and the object or O. In English sentences like The vase broke or The boy
ran, Ñthe vaseÒ and Ñthe boyÒ are S's; in a sentence such as The giraffe spotted the owl, Ñthe giraffeÒ is
an A and Ñthe owlÒ is an O.
Du Bois (1987) analyzed the distribution of lexical A, S, and O in elicited, ongoing discourse in the
ergative language of Sacapultec Maya. The corpus study revealed that only 2.8 percent of transitive
clauses involved two lexical NPs. Moreover, only 3.2 percent of A's represented discourse-new
entities, expressed by lexical NPs. On the other hand, 22.5 percent of S's and 24.7 percent of O's
represented such discourse-new entities. Du Bois posits two constraints: (1) a Q
UANTITY
GENERALIZATION: Ñavoid more than one new argument per clauseÒ (Du Bois 1987: 819; see also Dixon
1972, Givn 1975, Chafe 1987); and (2) the G
IVEN
A GENERALIZATION: Ñavoid new A's.Ò These two
constraints are jointly taken to define the Preferred Argument Structure cross-linguistically.
These findings have been replicated again and again in many unrelated languages, including English
(Iwasaki 1985), German (Schuetze-Coburn 1987), French (Lambrecht 1987), Hebrew (Smith 1996),
Mam (England 1983), Malay (Hopper 1988), Quechua (Payne 1987), child Inuktitut speech (Allen and
28.12.2007
19. Pragmatics and Argument Structure : The Handbook of Pragmatics : Blackwell Re...
Page 4 of 11
Schroder to appear), Papago (Payne 1987), and Tzeltal (Brown to appear). S and O both easily
accommodate discourse-new elements. The A slot is distinct in that it strongly prefers old or given
elements.
2
This split between S and O on one hand and A on the other is what generally defines
ergativity, whether it is morphologically marked or syntactically expressed. It is the discourse
properties, Du Bois (1987) argues, that form the basis for the categorization that results in all types of
ergativity.
It is worth asking whether the Quantity and Given A generalizations are both independently required.
One question that arises is whether the Given A generalization is ultimately just an effect of a
correlation between animates and topicality. That is, the A argument of transitive clauses strongly
favors animate entities, and animates are good candidates for topic status simply because human
beings like to talk about other human beings (Osgood 1980). A's are likely to be topical, and ongoing
topics are necessarily given: therefore, A's are likely to be given - thus, the Given A generalization.
There is, however, a consideration that mediates against this idea that the Given A generalization is
simply epiphenomenal. Languages strongly favor introducing new animate entities via an intransitive
clause whenever new animate entities are introduced. Du Bois (1987: 831) suggests that speakers opt
for Ñintransitive introduction followed by transitive narration.Ò That is, humans may be likely to make
animate entities topics, but that is not sufficient to explain why the A slot is avoided when animates
are not topical. Thus it seems that the Given A constraint does not follow directly from the prevalence
of animate topics.
A second question that arises is whether the two constraints could possibly be conflated into one.
Other than A, there is only one other argument available in the nuclear clause (either S or O), so the
Quantity generalization (avoid more than one new argument per clause) would seem to follow from
the Given A generalization. But the Quantity generalization may help to motivate why it should be that
new animate entities are often introduced via an intransitive rather than transitive clause, which again,
is the one aspect of the Given A constraint that does not follow from the discourse frequency of
animate topics. Of course the Quantity constraint could be satisfied in one of two ways in discourse
contexts in which a new animate participant is introduced: either an intransitive clause could be used
or a transitive clause with a given O could be used. In fact, as discussed below, the admittedly rare
transitive expressions with non-given A's do tend to have given O's. Thus, there is evidence that the
Quantity constraint is not a consequence of the Given A constraint. Moreover, a possibly related type
of Quantity generalization seems to be operative in accounting for object omission with normally
transitive verbs, as is discussed below.
Can the Given A constraint be derived completely from the Quantity generalization together with the
general tendency for animates to be ongoing topics and therefore given? The fact just noted, that
languages apparently prefer the intransitive mention of new animates rather than the transitive
mention with given O's, even though the Quantity constraint is satisfied equally well in either way,
provides one piece of evidence that the Given A constraint is not simply epiphenomenal. Moreover,
languages differ in the degree to which the Given A generalization holds (Van Valin and LaPolla 1997,
