Extended Data Fig. 5: Response timecourses exhibit diverse proportions of early and late timecourses.

Each subplot depicts results for a single condition at a single vertex (Dataset D1). The left shows FIR timecourses (black, with lines and error bars indicating mean and standard error across two condition-splits) and the overall fit of the TDM model (purple). The right shows beta estimates (bars and error bars indicate mean and standard error across six condition-splits). To select which cases to show, we first identified vertices whose R² under the TDM GLM is greater than 10%. We then examined the estimated betas and calculated their t-values (beta divided by standard error across condition-splits). We determined (i) all cases with a robust Early beta (t > 5) and a weak Late beta (absolute value less than 1/10 of the Early beta), (ii) all cases with robust Early and Late betas (t > 5) and where each beta is at least 9/10 of the other beta, and (iii) all cases with a robust Late beta (t > 5) and a weak Early beta (absolute value less than 1/10 of the Late beta). Finally, we randomly selected 20 cases from each of the three groups.