Peter C. Gordon

Pronouns, Names, and the Centering of Attention in Discourse

Centering theory, developed within computational linguistics, provides an account of ways in which patterns of interutterance reference can promote the local coherence of discourse. It states that each utterance in a coherent discourse segment contains a single semantic entity—the backward-looking center—that provides a link to the previous utterance, and an ordered set of entities—the forward-looking centers—that offer potential links to the next utterance. We report five reading-time experiments that test predictions of this theory with respect to the conditions under which it is preferable to realize (refer to) an entity using a pronoun rather than a repeated definite description or name. The experiments show that there is a single backward-looking center that is preferentially realized as a pronoun, and that the backward-looking center is typically realized as the grammatical subject of the utterance. They also provide evidence that there is a set of forward-looking centers that is ranked in terms of prominence, and that a key factor in determining prominence—surface-initial position—does not affect determination of the backward-looking center. This provides evidence for the dissociation of the coherence processes of looking backward and looking forward.

Memory interference during language processing.

The authors studied the operation of working memory in language comprehension by examining the reading of complex sentences. Reading time and comprehension accuracy in self-paced reading by college students were studied as a function of type of embedded clause (object-extracted vs. subject-extracted) and the types of noun phrases (NPs) in the stimulus sentences, including relative clauses and clefts. The poorer language comprehension performance typically observed for object-extracted compared with subject-extracted forms was found to depend strongly on the mixture of types of NPs (descriptions, indexical pronouns, and names) in a sentence. Having two NPs of the same type led to a larger performance difference than having two NPs of a different type. The findings support a conception of working memory in which similarity-based interference plays an important role in sentence complexity effects.

Memory-Load Interference in Syntactic Processing

Participants remembered a short set of words while reading syntactically complex sentences (object-extracted clefts) and syntactically simpler sentences (subject-extracted clefts) in a memory-load study. The study also manipulated whether the words in the set and the words in the sentence were of matched or unmatched types (common nouns vs. proper names). Performance in sentence comprehension was worse for complex sentences than for simpler sentences, and this effect was greater when the type of words in the memory load matched the type of words in the sentence. These results indicate that syntactic processing is not modular, instead suggesting that it relies on working memory resources that are used for other nonsyntactic processes. Further, the results indicate that similarity-based interference is an important constraint on information processing that can be overcome to some degree during language comprehension by using the coherence of language to construct integrated representations of meaning.

LEXICAL AND PRELEXICAL INFLUENCES ON WORD SEGMENTATION: EVIDENCE FROM PRIMING

The authors examined the interaction of acoustic and lexical information in lexical access and segmentation. The cross-modal lexical priming technique was used to determine which word meanings listeners access at the offsets of oronyms (e.g., tulips or two lips) presented in connected speech. In Experiment 1, participants showed priming by the meaning of tulips when presented with two lips. In Experiment 2, priming by the meaning of the 2nd word was found in such sequences (e.g., lips in two lips). Finally, Experiment 3 demonstrated that listeners do not show priming by lips when it is pronounced as part of tulips. The results of these experiments show that listeners sometimes access words other than those intended by speakers and may simultaneously access words associated with several parses of ambiguous sequences. Furthermore, the results suggest that acoustic marking of word onsets places constraints on the success of lexical access. To account for these results, the authors propose a new model of lexical access and segmentation.

The representation and processing of coreference in discourse

A model is presented that addresses both the distribution and comprehension of different forms of referring expressions in language. This model is expressed in a formalism (Kamp & Reyle, 1993) that uses interpretive rules to map syntactic representations onto representations of discourse. Basic interpretive rules are developed for names, pronouns, definite descriptions, and quantified descriptions. These rules are triggered by syntactic input and interact dynamically with representations of discourse to establish reference and coreference. This interaction determines the ease with which coreference can be established for different linguistic forms given the existing discourse context. The performance of the model approximates that observed in studies of intuitive judgments of grammaticality and studies using online measures of language comprehension. The model uses the same basic interpretive mechanisms for coreference within and between sentences, thereby linking the domain traditionally studied by generative linguists to domains that have been of concern primarily to psychologists and computational linguists.

