Information Structure in Lexical-Functional Grammar: The Discourse-Prosody Correspondence in English and Serbo-Croatian

Rob O'Connor

Abstract

Proceedings of LFG05; CSLI Publications On-line

This thesis is primarily a study of the relationship between intonation and information structure from the point of view of LFG. On one hand, I portray intonation as an aspect of p(rosodic)-structure, a rarely investigated side of LFG. On the other hand, I adopt the notion, expressed by Lambrecht (1994), of information structure as the pairing, in a given discourse context, of a proposition with a 'lexicogrammatical structure'. In LFG terms, however, describing information structure as a 'pairing' suggests a correspondence between dimensions, rather than the separate dimension that is usually implied in LFG work. I adapt Lambrecht's characterisation and propose that the term 'information structure' covers the mapping between certain dimensions of the grammar. In particular, it covers the mappings between d(iscourse)-structure, representing what Lambrecht refers to as the 'pragmatic structuring' of a proposition, and those dimensions which are equivalent to 'lexicogrammatical structures' such as c(onstituent)-structure and p(rosodic)-structure. This is represented schematically in (1).

(1)	
			        /   p-structure
                               /
   d-structure --- information ---- c-structure
                   structure   \
                                \   lexicon/morphology

In theory, p-structure represents the whole range of prosodic phenomena. In terms of the prosodic expression of information structure, however, only intonational phenomena are relevant. I propose a version of p-structure that organises tonal events, i.e. pitch accents and edge tones, into an Autosegmental-Metrical style prosodic tree representation of an intonational contour, or tune. The p-structure for example (2) is given in (3).

(2)   L+H*  !H*          L-           H*H*    L+H*   L-             L+H*  L-L%
       |     |           |            |  |     |     |               |    | |
   [[Capote died Saturday]ip [at the Bellaire home of]ip [Joanne Carson]ip]IP

                                                     (Beckman & Ayers 1997: 27)
(3)                                      TUNEIP
                                 /         |                    \         \
                             /             |                          \     \
               tuneip                     tuneip                     tuneip  t%  
          /   /       \               /  /    \    \                 /   \   |
       t*   t*           t-          t* t*     t*    t-              t*   t- |    
       |     |           |            | |      |     |               |    |  |
      L+H*  !H*          L-           H*H*    L+H*   L-             L+H*  L- L%
       |     |           |            |  |     |     |               |    |  |
   [[Capote died Saturday]ip [at the Bellaire home of]ip [Joanne Carson]ip]  IP

A 'TUNEIP', or 'intonational phrase (= IP) tune' consists of at least one 'intermediate phrase (= ip) tune', or tuneip, plus an IP boundary tone (t%). In turn, tuneip consists of at least one pitch accent (t*) plus a phrase accent (t-). P-structure can also be expressed in terms of 'tune structure rules' as in (4).

(4) n >= 1
    (a)  TUNEIP  -->   tuneipn  t%
    (b)  tuneip  -->   t*n      t-

Turning to d-structure, I represent this in AVM format as shown in (6) for the answer in example (5).
(5) Q: What happened to your car?
A: It broke DOWN.
(Lambrecht 1994: 223)

(6) [ FOCUS   {'break down'}
      ACTVN-F +
      TOPIC   {'it'}
      ACTVN-T -
      OPEN    -               ]

In (6), in addition to the discourse functions, TOPIC and FOCUS, I have introduced two discourse features, +/-OPEN and +/-ACTVN [= 'activation']. The first, +/-OPEN, relates to the notion that clauses and sentences express, among others things, the status of propositions as 'complete' (e.g. declaratives) or 'incomplete' (e.g. interrogatives). The second feature, +/-ACTVN, relates to the status of referents of linguistic expressions as being either 'active' or 'non-active', a notion used by Lambrecht (1994), following Chafe (1987), in place of the 'old' versus 'new' distinction between referents. This 'activation' status of referents is independent of the distinction between 'old' and 'new' information. In presenting my thesis I will concentrate on the formalisation of the interaction between d-structure and p-structure in terms of a mapping function by which information about one dimension, in this case d-structure, is annotated on to specific parts of another dimension, in this case p-structure. I will also address how this mapping interacts with the mapping between d-structure and c-structure.