Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Interplay Between Sound Perception and Lexical Prediction in Speech Accent Processing: An ERP Study
Poster Session E, Sunday, September 14, 11:00 am - 12:30 pm, Field House
Shang-En Huang1, Seana Coulson1,2; 1University of California, San Diego, 2Kavli Institute for Brain and Mind
Research on the perception of accented speech offers a valuable opportunity to explore the inter-relationship between acoustic and more semantically driven aspects of speech processing. The extent that contextual factors influence the perception of accented speech is currently unclear, and some controversy surrounds how less familiar speech sound variations from a foreign accent can influence the generation of lexical semantic predictions. To explore these issues, we recorded EEG from 40 native English speakers as they listened to sentences spoken with either a local or foreign accent. Each sentence frame had both a predictable (high cloze) or a less predictable (low cloze) ending, and all target words occurred in both the predictable and less predictable conditions. The use of multiple stimulus lists ensured that each participant heard only a single variant of each stimulus. Mixed effects models were used to analyze the latency and amplitude of the N1, P2, and N400 components elicited by the first word in each sentence and the sentence final (target) words. Contextual expectations are minimal at the first word, thus these analyses addressed whether foreign accents pose challenges for native listeners via a fixed effect of Accent and a random intercept for subjects. Analysis indicates the N1 and P2 components peaked later for the foreign accent (N1: p<.05; P2: p<.001); the P2 amplitude was larger for the local accent (p<.001); and the N400 was similar in both latency (p=.405) and amplitude (p=.373) for both accents. These results suggest that perceptual processing of foreign accented speech is less efficient, but its impact on lexical processing is negligible. To examine how semantic predictions modulate accent effects, we conducted similar analyses of ERPs elicited by sentence final words using models with fixed effects of Accent and Predictability and a random intercept for subjects. Results revealed larger P2 amplitudes for high cloze than low cloze words (p<.05), suggesting acoustic processing is more successful when supported by sentential context. Sentence final P2 peaked earlier for local accents (p<.05), indicating more efficient processing occurred irrespective of contextual support. Interestingly, sentence-final words elicited larger amplitude P2 in foreign than local accented speech (p<.05), suggesting listeners may rely more on predictive processing for the perception of foreign accents. Accent effects were observed on the latency of the N400 (earlier for local; p<.001) and Predictability effects were observed on N400 amplitude (larger for low cloze; p<.001). Planned comparisons revealed Predictability effects for words in both local (p<.001) and foreign (p<.001) accented speech, indicating participants were able to use contextual information to make lexical predictions regardless of accent. Lexical predictions from sentence context thus impacted the amplitude of the P2 component, but not the timing. Conversely, the familiarity of local accents impacted the timing of the N400, with negligible impact on either its overall amplitude, or the size of N400 predictability effects. Though accent and predictability effects were largely independent, results suggest listeners use context to facilitate acoustic feature extraction from foreign accented speech.
Topic Areas: Speech Perception, Meaning: Lexical Semantics