Higher reward leads to more temporally precise task-related responses.

(A) Distributions of response (“Resp.”) times per trial, estimated as positions of the corresponding template match peaks (“Match Peak pos.”), same data set as in Fig 2A–2E. Distributions are separated by reward size, with color coding as indicated in the key (common to (A, B) and to all later figures). Data are shown as a vertical “violin plot” histogram with numbers of trials increasing from 0 (middle) upwards for high reward (“Hi”) and downwards for low (“Lo”). Similar displays are used for all such comparisons of distributions to avoid clutter (e.g., with interleaved histograms). Clustering of response times per reward size was quantified as the 2 standard deviation (“2x StdDev”) width of timing distributions. (B). Distributions of response amplitudes per trial, estimated from template matching (“Tmplt Match”), shown separated by reward size following the same conventions as in (A). Response amplitude per reward size was quantified as the median (indicated by arrowheads; medians are indicated similarly in all later amplitude distributions). Significance (p-values) in (A,B) were obtained from bootstrap with 10,000 resamples. (The 2x StdDev values per reward size [“hi”,”lo” in panel A] and median amplitude per reward size in panel B shown in these and other panels are not the sample medians from the measured distributions but rather are medians obtained from the same bootstrap procedure used to get p-values.) (C) Comparing the 2 standard deviation width of the response time distribution for all correct high-reward trials versus that for all correct low-reward trials, per experiment (“expt”; N = 30). (D) Comparing median response amplitudes for all correct high- versus all correct low-reward trials, per experiment. p-Values in (C), (D): Wilcoxon signed rank tests for pairwise comparisons (data in S11 Data).