Poster D2, Monday, March 26, 8:00-10:00 am, Exhibit Hall C
Decoding natural continuous speech in native and non-native brain
Shweta Soni1, Matthew S Tata1; 1Centre for Canadian Behavioural Neuroscience, University of Lethbridge, Alberta
Identifying the boundaries of speech events such as syllables, phonemes and words is crucial in order to comprehend speech and might be affected by the differential familiarity with spoken languages. Here, we attempt to investigate 1) how differently speech features are encoded in native and non-native English speakers during continuous speech perception, and 2) can the predictability of these speech features be gained over successive presentations of the same speech segment. In this study, non-native late learners of English were compared with native-English speaking Canadians. We found that non-native speakers showed significant differences in behavioral performance as compared to natives along with improvement in performance as a function of repetition of the speech sample. Furthermore, the mapping between different features of speech stimuli and neural responses was modeled using linear regression and then these models were examined on their ability to predict EEG data as well as stimulus (e.g., envelope). We found that non-natives were poor on reconstructing speech envelopes and encoding low-level spectrotemporal features than natives but improved over successive speech presentations. We suggest that the brain’s mechanisms of interpreting linguistic information might depend on familiarity with the deep structure of a language. We will also discuss the role of higher-level processes like phonetic feature processing in speech perception.
Topic Area: ATTENTION: Auditory