Neural decoding reveals the functional anatomy of auditory integration and competition in speech perception

Resumen

At the beginning of a word, many continuations are possible – for example, when hearing packet, listeners may consider pass and package. Cognitive science shows that listeners in this situation immediately consider multiple words and narrow down the set of candidates via competition. However, cognitive neuroscience has not yet been able to identify the cortical locus of this competition. One potential source of this difficulty is the hypothesis (tested here) that representations of unfolding auditory stimuli take the form of distributed patterns of activity, rather than the overall quantity of activity. We investigated competition dynamics by combining direct recordings from the cortex of humans listening to spoken words with machine learning to decode those distributed patterns. Two critical first steps in investigating the neural basis of these competition dynamics were to identify regions a) whose pattern of activity supports multiple candidates; and b) whose pattern of activity persists over time (after stimulus offset). We characterized both properties in six language regions by using machine learning to assess which words were under consideration in each area every 25 ms during listening. Auditory areas – early in the processing stream – exhibited competition dynamics lasting even after stimulus offset. This pattern was not seen in higher-level language areas. Thus, competition among distributed patterns of activity in auditory sensory areas may underlie human word recognition.

Áreas y palabras clave

Cómo citar (APA 7)

Rhone, A., Farris-Trimble, A., Nourski, K., Kawasaki, H., III, M. A. H., & McMurray, B. (2022). Neural decoding reveals the functional anatomy of auditory integration and competition in speech perception. https://doi.org/10.31234/osf.io/bd6eh