The main goal of this presentation is to outline a new approach to the interface between two systems: "the narrow" syntax and discourse, and to show how this approach can account for a number of observations in child speech and in the speech of agrammatic Broca's aphasics.
The relevance of some extra-syntactic mechanisms for apparently "purely" syntactic phenomena is well known but has been rarely discussed in terms of the interface between syntax and discourse. Some classical examples include Pesetsky's (1987) observation about the differences in acceptability of D-linked and non-D-linked Wh-questions, as in (1).
(1) a. *Mary asked what who read vs. b. Mary asked what which man read
Continuing along this line, Rizzi (1990) and Cinque (1991) show that only D-linked Wh-questions can be extracted from weak islands:
(2) ??A chi ti chiedi quanti soldi hai dato?
To whom do you wonder how much money you gave
A quale dei tuoi figli ti chiedi quanti soldi hai dato?
To which one of your children do you wonder how much money you gave
The relevance of discourse factors can be also observed in special registers that allow non-finite main clauses, as in (3), (4) and (5).
(3) Korol' rasskazal anekdot. Princessa xoxotat'! [Russian]
King has-told a joke. Princess to laugh!
(4) John dance??? Never!!!
(5) McDonalds to serve beer.
There are other registers in normal adult speech that allow omissions of otherwise required material, e.g. the diary style (see Haegeman 1990) and determiner omission:
(6) Got up at 7. Took shower. Had breakfast. Left.
(7) Leuk huisje heb je! [Dutch]
nice house you have!
Importantly, strong contextual background is required for the acceptability of such utterances (e.g. the diary style is possible only under assumption that it is known that it is a diary!).
In the domain of definite DPs, it has been observed that there are cases of apparent Principle B violation (for discussion see Reinhart 1983, Heim 1998, among others), as in (8) and "the first time use" of definite descriptions as in (9).
(8) Who is this speaker? Must be Zelda. She praises her to the sky.
(9) I've read an interesting book recently. The author was well known to me.
Once again, in both cases it is the availability of certain discourse conditions that makes such cases acceptable.
Finally, the concept of "discourse-new" information (Prince 1981) has been shown to play a crucial role in the distribution of associates in expletive constructions (Birner and Ward 1998):
(10) a. There were the same people at both conferences
b. There was the usual crowd at the beach.
In the field of psycholinguistics, a significant amount of research has shown the relevance of discourse - related operations for real - time processing (e.g. Frasier (submitted), Pinango et al 2001), for child language (e.g. Thornton 1995, Avrutin 1994, 1999, among others) as well as for agrammatic production and comprehension (for review see Avrutin 2000). Additional evidence is presented in Zuckerman et al and Baauw et al (BU conference 2001).
To summarize, there appears to be a sufficient body of data that warrants the necessity to look more carefully at the interaction of "purely" syntactic operations (operations within the domain of "narrow syntax") and the operations applying at the level where the referential properties of linguistic expressions are represented (the assumption being that the concept of "referential property" is extra-syntactic as the narrow syntax operates on formal syntactic features which do not encode referential capacity).
I will proceed in the following way.
1. We begin with the discussion of what it means for an element to be "referential". This notion, while widely used in linguistic theory (as, say, in the case of D-linked phrases, or +/- R expressions in Reflexivity) has not been clearly formulated. I will therefore introduce the notion of "referential potential" that is related to the ability of a linguistic element to introduce - or be represented by - a corresponding discourse entity.
2. Once we are dealing with discourse entities, it is necessary to have some model of what such entities are, how they interact with each other, and what are possible constraints on operations that apply to them. As a basis, I use Heim's (1982) File card model that has "file cards" (basically, information units) as primitives. The existing model will be extended to include both Individual file cards and Event file cards. The basic idea is to capture the intuition that discourse contains both individuals and events as entities and both of them can be referred to.
3. The next step - which will be crucial for a new view of the discourse anaphora and a reformulation of the binding conditions - is to show that discourse (or, rather, information) units have structure. It will be shown that both Event and Individual file cards should be analyzed as containing parts relevant for information packaging (Frame, Heading, (optional) Property). Functional categories D and T introduce, correspondingly, Individual and Event Frames, lexical categories N and V introduce Headings.
4. Now that we have established a connection between syntactic symbols (N, V, D, T) on the one hand and corresponding information symbols (file cards and their parts) on the other, we can take a closer look at how entities interact with each other at the level of "discourse" (information) structure. Viewing discourse anaphora as a mechanism of transferring information from one file card to another, we can characterize constraints on information transfer. Depending on their information status, which in turn is related to the morphosyntactic composition of corresponding syntactic elements, information units (file cards) can be Weak or Strong. Informationally Weak cards may not exist as independent information units, and are collapsed with previously introduced entities (the case of Dutch 'zich', Icelandic 'sig' and other SE type pronominals). Strong cards occupy an independent place in the informatin structure, but may require receivng information from other entities (this is the essence of discourse anaphora, as for pronouns 'he', 'she', etc.). Reflexive 'himself', 'zichzelf' and other elements containing SELF morpheme will be shown to have a very special status of "standing for" their antecedents (as in Reuland's 2001 Primitives of Binding framework.)
5. Such constraints will be shown to capture generalizations that used to be explained by the binding theory. In other words, in line with some current views in linguistic theory, rules responsible for the distribution of pronominals will be shown to be part of the domain outside the narrow syntax. These rules, naturally, still belong to the language-specific computational system, but to that part of the system that operates on non-syntactic symbols.
6. I will then discuss some differences between production and comprehension of structures containing pronominals, adopting Levelt's model of speech production. The main question for the production that I address is how we choose the appropriate pronominal lemma.
7. The developed model will now be applied to a variety of psycholinguistic
data, specifically to the omission of functional categories (D and T) in child
and aphasic speech and to some errors in comprehension sentences with pronouns
(also by children and agrammatic aphasics). I will propose that many errors
in the speech (and comprehension) of these populations are due to limited resources
(not a grammatical deficit), which results in relying on alternative means of
conveying information. For example, the omission of Determiners will be viewed
as introduction an Individual Frame by alternative means, specifically through
discourse presupposition. Similar claims can be made about omission of Tense,
which is responsible for the introduction of an Event Frame. I will also discuss
recent results of a study with Dutch speaking children that invesigated their
comprehension of pronouns in ECM constructions and with contrastive stress.
If you are interested in meeting with the speaker, please contact Scott Fults or Nina Kazanina
To go back to the Colloquium schedule, click
here.