Fig. 4: State-dependent encoding model better predicts neuronal activities.
From: Human motor cortex encodes complex handwriting through a sequence of stable neural states
a–c, R2 of the model that explains neural activity using state-dependent directional tuning (with neurons across 20 sessions, M ± s.d.). The state-dependent directional tuning model significantly enhances the R2 for neurons (n = 157) with reliably tuning to characters, and the mean R2 increased from 0.0873 to 0.2873 (state-dependent versus random states, paired two-tailed t-test, t(156) = 13.94, P = 3.37 × 10−29, Cohen’s d = 1.092; state-dependent versus state-independent, paired two-tailed t-test, t(156) = 14.66, P = 3.897 × 10−31, Cohen’s d = 1.265) (a). For simple-tuning neurons (n = 65), the mean R2 increased from 0.1748 to 0.3715 (state-dependent versus random states, paired two-tailed t-test, t(64) = 8.799, P = 1.271 × 10−12, Cohen’s d = 1.086; state-dependent versus state-independent, paired two-tailed t-test, t(64) = 9.622, P = 4.716 × 10−14, Cohen’s d = 1.291) (b). For complex-tuning neurons (n = 92), the mean R2 increased from 0.0254 to 0.2279 (state-dependent versus random states, paired two-tailed t-test, t(91) = 10.77, P = 6.226 × 10−18, Cohen’s d = 1.353; state-dependent versus state-independent, paired two-tailed t-test, t(91) = 11.02, P = 1.904 × 10−18, Cohen’s d = 1.527) (c) Source data.
