Electrophysiology of gender agreement comprehension: evidence from Hindi

Abstract

Languages of the world use several dependencies to denote structural relations between words. One such dependency is agreement, wherein the co-occurring words in a sentence have similar values of certain features. Despite being well-researched in the fields of linguistics as well as sentence processing, previous studies on agreement processing have studied either how processing of the phi-features (person, number, and gender) vary or how agreement is processed across different constituencies (modifier-head, verb-argument). Further, these studies have revealed contradictory results both cross-linguistically as well as within a particular language. Moreover, even amongst the phi-features, person and number have garnered more attention than gender. In light of the existing research, two interesting questions crop up: The first being whether gender agreement processing is a monolithic universal phenomenon and the second being if there are variations in the neurophysiological correlates of gender agreement processing, what could those variations be attributed to. In this thesis, we try to find answers to the first issue and understand the source of variations, if they occur, using the oft-used EEG/ERP technique in an experimentally understudied language, namely, Hindi. We investigate the neurophysiological correlate of verbal gender agreement in Hindi, whereby the verb agrees with the highest nominative-marked argument in the sentence (Kachru, 2006; Mohanan, 1994; Bickel and Yadava, 2000). In a series of three visual ERP experiments, we try to find answers to the following questions. In the first experiment, using intransitive sentences, we scrutinize whether referential and grammatical gender influence the gender agreement computation process similarly or not. The results of this experiment revealed a long-latency N400 for the grammatical gender condition (inanimate subject) and a P600 for the referential gender condition (animate subject), thus pointing towards referential and grammatical gender modulating gender agreement processing differently. In the second experiment, we probe whether subject- and object-verb agreement evoke similar or different electrophysiological signatures, thereby examining whether verb agreement in Hindi is dependent on the type of argument the verb agrees with. A long-latency N400 was evoked for both subject and object verb agreement mismatches. This result suggests that the underlying mechanism in verb agreement is similar irrespective of the argument with which the verb enters into an agreement relation. In the third and final experiment, we examined the typologically different “default agreement” construction in Hindi to understand how gender agreement is processed in such a construction. Moreover, we also wished to determine if differences in intrinsic argument properties had an influence on the gender agreement computation mechanism. Hence, we also manipulated the object animacy (animate vs. inanimate). An N400-P600 was elicited for the animate-inanimate condition, whereas a P600 was evoked for the animate-animate condition. The qualitatively different results of the two mismatch conditions reflect that intrinsic argument properties such as argument animacy play a role in the gender agreement comprehension process. Collectively, the results reveal that verbal gender agreement in Hindi is not a monolithic phenomenon but is sensitive to fine-grained distinctions, including the intrinsic properties of arguments. The empirical evidence from the thesis adds to the body of cross-linguistic work on language comprehension and underlines the need for further research to understand the fine-spun differences in agreement processing.