Extended Data Fig. 3: Temporal profile, raw encoding R2, and noise ceiling.
From: Dissecting neural computations in the human auditory pathway using deep neural networks for speech
a) The temporal receptive field (absolute beta weights of the spectrotemporal encoding model) of each individual speech-responsive unit/electrode. (Gray shaded areas indicate random permuted distributions of the averaged TRF across all units/electrodes, same as Fig. 2c. b) The histogram of the optimal delay window lengths corresponding to models in Fig. 2a. c) raw prediction R2 of different models. Dashed line: noise ceiling estimated from 10 repeated trials. d) Distribution of the normalized brain prediction score of each model across individual units/electrodes. Dashed line: noise ceiling estimated from 10 repeated trials. Red star (*) indicates the best model for each area, black dot (.) indicates other models that are not statistically different from the best model (p > 0.05, two-sided paired t-test; n = 50 neurons for AN; n = 100 neurons for IC; n = 53 electrodes for HG; n = 144 electrodes for STG). Box plot shows the first and third quantiles across electrodes, orange line indicates the median, black line is the mean value, and whiskers indicate the 5th and 95th percentiles.
