Frontiers in Psychology (Apr 2014)
Cataphora processing in agrammatic aphasia: Eye movement evidence for integration deficits
Abstract
Background Individuals with agrammatic aphasia show selective comprehension impairments in complex sentences with long-distance dependencies such as Wh-filler-gap dependencies (WhFG). Recent studies suggest that this difficulty arises from failure to integrate grammatical and lexical information (Lexical Integration Deficit Hypothesis [LIDH]; Thompson & Choy, 2009). This study further evaluates LIDH by investigating cataphora processing. Cataphoric dependencies involve a long-distance relation between a preceding pronoun (dependent element) and its antecedent (licensing element), as in (1a). They are processed similarly to WhFG where the parser searches for the licensing element upon encountering the dependent element (Kazanina et al., 2007). Importantly, cataphora resolution is sensitive to structural (Binding Condition C) and lexical (gender) information, thereby allows examinations of the interaction between grammatical and lexical information. Method In an eye-tracking self-paced reading experiment, 6 individuals with agrammatic aphasia and 15 controls were tested using a paradigm where the c-command relation (c-command vs. non-c-command), the gender congruency (gender-matched vs. gender-mismatched) between the pronoun and potential antecedent, and the gender specification (stereotypical vs. definitional) were manipulated, as illustrated below. (1) a/b. After he/*she read the paper, the king happily went... c/d. He/she read the paper after the king happily went... e/f. After he/she read the paper, the minister happily went... g/h. He/she read the paper after the minister happily went... In (1a/b), the pronoun is grammatically linked to the potential antecedent (the king). Furthermore, because the gender of the antecedent is definitional (king-male), in (1b) we expect the parser to show a reading time slowdown at the gender-mismatched antecedent (i.e., gender-mismatched effect; GMME). In (1e/f) the antecedent is gender-ambiguous, but understood stereotypically as male or female (minister-male). Since the interpretation of stereotypical-gender nouns is affected by the gender of the pronoun in healthy individuals (Kreiner et al., 2008), in (1f), minister is interpreted as female. Consequently, this process requires additional integration costs. The LIDH predicts that since the integration process is impaired in agrammatic aphasia, the GMME is expected for patients but not for controls. In (1c/d/g/h), since pronoun-antecedent relation is impossible, no GMME is expected. Results Analyses of second-pass reading times revealed a significant interaction of c-command x gender x noun type at the critical antecedent region for the controls, associated with a GMME in the non-c-command/mismatched/definitional condition, but not in the stereotypical condition. Similarly, the patient group showed the same three-way interaction, but in the spillover region. However, GMMEs in both definitional and stereotypical c-command/mismatched conditions were found, with a greater GMME in the stereotypical compared to the definitional condition. Conclusion These findings indicate that the patients exhibited similar eye-movement patterns to that of the controls when processing definitional-gender nouns. However, unlike controls, they failed to use contextual gender information to revise gender stereotypes. These patterns support the LIDH.
Keywords
