Fig. 5: Social vector representations are modified by training in a pursuit task.

a Experimental paradigm of the pursuit test in a circular track. Scale bar: 10 cm. b Distribution of the latency to obtain a reward across days (5 imaged mice; mean ± s.e.m.), centered around the transition between naïve and trained (day 0). c Distribution of latency for naïve and trained sessions (n (naïve) = 35 sessions, n (trained) = 117 sessions; median ± i.q.r., Mann–Whitney test, two tailed, Mann–Whitney U = 73, Cliff’s Delta = 0.96, p = 2.8 × 10⁻²⁷). d Representative examples of cells classified as egoSVC in trained (right) but not in naïve (left) sessions (one cell per row). Each example includes the rate map (left), the trajectory with events (center) and the rate vs. angle map. Maximum rate (P) and information content (I) are indicated. Scale bar: 20 cm. e Subpanels show for naïve (blue; n = 27 sessions) and trained (green; n = 81 sessions) conditions: proportion of cells classified as egoSVC, mean calcium event rate, mean information content, number of fields and field size (median ± i.q.r., p-values for two-tailed Mann–Whitney tests indicated, Mann–Whitney U = 897, 860, 637, 447.5, 370.5; Cliff’s Delta = 0.18, 0.22, −0.42, 0.60, 0.67). f Cross-validated absolute decoding error for naïve and trained sessions (n (naïve) = 26 sessions, n (trained) = 79 sessions, mean ± s.e.m., two-tailed Mann–Whitney test, Mann–Whitney U = 440, Cliff’s Delta = 0.57, p = 6.0 × 10⁻⁶). In box plots, the central line indicates the median, and the bottom and top edges of the box mark the interquartile range. Whiskers extend from −1.5 × i.q.r. to +1.5 × i.q.r. from the closest quartile, where i.q.r. is the interquartile range. Black dots mark outliers. Source data are provided as a Source Data file.