Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Emergence of lexical stress encoding in the human temporal cortex
Poster Session A, Wednesday, September 30, 11:00 am - 1:00 pm, Wangari Maathai
Ilina Bhaya-Grossman1, Yulia Oganian2, Emily Grabowski3, Edward Chang4; 1Stanford University, 2University of Tübingen, 3University of California, Berkeley, 4University of California, San Francisco
Lexical stress, which distinguishes the noun “PRE-sent” from the verb “pre-SENT”, critically facilitates segmentation and word recognition in speech. Lexical stress is perceived by listeners as categorical patterns of alternating stressed and unstressed syllables, yet it is cued by multiple, distinct acoustic features, including duration, intensity, and pitch. The neural mechanism by which stress is represented, either as an abstract linguistic category independent of its acoustics, or as an emergent integration of multiple acoustic cues, remains unknown. Using high-density intracranial recordings from eleven English-speakers passively listening to natural sentences, we found that neural populations on the human temporal cortex encode lexical stress category. A majority of neural populations that encoded stress category also encoded at least one acoustic cue to stress, most often intensity or pitch, suggesting that categorical stress encoding arises from cue integration. Single-neuron recordings from 18 Neuropixel probes revealed that individual neurons largely do not encode stress category, indicating that this representation emerges only at the neural population level. In eight Spanish speakers listening to the same (English) speech, stress encoding was significantly weakened, suggesting its encoding depends on language experience. Finally, using an active task in which participants identified stressed syllables, we found that stress-encoding neural populations are sensitive to prior syllable context. Together, these findings demonstrate that categorical stress encoding on the human temporal cortex emerges at the population level through language- and context- dependent integration of multiple acoustic cues.
Topic Areas: Prosody, Speech Perception