Extended Data Fig. 5. Choice-related activity in somatosensory cortex.
(a) Averaged choice kernel maps for EMX mice during the initiation and delay period. Dashed circles show location of somatosensory whisker (SSWhisker), somatosensory nose (SSNose), and frontal cortex. SSWhisker (blue trace) and SSNose (red trace) were constantly positive or negative, respectively, even during the initiation period. In contrast to frontal cortex (yellow trace), both areas were only weakly modulation by the stimulus onset (gray box). (b) Choice kernel maps for IT mice during the delay period. Dashed circles show location of auditory, SSWhisker, and frontal cortex. Choice-related activity in SSWhisker (red trace) increased over the course of the trial. No choice-related modulation was apparent in frontal cortex. (c) Choice kernel maps for PT mice during the delay period. Conventions as in (b). Choice-related activity strongly increased in frontal cortex after stimulus onset and was weaker in other cortical areas. (d) 10x cross-validated decoder performance, predicting animal’s left/right choices at different times during the trial. In all PyN types, decoder performance was above chance at all times, including the initiation period before the stimulus (gray box). This suggests that, in some trials, animals follow a pre-conceived choice that is stimulus-independent and can be decoded from cortical activity. Decoder performance was highest in the response period (dashed vertical line) when animals performed licking movements. (e) Contralateral choice weight maps during the delay period (same as in Fig. 6c). Dashed circles show the location of somatosensory whisker cortex (SSWhisker). In all PyN types, choice weights in SSWhisker were increased in the initiation period before the stimulus (gray box). A potential explanation could be that pre-stimulus choices are reflected in choice-specific whisker movements. However, choice signals in SSWhisker persisted when removing movement-related activity from the imaging data (Supplementary Fig. 12). Whisker or other facial movements might therefore be too subtle to be captured by our analysis or choice signals in SSWhisker reflect non-overt choice-related activity.