Extended Data Fig. 6: Cue-evoked dopamine release encodes cue-value RPEs.
From: Striatum-wide dopamine encodes trajectory errors separated from value
a, Non-TE DA cue responses averaged across sites and mice using a linear mixed-effects model (470: df = 315, n = 316 sites, 10 mice; 405: df = 230, n = 231 sites, 7 mice). b, Maximum cue-evoked z-scored ΔF/F during early learning for a subset of mice (df = 79, n = 80 sites, 3 mice). c, Violin plot of maximum non-TE coefficients for mice with quasi-simultaneous 470 nm and 405 nm recordings (two-sided t-test on wavelength effect: p = 1.33 × 10−18, t460 = –9.19, n = 231 sites, 7 mice). d, DA reward-history coefficients averaged as in a for mice with >30 unrewarded trials (470: df = 260, n = 261 sites, 8 mice; 405: df = 175, n = 176 sites, 5 mice). e, Z-scored ΔF/F aligned to cue onset following rewarded (red) or unrewarded (blue) trials, averaged across sites and mice (df = 260, n = 261 sites, 8 mice). f, Violin plot of maximum ΔF/F within 1 s from cue onset for trials in e (two-sided t-test on reward-history effect: p = 0.0052, t520 = –2.80, n = 261 sites). g, DA cue-identity coefficients averaged as in a (mixed-effects intercept: df = 315, n = 316 sites, 10 mice). Inset, maximum cue-identity coefficients vs contralateral bias index (Pearson r = 0.63, p = 0.025, n = 10 mice). h, Z-scored ΔF/F for cues signaling contralateral (red; bias cue in 9/10 mice) or ipsilateral (blue) directions, averaged as in e. i, Violin plot of maximum cue-identity coefficients (two-sided t-test on cue-identity effect: p = 6.77 × 10−7, t630 = –5.02, n = 316 sites, 10 mice). j, Mean number of trials with initial velocity directed contralaterally (red) or ipsilaterally (blue) to implant (two-sided paired t-test: p = 0.0037, t9 = −3.89, n = 10 mice). k, Reward rate for contralateral (red) vs ipsilateral (blue) cue trials (two-sided paired t-test: p = 0.0087, t9 = 3.33). l, Time to reward for contralateral (red) vs ipsilateral (blue) cue trials (two-sided paired t-test: p = 0.0066, t9 = −3.51). Shaded regions and error bars denote 95% confidence intervals; thin lines, individual-mouse averages. Box plots: centre, median; box bounds, first and third quartiles; whiskers, extremes value less than 1.5 x interquartile range from bounds.
