a, Lick raster plots for an example mouse during the first (top) and the last training day (bottom). During food-approach training, a 9-second, 10-kHz tone is associated with the availability of condensed milk at a metal spout. For the lick raster plots, licks were plotted at 5 licks/sec when the sensor was tonically triggered by licking; this does not affect the quantifications in panels b-c. b, Summary data for lick probability during training. Relative lick probability is the average probability of licking the spout within a 4.5-second window during the stimulus, divided by the lick probability during the 20 seconds before or after the stimulus. Mice lick the spout specifically in response to the tone on the fifth day of training (relative lick probability > 1, P = 0.002, one-sided permutation test) but not on the first day (P = 0.09). c, Summary data for reward-port approach probability during training. Relative approach probability is the average probability of moving from the back of the conditioning box to the side where the spout is located in response to the tone, divided by the approach probability at other random time points during the session. Mice approach reward specifically in response to the tone on the fifth day of training (relative approach probability > 1, P = 0.02, one-sided permutation test) but not on the first day (P = 0.41). For panels b and c, gray lines are individual mice and the green line is the mean. N = 5 mice.