Fig. 4. An encoding model uncovers task-specific differences across pyramidal neuron types.

a, Schematic of the encoding model. Top, task-related and movement-related variables account for fluctuations in cortical activity. Bottom, weights for each variable define a spatiotemporal event kernel, revealing cortical activity in response to a specific event (example shows right stimulus kernel in EMX mice). b, Average maps of cvR². The model accurately predicted cortical variance for all PyNs, with over 90% explained variance in the frontal cortex of PT mice. c, cvR² from two models, using only movement (‘Movement’) or task (‘Task’) variables. In all groups, movements were more predictive than task variables and accounted for the majority of the explained variance of the full model (dashed lines). Circles denote sessions. d, Contributions of movements, stimulus and choice to the model’s total explained variance. Although movements contributed the most, stimulus and choice also made sizable contributions.