Fig. 3: Neuronal activity is most strongly influenced by spatial position.
From: A persistent prefrontal reference frame across time and task rules
a Position is the best predictor for calcium activity. Left: Explained variance of GLMs using all or single predictors (F = 93.71, p = 3*10−12; full (white) vs position (green): t = 8.19, p = 8*10−5, full vs speed (purple): t = 11.23, p = 10−5, full vs goal (blue): t = 10.05, p = 2*10−5, position vs speed: t = −9.42, p = 3*10−5, position vs goal: t = 9.16, p = 4*10−5, speed vs goal: t = 5.82, p = 6*10−4). Right: Decrease in explained variance with a single predictor shifted in time (F = 87.99, p = 10−8; position vs speed: t = −9.39, p = 3*10−5, position vs goal: t = 9.62 p = 3*10−5, speed vs goal: t = 3.67, p = 0.008). b Most cells encode for position (green; speed: purple; goal: blue) and for one or two variables in total. Left: Percentage of cells significantly modulated by the different predictors (F = 38.34, p = 2*10−6; position vs speed: t = −7.08, p = 2*10−4, position vs goal: t = 8.70, p = 5*10−5, speed vs goal: t = 2.71, p = 0.030). Right: Number of variables they are modulated by (F = 16.35, p = 10−5; 0 vs 1: t = −4.80, p = 0.002, 0 vs 2: t = −2.74, p = 0.029, 0 vs 3: t = 1.84, p = 0.108, 1 vs 2: t = −0.02, p = 0.984, 1 vs 3: t = 5.83, p = 6*10−4, 2 vs 3: t = 16.67, p = 7*10−7). c The contribution of predictors is stable across days. There is a main effect of model type (full model: black; position: green, speed: purple; goal: blue), with position outperforming the other single predictor models (left, F = 210.13, p = 8*10−16; full vs position: t = 6.62, p = 3*10−4, full vs speed: t = 15.93, p = 9*10−7, full vs goal: t = 14.64, p = 2*10−6, position vs speed: t = −15.32, p = 10−6, position vs goal: t = 15.43, p = 1*10−6, speed vs goal: t = 6.78, p = 3*10−4) and the largest decrease in explained variance for the position-shifted model (right, F = 245.46, p = 10−11; position vs speed: t = −12.29, p = 10−6, position vs goal: t = 17.28, p = 5*10−7, speed vs goal: t = 3.32, p = 0.013). In both analyses, ANOVA identified no effect of time (F = 1.49, p = 0.146; F = 1.12, p = 0.358) nor model*time interaction (F = 1.13, p = 0.287; F = 0.96, p = 0.526), suggesting stable contribution of the individual predictors over time. Thick lines indicate mean across animals, thin lines are individual mice. One-way repeated measures ANOVA (a) and (b), two-way repeated measures ANOVA (c), followed by paired t-tests with Šidák correction. Boxplots show median and lower/upper quartile of data, circles show individual mice (n = 8 for all comparisons). Source data are provided as a Source Data file.