Lambrecht 2001). For example, the constraint is near absolute in spoken French insofar as subject
arguments cannot be focal and non-focal subjects tend to be given (Lambrecht 1995), whereas
English allows new, focal A's in certain circumstances (discussed below). Thus it seems that the
Quantity and Given A generalizations are both required to describe the data.
The Given A and Quantity generalizations seem to be accurate cross-linguistically, and their existence
is likely motivated, if not predicted, by processing and discourse factors. The Quantity generalization
may be based on some kind of ease of processing generalization, although the specific explanation
has not yet been identified. As noted above, the Given A generalization goes beyond the
conversational tendency to make humans topical; still, the motivation for the generalization
undoubtedly lies in this tendency. In fact, the tendency for humans to make other animate beings
topics results in a tendency for both A and S to be topical more often than O in many languages.
Thus, sentences in which the logical subject represents the topic and the predicate represents the
comment or assertion about that topic represent the most frequent pattern and can, therefore, be
considered the canonical or unmarked construction type (Kuno 1972, Horn 1986, Chafe 1994,
Lambrecht 1994).
28.12.2007
19. Pragmatics and Argument Structure : The Handbook of Pragmatics : Blackwell Re...
Page 5 of 11
5 Sentence Focus
5 Sentence Focus Constructions
Constructions
Languages typically have special constructions that allow for non-canonical packaging of information.
Lambrecht (1994) defines SENTENCE FOCUS (SF) CONSTRUCTIONS as constructions that are formally
marked as expressing a pragmatically structured proposition in which both the subject and the
predicate are in focus. As Lambrecht (1995) notes, the function of SF constructions is presentational -
namely, to present an entity or an event into the discourse (cf. also Sasse's (1987) entity-central vs.
event-central thetic sentences). An English SF construction that introduces an event into the discourse
is characterized by having pitch accent only on the logical subject, and not on the predicate phrase,
as in (9).
(9) What happened?
a. Her SON is sick.
b. Her BIKE broke down.
c. My SHOULDER hurts.
d. ZACH called.
e. Her HUSBAND left her.
Lambrecht (1994) observes as well that the subject in this construction is not topical and cannot be
pronominal. For example, (10) can only be interpreted with a narrow focus on the subject argument
(an ARGUMENT FOCUS reading) and does not permit a sentence focus interpretation:
(10) HE is sick. (possible context: A: Is she sick? B: No, HE is sick)
The predicate in the SF construction typically has semantics that are compatible with presentation,
with SF constructions cross-linguistically favoring certain unaccusative verbs such as arrive, come,
die, and disappear. SF expressions are rarely transitive, consistent with the Given A generalization,
since the focal intransitive subject is an S and not an A. When SF expressions are transitive (e.g. (9e)),
the object nominal strongly tends to be pronominal (Lambrecht 1995), in accord with the Quantity
generalization.
In sections 4 and 5, we have seen that information structure properties motivate the existence of a
dominant argument structure type cross-linguistically, the specific properties of which are
conventionalized differently and to different extents in different languages. The need for a full range
of expressive power motivates the existence of marked construction types such as the SF
construction.
The ditransitive construction, as seen in (1) above, can be used to illustrate the potentially far-
reaching role of information structure in the grammar of argument structure.
6 Information
6 Information Structure and the Ditransitive
Structure and the Ditransitive
In both corpus and experimental studies, Arnold et al. (2000) found that both newness and heaviness
play a role in determining the choice of the ditransitive over the dative construction, where heaviness
is determined by number of words, and newness by lack of previous mention in the discourse
3
(see
also Givn 1979, 1984, Dryer 1986, Thompson 1990). For present purposes, we will interpret these
generalizations as implying that the recipient argument must be topical, not focal:
(11) Subj V Obj1(topical) Obj2
She kicked him the ball
The idea that the ditransitive constrains the recipient argument to be non-focal may ultimately help
account for certain interesting facts about how the ditransitive construction interacts with long-
distance dependency constructions and the passive construction in English. In particular, notice that
the recipient argument of the ditransitive cannot readily appear in a long-distance dependency
relation:
28.12.2007
[ Pobierz całość w formacie PDF ]

zanotowane.pl

doc.pisz.pl

pdf.pisz.pl

anio102.xlx.pl