Item

Abstract

Communicative situations typically provide listeners with contextual information. One such information is speaker identity. Listeners are able to adapt to the individual language use of a particular speaker and can use this information in order to generate expectations. This can be observed even for syntactic structure. In this fMRI study, we asked how top-down effects of expectations are implemented on a neural level and how they affect syntactic processing. Twenty-eight participants were presented with German sentences spoken by two different speakers. Speakers differed in the probabilities by which they produced sentences with a particular syntactic structure [Subject-Object-Verb (SOV) vs. Object-Subject-Verb (OSV)]. One speaker had a high probability to produce SOV sentences and a low probability to produce OSV sentences (SOV-Speaker), and vice versa for the other speaker (OSV-Speaker). After exposure to the speakers in a training session, listeners were able to generate expectations about a sentence structure solely on the basis of speaker identity. These expectations were then investigated in a test session that was conducted in the MRI scanner. In order to disentangle syntactic processing and syntactic expectations we used two types of trials: (1) Unambiguous sentences where the syntactic structure information was provided by the case-marking of the determiners and expectations were based on speaker identity, and (2) ambiguous sentences where the determiners were replaced with noise and therefore only speaker information but no syntactic structure information was provided. Performance in the comprehension task showed an interaction between Speaker and Structure. The typical performance advantage of the easy SOV sentences over the complex OSV sentences was reduced for the OSV-Speaker compared to the SOV-Speaker . Analysis of the unambiguous trials showed a main effect of syntactic structure with increased activation for the OSV structure in a left-lateralized network including the IFG (BA44), pMTG, precentral gyrus and preSMA. In a next step, we analyzed sentences on basis of whether they were expected or unexpected and not based on their syntactic complexity. Unexpected structures showed increased activation in the right IFG (BA44) and bilateral network comprising the MFG, Angular Gyrus, preSMA and IFG (BA47). Crucially, analysis of the ambiguous trials allowed to investigate speaker-specific expectations in the absence of syntactic structure information. We observed increased activation for the OSV-Speaker compared to the SOV-Speaker in the bilateral Insula and Putamen and the left ACC/preSMA. While the processing of complex syntactic structures activated a left-lateralized fronto-temporal network, the processing of unexpected structures (independent of complexity) activated a bilateral frontal-parietal network with a focus to the right hemisphere. This may index the recruitment of attentional mechanisms when an expectation is violated. Interestingly, in absence of any explicit syntactic structure information speaker-expectations were shown to activate cortical and sub-cortical brain regions. In case of the OSV-Speaker this may reflect a controlled modulation of linguistic processing when the default syntactic structure needs to be inhibited in favor of the structure that is expected on basis of the speaker. These results give insight into the neural implementation of expectations in a communicative context.