Fig. 4. Identifying ENN units (i.e., fMRI vertices) containing relevant task rule (context) and motor response (behavior) representations.

a We identified the task rule input and motor output representations in empirical data using MVPA and univariate task activation contrasts. b A 12-way decoding of each of the task rules (across the 3 rule domains) using task activations (estimated from a task GLM) during the encoding period of the C-PRO task. We applied this 12-way decoding to every parcel, given that task rule activations have been previously shown to be widely distributed across cortex⁸. Chance decoding was 8.33%; statistical maps were thresholded using an FDR-corrected p < 0.05 threshold. c To identify the motor/output activations, we performed a univariate contrast, contrasting the middle versus index finger response activations for each hand separately. Finger response activations were estimated during the response period, and univariate contrasts were performed on a vertex-wise basis using all vertices within the somatomotor network⁶⁵. Contrast maps were statistically thresholded using an FDR-corrected p < 0.05 threshold. The resulting finger activations matched the placement of finger representations in the well-established somatomotor homunculus in the human brain.