Fig. 4: PfFL neural activity encodes reach direction early, but not late, in learning. | Nature Communications

Fig. 4: PfFL neural activity encodes reach direction early, but not late, in learning.

From: Dissociable roles of distinct thalamic circuits in learning reaches to spatial targets

Fig. 4

a Predicted initial vector direction of combined held-out test data from 10 models trained on PfFL neural activity using different train/test splits, plotted against the corresponding true initial vector direction. Models trained on an early session in block 1. b Same as (a), for VALFL animal. c Circular histogram showing the true (left) and predicted (right) initial vector directions from a representative PfFL animal data in (a). d Circular histogram showing the true (left) and predicted (right) initial vector directions from a representative VALFL animal data in (b). e Averaged coefficients of determination (R2s) of model predictions for PfFL (purple), VALFL (green), and models trained on shuffled vector data (gray) for 5 days of training in each block (averaged over two blocks). 2-way repeated measures ANOVA: day effect, F(4,32) = 4: *p = 0.0115, region effect, F(1,8) = 17: *p = 0.0031. Šídák’s multiple comparisons test: PfFL early-mid vs VALFL early-mid, *p = 0.021. PfFL mid vs VALFL mid, *p = 0.043. PfFL late vs VALFL late, *p = 0.029. For all other day comparisons, p > 0.05. For shuffled data: 2-way repeated measures ANOVA, day and region effects, p > 0.05. f Averaged coefficients of determination (R2s) of model predictions for PfFL using neural activity pre (black) and post (purple) onset of movement. 2-way repeated measures ANOVA: time from onset effect, F(1,4) = 62: *p = 0.0014. Šídák’s multiple comparisons test: pre vs post (early), *p = 0.0008, pre vs post (early-mid), *p = 0.0084, pre vs post (mid), *p = 0.0052, pre vs post (mid-late), p > 0.05, pre vs post (late), *p = 0.0001. g Same as in (f), for VALFL neural data pre (black) and post (green) movement onset. 2-way repeated measures ANOVA: time from onset effect, p > 0.05. h Averaged coefficients of determination (R2s) of multivariate model predictions for PfFL (purple), VALFL (green), early day block average. 2-way repeated measures ANOVA for PfFL: Model effect, F(4,32) = 19: *p < 0.0001. Šídák’s multiple comparisons test for PFFL: direction vs full model, *p = 0.0353. Direction vs speed/distance, *p < 0.0001. Full model vs speed/distance, *p = 0.0219. Direction/distance vs speed/distance, *p = 0.0018. Direction/speed vs speed/distance, *p = 0.0025. For all other comparisons, p > 0.05. Šídák’s multiple comparisons test for VALFL: Direction vs speed/distance, *p = 0.0001. Direction/distance vs speed/distance, *p = 0.0056. Direction/speed vs speed/distance, *p = 0.0023. For all other comparisons, p > 0.05. e–g Center and error bars indicate mean ± SEM. See also Supplementary Fig. 8, 9. Source data are provided as a Source Data file.

Back to article page