Extended Data Fig. 3: Two distinct phases of learning, motor learning in Phase 1 and associative learning in Phase 2.

In Phase 1, we train hungry mice to perform stereotyped forelimb reaches to obtain the food pellet. In this phase, food pellets are presented at random times. In Phase 2, we train mice to associate the cue with the presentation of the food pellet. Hence, in this phase, mice learn to associate the cue with the forelimb reach. a, Reach outcomes from an example mouse over Phases 1 and 2. Each trial is one cue presentation. Drop means the mouse dislodged the pellet but failed to consume it. Miss means the mouse reached but did not touch the pellet. Success means the mouse successfully grabbed and consumed the pellet. Failures (drops and misses) decrease during Phase 1 motor learning. No further improvements in success rate in Phase 2. b, 3D paw tracking at 255 frames per second (fps). left, Average trajectory of reaches from Phase 2 sessions from an example mouse (n = 412 reaches from 3 sessions). All reaches aligned to the time when the forepaw is part-way to the pellet during the initial ballistic movement of the forelimb toward the pellet. right, Example single reaches during Phase 1 from the same example mouse, showing variable trajectories and a non-stereotyped reach. c, Reaches from same example mouse as b during Phase 2 after pairing the cue with the food pellet. left, Example single reach trajectories overlaid. right, Reach rate over time aligned to the cue (blue bar represents the cue). top to bottom, Each row is a different example session from beginner, intermediate, and expert stages of learning about the cue. Note no further refinement of reach trajectories, despite the mouse shifting the timing of the reach to the time window immediately after the cue.