Presentation

Hatice Zora¹, Valéria Csépe²; ¹Max Planck Institute for Psycholinguistics, Nijmegen, Netherlands, ²Brain Imaging Centre, Research Centre for Natural Sciences, Budapest, Hungary

Perception is an inferential process rather than emerging from a purely sensory operation. To derive an optimal interpretation of incoming data, the brain unifies sensory input with prior knowledge and experiences. Research into the neural origins of perception asserts a generative model where top-down predictions modulate the processing of bottom-up sensory input, and principally, stimuli that are predicted elicit a reduced neural response. Despite becoming increasingly influential in cognitive neuroscience, the precise account of these predictive models of perception remains debated. In this study, we aim to contribute to the debate at issue by investigating how neural circuits rely on present and past predictions to facilitate speech perception. Specifically, we look into inferential process of lexical access as involving simultaneous predictions at different domains, inter alia, phonological and semantic. As a working model, we chose prosody that codes not only phonological but also semantic information by special voice modulations, and as such, offers a unified model of bottom-up sensory and top-down inferential processes. Using the Mismatch Negativity (MMN) ERP component that is often referred to as a signature of predictive coding in auditory modality, we examined neural responses to deviations in lexical stress pattern, a structural property of a word that specifies which syllable in the word is more salient, in a passive auditory oddball paradigm. The stimuli were spoken Dutch words, which are segmentally identical but contrastive in their stress patterns, as in canon [ˈkaːnɔn] ‘cannon’ vs kanon [kaːˈnɔn] ‘canon’, and pseudo-words, imitating the acoustics of words such as tanon* [ˈtaːnɔn] vs. [taːˈnɔn]. While operating only at the phonological level in pseudo-words, lexical stress pattern leads to a semantic category change besides phonological variation in words. The preliminary results from nineteen native speakers of Dutch indicate MMN responses to both words and pseudo-words. The MMN response to pseudo-words might reflect a prediction error based on the most recent phonological input, which in turn can be explained through neural adaptation and local inhibition of cortical activation. The MMN response to words, on the other hand, is best justified by the unification of previously established semantic memory traces with phonological information. This evidence denotes simultaneous inference from sensory and cognitive sources, and a combination of predictive and integrative functions for a meaningful representation of speech.

Topic Areas: Prosody, Speech Perception