From minimal dependencies to sentence contexts: neural correlates of agreement processing

Abstract

Language comprehension is incremental, involving the integration of formal and conceptual information from different words, together with the need to resolve conflicting cues when unexpected information occurs. However, despite the extensive amount of findings regarding how the brain deals with these information, two essential and still open questions are 1whether the neural circuit(s) for coding syntactic and semantic information embedded in our linguistic code are the same or different, and 2whether the possible interaction(s) between these two different types of information leaves a trace in the brain response. The current thesis seeks to segregate the neuro-anatomical substrates of these processes by taking advantage of the Spanish agreement system. This system comprised those procedural mechanisms concerning the regular assignment of the number [singular, plural], person [first, second and third] and/or gender [feminine, masculine] information, associated with different sentence constituents. Experimental manipulations concerning different agreement features and the elements involved in an agreement relation, allowed us to characterize the neural network underlying agreement processing. This thesis comprised five experiments: while experiments I and II explored nominal dependencies in local as well as non-local relations, experiments III, IV and V explored subject-verb relations in a more complex sentence context. To distinguish between purely syntactic mechanisms and those where semantic and syntactic factors would interact during language comprehension, different types of agreement relations and/or agreement features were manipulated in well- and ill-formed constructions. The interaction effect between the different factors included in each experiment was always the critical comparison. In general, our results include firstly a functional dissociation between well-formed and ill-formed constructions: while ill-formed constructions recruited a bilateral distributed fronto-parietal network associated to conflict monitoring operations, not language specific, well-formed constructions recruited a left lateralized fronto-temporo-parietal network that seems to be specifically related to different aspects of phrase and sentence processing. Secondly, there was an anterior to posterior functional gradient associated to the middle and superior temporal cortex that consistently appears across experiments. Specifically, while the posterior portion of the left MTG-STG seems to be related to the storage and retrieval of lexical and morpho-syntactic information, the anterior portion of this region was related to syntactic-combinatorial building mechanisms. Critically, in the most anterior part of the left temporal cortex, corresponding with the middle and superior temporal pole, form-to-meaning mapping processes seems to be represented. Thirdly, the response of the left temporal cortex appears to be controlled by left inferior frontal regions (LIFG). Finally, left parietal regions such us the angular gyrus showed increased activation for those manipulations involving semantic factors (e.g., conceptual gender and Unagreement constructions), highlighting its crucial role in the processing of different types of semantic information (e.g., conceptual integration and semantic-discourse integration). Overall, these findings highlight the sensitivity of the agreement system to syntactic and semantic factors embedded into an agreement relation, opening new windows to the study of agreement computation and language comprehension.