Similarity-Based Interference During Language Comprehension: Evidence from Eye Tracking During Reading

The nature of working memory operation during complex sentence comprehension was studied by means of eye-tracking methodology. Readers had difficulty when the syntax of a sentence required them to hold 2 similar noun phrases (NPs) in working memory before syntactically and semantically integrating either of the NPs with a verb. In sentence structures that placed these NPs at the same linear distances from one another but allowed integration with a verb for 1 of the NPs, the comprehension difficulty was not seen. These results are interpreted as indicating that similarity-based interference occurs online during the comprehension of complex sentences and that the degree of memory accessibility conventionally associated with different types of NPs does not have a strong effect on sentence processing.

Pronominalization and discourse coherence, discourse structure and pronoun interpretation

Two self-paced reading-time experiments are reported that examine the time course of pronoun interpretation processes based on local discourse structure and on world knowledge. The characterization of local discourse structure is based on recent work on centering, which provides a specific formulation of how the ways in which sentences make reference to common entities determines the coherence of discourse segments and how discourse structure influences interpretation of ambiguous pronouns. The results of the first experiment show that readers generate a default interpretation of a pronoun based on features of local discourse structure, and that that default interpretation is later confirmed or overridden by knowledge-based processes. The results of the second experiment show that local discourse structure continues to influence pronoun interpretation even when the semantic information that ultimately compels interpretation occurs before the pronoun. These results support the view that processes acting on local discourse structure play a powerful role in guiding language comprehension.

Speech production: Motor programming of phonetic features

Abstract Three experiments are reported on the role of phonetic features in motor programs for speech production. Each experiment involved a type of response-priming procedure. The procedure required subjects to prepare a specified primary vocal response that contained one or two vowel-consonant syllables (e.g., “up,” “ub,” “ut,” and “ud”). After the preparation interval, the subjects either produced the primary response upon command or else switched to produce another specified secondary response instead. Response latency and accuracy were measured as a function of the relationship between the phonetic features of the primary and secondary responses. Longer latencies and more errors occurred when the secondary response had place-of-articulation or voicing features identical to those of the primary response. The results may be interpreted in terms of an interactive-activation model. It appears that phonetic features play a significant role during the compilation of articulatory motor programs, and that preparation to produce an utterance inhibits the programming of other utterances with similar features. This outcome complements and extends conclusions derived from analyzing naturalistic slips of the tongue and coarticulation phenomena. The interactive-activation model of speech production provides a link with theoretical accounts of speech perception, word recognition, and manual movement.

Electrophysiological Evidence for Reversed Lexical Repetition Effects in Language Processing

Effects of word repetition are extremely robust, but can these effects be modulated by discourse context? We examined this in an ERP experiment that tested coreferential processing (when two expressions refer to the same person) with repeated names. ERPs were measured to repeated names and pronoun controls in two conditions: (1) In the prominent condition the repeated name or pronoun coreferred with the subject of the preceding sentence and was therefore prominent in the preceding discourse (e.g., John went to the store after John/he ); (2) in the nonprominent condition the repeated name or pronoun coreferred with a name that was embedded in a conjoined noun phrase, and was therefore nonprominent (e.g., John and Mary went to the store after John/he ). Relative to the prominent condition, the nonprominent condition always contained two extra words (e.g., and Mary), and the repetition lag was therefore smaller in the prominent condition. Typically, effects of repetition are larger with smaller lags. Nevertheless, the amplitude of the N400 was reduced to a coreferentially repeated name when the antecedent was nonprominent as compared to when it was prominent. No such difference was observed for the pronoun controls. Because the N400 effect reflects difficulties in lexical integration, this shows that the difficulty of achieving coreference with a name increased with the prominence of the referent. This finding is the reverse of repetition lag effects on N400 previously found with word lists, and shows that language context can override general memory mechanisms.

Processing of Reference and the Structure of Language: An Analysis of Complex Noun Phrases

Five experiments used self-paced reading time to examine the ways in which complex noun phrases (both conjoined NPs and possessive NPs) influence the interpretation of referentially dependent expressions. The experimental conditions contrasted the reading of repeated names and pronouns referring to components of a complex NP and to the entire complex NP. The results indicate that the entity introduced by a major constituent of a sentence is more accessible as a referent than the entities introduced by component noun phrases. This pattern of accessibility departs from the advantage of first mention that has been demonstrated using probe-word recognition tasks. It supports the idea that reduced expressions are interpreted as referring directly to prominent entities in a mental model whereas reference by names to entities that are already represented in a mental model is mediated by additional processes. The same interpretive processes appear to operate on coreference within and between sentences.