Extended Data Fig. 9: Influence subspace contributions across neurons.

Here used the canonical vectors defined using the inactivation effect maps of individual muscles. a, Cumulative distributions of weights for the activity of individual neurons in canonical vectors defined by CCA using the inactivation effect maps of individual muscles. Weights were normalized so the median value for each mouse is 1 to enable comparison across mice, as different numbers of neurons were recorded in each mouse. Tests for multimodality were not significant for all distributions (p-values ranged from 0.64 to 0.99, Hartigan’s dip test). Results indicate a broad distribution of neuronal weights for each muscle in each mouse. b, Cumulative distribution of norms of the four-element vectors composed of each given neuron’s canonical vector weight from alignment with the inactivation effect maps of individual muscles. Tests for multimodality were not significant for all distributions (p-values ranged from 0.34 to 0.99, Hartigan’s dip test). Results again indicate a broad distribution of neuronal weights for each muscle in each mouse. c, d Same as a, b, but weights (c) and weight vector norms (d) are scaled by each neuron’s mean firing rate during climbing. Tests for multimodality were not significant for all distributions (p-values ranged from 0.90 to 0.99, Hartigan’s dip test). e, Scatterplot of weight vector norms as in b versus the depth below pia assigned to the waveform centroid of each neuron. Red whisker plots span from the first to the third quartile for all neurons assigned to each laminar group (those within dotted boundaries). Data from three mice are combined. The distributions of weight vector norms for neurons localized to layers 5 (p = 3 × 10⁻⁵, two-sided Wilcoxon’s rank-sum test) and 6 (p = 7 × 10⁻⁴) were significantly higher than those localized to superficial layers. f, Same as e, but weight vector norms are scaled by each neuron’s mean firing rate during climbing. The distributions of weight vector norms for neurons localized to layers 5 (p = 5 × 10⁻¹⁴, two-sided Wilcoxon’s rank-sum test) and 6 (p = 3 × 10⁻⁷) were significantly higher than those localized to superficial layers.