Use this URL to cite or link to this record in EThOS:
Title: The time course of the influence of implicit causality information on resolving anaphors
Author: Stewart, Andrew James
Awarding Body: University of Glasgow
Current Institution: University of Glasgow
Date of Award: 1998
Availability of Full Text:
Access from EThOS:
Access from Institution:
This thesis asks whether initial anaphor processing proceeds in a restricted manner with reference only to a well defined set of information or whether it is the case that all factors that are potentially relevant for resolving an anaphor exert a processing influence at the same time. In an attempt to adjudicate between these possibilities, we focus on the nature of the processing influence of implicit causality information on anaphor resolution. Following a summary in Chapter 1 of issues concerning possible cognitive architectures and a review in Chapter 2 of previous work on anaphor resolution, we propose a two-stage model of anaphor resolution. We propose that the first stage involves co-indexation between anaphor and antecedent and is informed by low-level factors. We claim this stage behaves in a modular or restricted manner. The second stage involves integrative processing and behaves in a nonmodular or unrestricted manner. We suggest that it is at this second stage of processing that implicit causality influences anaphoric processing. Implicit causality (Garvey and Caramazza, 1974) is a property associated with a particular set of verbs which, in sentence fragments such as (1) and (2) below, influences interpretation of the ambiguous pronoun. (1) John fascinated Bill because he ... (2) John blamed Bill because he ... The verb 'fascinate' is classified as an NP1 biasing verb as it biases towards the character occupying the first Noun Phrase as the locus of cause. Similarly, the verb 'blame' is an NP2 biasing verb as it biases toward the character occupying the second Noun Phrase as the locus of cause. Readers prefer to interpret the pronoun as coreferential with the character predicted by the verb. Previous work has demonstrated the influence of implicit causality in both language production (e.g. Garvey and Caramazza, 1974) and comprehension (e.g. Caramazza, Grober, Garvey and Yates, 1977). A reading penalty arises when the information in the subordinate clause conflicts with the verb bias, i.e. when there is a mismatch between implicit and explicit causes, as in example (3). (3) John blamed Bill because he hated Bill. However, several major methodological criticisms can be raised against previous work examining the influence on comprehension of implicit causality. Variations in factors such as sentence length, sentence plausibility and non-homogeneity of strength of verb biases may have confounded previous research. Experiments la and lb in this thesis were used to create materials controlled for plausibility and strength of bias. Average length of the experimental sentences was equated across conditions. From an initial set of 50 verbs examined in Experiments la and lb, we selected 24 that were strongly biasing and of equivalent plausibility for each cause. An initial self-paced reading experiment (Experiment 2) demonstrated an implicit causality congruency effect with our materials on whole sentence reading times. Experiments (3) and (4) involved presenting the experimental materials in two halves, with the split occurring following the anaphor (see (4) and (5) below). We added an additional between experiment factor of question type which encouraged either deep or shallow processing. (4) John fascinated Bill because he/John was full of interesting stories. (5) John fascinated Bill because he/Bill was easily entertained. Each sentence was presented as two fragments with the split following the anaphor. If implicit causality exerts an early influence on processing we would expect to find evidence of the congruency effect on reading times to the first fragment. The only effect we found on reading time to the first fragment was a repeat name penalty resulting from repetition of the first mentioned character's name. This did not interact with verb bias suggesting implicit causality does not influence interpretation of the anaphor when it is first encountered. Reading times to fragment 2 showed an effect of implicit causality. Our between experiment manipulation led to a reduction in the strength of the implicit causality congruency effect under circumstances where shallow processing was encouraged. In other words, the relative difficulty associated with reading sentence continuations going against the bias of the verb was reduced when readers were engaged in shallower reading. The depth of processing manipulation had no effect on the magnitude of the repeat name penalty. This suggests that these phenomena may be arising from processing at different stages within the system.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: Language processing; Parallel processing