Fig. 6: Recurrent dynamics and connectivity within the sensorimotor module.

a Neural trajectories during the intertrial interval (ITI) colored by rule, visualized in the space spanned by the first three principal components. Black dots represent the start of the ITI. Only trials following a correct trial were included. b Neural trajectories during the response period colored by response location, visualized in the space spanned by higher order principal components. Black dots represent the start of the response period. Only trials following a correct trial were included. c The principal angle between the subspaces spanned by neural trajectories during different task rules (gray distribution represents the principal angle obtained through shuffled data, see Methods). Each data point represents one trained network. d The principal angle between the subspaces spanned by neural trajectories during different responses (gray distribution represents the principal angle obtained through shuffled data, see Methods). Each data point represents one trained network. e The connectivity biases between different rule-selective populations across models. Exc: excitatory neurons. f The same as e but for different response location-selective populations. Exc: excitatory neurons. g The same as e but for different shared feature-selective populations. Exc: excitatory neurons. h. The results in e–g show that neural populations selective for different rules, response locations and shared features mutually inhibit each other. Data in c–g are shown for networks with subtractive dendritic nonlinearity. Networks with divisive dendritic nonlinearity show similar result (Supplementary Fig. 10).