Extended Data Fig. 10: Distractor-impact on persistent activity in ALM is learning-dependent.

Spike rate modulations in left ALM in distractor-naive versus distractor-trained mice, in response to weak-distractors computed using correct trials. a, Spike rates of example cells and b, Activity along the Choice mode in the presence of weak distractors in distractor-naive (top, n = 1356 cells) and distractor trained mice (bottom, n = 1061 cells). Lick-left trajectory without stimulation (red), and with weak distractors during early-delay (gray), or late-delay (black). Lick-right trajectory during sample-epoch stimulation (blue). Note that weak distractors had a persistent effect in ALM of distractor-naive mice. Specifically, distractors shifted the activity from lick-left trajectory towards lick-right trajectory. This was evident from shifts in projections on Choice mode (b) and from activity of individual cells that contributed to this mode (a). In contrast, in distractor-trained mice, the effect of distractors was transient. c, Impact of weak distractors on Choice mode in distractor-naive (gold) and distractor-trained mice (purple). Trajectories were aligned to the onset of each distractor. Data is shown as average across distractors with different onset times and sessions ± s.e.m. (shaded). In distractor-naive mice the effect of distractor persisted for at least 0.8 s (the time interval from the late-delay distractor onset and the Go cue), whereas in distractor-trained mice the activity recovered to the unperturbed trajectory. d, Schematics of putative attraction basins in distractor-naive (left) and distractor-trained (right) mice (cartoon). Shallow basin of attraction in distractor-mice mice allowed sufficiently strong stimuli to switch the neural activity from one basin to another.