Extended Data Fig. 2: Signal-to-noise ratios for determining recording site inclusion.

a, Power spectral density (PSD) plots for example sites with high (included sites, top) and low (excluded sites, bottom) signal-to-noise ratios (SNRs). Horizontal line indicates the noise power density, estimated as the average power density within the 10-14 Hz frequency band (see Methods). b, Average baseline-corrected ΔF/F aligned to cue onset on all congruent trials for the same example fibers in a. c, Peak cue-evoked ΔF/F as a function of log SNR, each data point represents one recording site, n = 638 sites across 10 mice. The vertical line represents the point where the signal is equal to the noise (SNR = 1). Recording sites to the left of the vertical line (SNR < 1) were excluded from future analyses. d, Violin plots showing the log-transformed SNR for sites classified (via micro-CT localization) as within (n = 356 sites) or outside of (n = 182 sites) the striatum. SNR > 1 in blue, SNR < 1 in red. Thin lines represent average SNR for individual mice. Error bar shows mean and 95% confidence interval estimated from linear mixed effects models (in striatum: t₃₅₅ = 9.60 p = 4.33 × 10⁻¹⁹, n = 356 sites, 10 mice; out striatum: t₁₈₁ = 1.35, p = 0.53, n = 182 sites, 10 mice, two-sided t-test on model intercepts). Effect of recording site classification on SNR estimated from a linear mixed effects model, two-sided t-test t₆₃₆ = 8.62, p = 1.62 × 10⁻¹⁶, n = 638 sites, 10 mice. P-values were post hoc Bonferroni corrected for three comparisons.