Fig. 5: Learning new tasks results in new preparatory activity.
From: A combinatorial neural code for long-term motor memory
a, Possible outcomes of preparatory states over continual learning. b, Task structure of task context 1, 2 and 3. In task context 3, mice report frequency of a pure tone using directional licking after a delay epoch. Time interval, 31.50 ± 9.65 days between task contexts 1 and 2, 66.75 ± 26.36 days between task contexts 1 and 3; mean ± s.d. across fields of view. c, Selectivity index in task contexts 1, 2 and 3 for neurons with significant trial-type selectivity (P < 0.001, two-tailed t-test) during the delay epoch. Red, neurons preferring lick left; blue, neurons preferring lick right. Significant selectivity and trial-type preferences are determined in task context 1. d, Schematic of movement-specific activity trajectories in activity space and CDDelay vectors across task contexts. Red and blue shades, preparatory states for lick left and lick right, respectively. e, ALM activities from an example field of view projected on the CDDelay from task context 1. Thick lines are the mean and thin lines show single trials. f, Decoding accuracy of the CDDelay from task context 1 tested on task contexts 1, 2 and 3 (n = 8 fields of view, 3 mice). Task context 1 versus 3, ***P = 5.56 × 10−5. Decoding accuracy of the CDDelay from task context 2 (n = 8 fields of view, 3 mice). Task context 2 versus 3, ***P = 6.08 × 10−4. Decoding accuracy of the CDDelay from task context 3 (n = 8 fields of view, 3 mice). Comparing to task context 1 decoder, **P = 0.0022; comparing to task context 2 decoder, **P = 0.002. Two-tailed paired t-test. Data are mean ± s.e.m.
