Fig. 5: A context-specific functional group reorganizes during prey capture.

a–d Monkey TY. a The kinematics + reachFN encoding model versus the network encoding model trained with spontaneousFN network features and tested on reachFN network features. Units above the dotted line are defined as the context-specific functional group (green, 39/175 units), and the remaining are context-invariant units (orange, 136/175 units). b The context-specific functional networks, with reachFN1 on top and spontaneousFN on bottom. The color scale of edge weights is displayed under (a). c The context-invariant reachFN1 and spontaneousFN. d The cumulative distribution of the difference in edge weights between reachFN1 and spontaneousFN for the context-specific, context-invariant, and full (purple) groups. Distribution comparisons are inset as colored pairs of boxes. The * indicates significant differences in the median with p < 0.01 (two-sided median test). All comparisons had p ≈ 0.0. e–h Corresponding results for MG, with 9/73 units in the context-specific functional group. h The * indicates significant differences in the median with p < 0.01 (two-sided median test), as in (d). P-values are 6.9 × 10⁻¹⁴ for context-specific vs. context-invariant, 7.7 × 10⁻¹⁴ for context-specific vs. full, and 5.7 × 10⁻¹² for context-invariant vs. full. Source data are provided as a Source Data file.