Extended Data Fig. 10: Effect of fear conditioning and familiarity on drift.

a, Conditioning experiment. Day −1: present one odorant paired with shock (CS+) and a second without shock (CS−) in a conditioning chamber. Days 0 and 16: administer conditioned (CS+ and CS−) and four additional neutral odorants to head-fixed mouse while recording neural signals and measuring behavioural responses in a virtual burrow assay²⁹. Days 1–15: record neural signals in head-fixed mouse without test odorant administration. b, Behaviour. Left, trial-averaged ingress amplitude (n = 5 mice) across time on days 0 and 16 on trial blocks 2–7 (shading, 95% CI). Grey bar, 4-s odorant stimulus epoch. Right, mean ± 95% CI ingress amplitude during the final second of the odorant epoch on blocks 2–7. For days 0 and 16, CS+ vesus CS− and CS+ versus neutral, P < 1.4 × 10⁻³, Wilcoxon rank-sum. c, Neurophysiology. Left, scatter plots showing single-unit response magnitude for all three stimulus classes (mean spontaneous baseline-subtracted evoked responses computed during the odorant stimulus epoch) of odour–unit pairs on day 0 versus day 16 (CS+: n = 148 odour–unit pairs, CS−: n = 129 odour–unit pairs, neutral stimuli: n = 482 odour–unit pairs, data pooled across 5 mice). Black dashed line, unity; blue dotted line, linear regression; blue shading, 95% CI. Regression was performed across all odour–unit pairs that showed a significantly modulated response on at least one of the two days (Wilcoxon rank-sum, α = 0.001). Middle, cumulative distributions; right, mean ± 95% CI (n = 5 mice) of corrected angles for all three classes of stimulus. For all comparisons, P > 0.05 (Wilcoxon rank-sum). We note that classical conditioning reduces within-day variability (unpublished observations). Thus, the odour–unit pair response correlations reported here, which are not corrected for within-day variability, are higher than in other experiments (for example, Fig. 2a), but measures that correct for within-day variability, such as corrected angle or drift rate, are comparable. d, Familiarity experiment. Mice were presented with a panel of four neutral odorants daily over a 32-day interval (days −16 to 16; familiar). Starting on day 0, a panel of unfamiliar odorants was presented at 8-day intervals. e, Mean odour-evoked response magnitude (spontaneous baseline-subtracted, computed during the odorant stimulus epoch) of odour–unit pairs across intervals of 8 days (left, familiar odorants: n = 741 odour–unit pairs, unfamiliar odorants: n = 1,137 odour–unit pairs) and 16 days (right, familiar odorants: n = 371 odour–unit pairs, unfamiliar odorants: n = 570 odour–unit pairs), data pooled across 5 mice. Black dashed line, unity; blue dotted line, linear regression; blue shading, 95% CI. Regression was performed across all odour–unit pairs that showed a significantly modulated response on at least one of the two days (Wilcoxon rank-sum, α = 0.001). f, Across-day classification accuracy (4-way, SVM, linear kernel, L2 regularization, scaled between chance and maximum within-day performance to account for slight differences in within-day performance between the two conditions). Solid lines, mean; shading, 95% CI. g, h, Cumulative distribution (g) and mean ± 95% CI (h) of corrected angles from n = 5 mice. Unfamiliar, ρ = 0.48, P = 1.0 × 10⁻⁴; familiar, ρ = 0.23, P = 0.08.