Fig. 1: Diverse optimism in value coding across ACC neurons.

a, Top, on each trial, subjects chose between two cues of neighboring probability value. Bottom, each probability value could be denoted by two stimuli, resulting in two stimulus sets (see ref. ⁸ for task details). b, Example responses from three separate neurons demonstrating different levels of optimism. In each plot the mean firing rate is plotted as a function of time and split according to the chosen value (probability) level. There are four chosen values (0.3–0.9 probability) because subjects rarely chose the 0.1 probability level (choice accuracy was at ceiling: 98%). Insets demonstrate that the firing rate is a nonlinear function of value. Mean firing rate (z-scored across trials) in a 200- to 600-ms window after cue onset is plotted as a function of the four values. Reversal points are the interpolated values at which there is 0 change from the mean firing rate, an index of nonlinearity. Shaded regions and error bars denote s.e.m. c, Histogram showing a diversity of reversal points across ACC RPE-coding neurons. Coloring denotes optimism as defined by reversal point, with red being more optimistic. d, Scatter plot showing reversal points estimated in half of the data strongly predicted those in the other half. Each point denotes a neuron. Inset, log(P values) of Pearson’s correlation between 1,000 different random splits of the data into independent partitions. Across partitions, the mean R = 0.44 and geometric mean of the P values was P = 0.003 (black line). Bootstrapping to obtain a summary P value was also significant (P < 0.01). e, Scatter plot showing reversal points estimated in stimulus set 1 strongly predicted those in stimulus set 2 (R = 0.41, P = 0.009). Each point denotes a neuron. f, AP topographic location of the neuron predicted its reversal point, with more anterior ACC neurons being more optimistic (R = 0.37, P = 0.016). As we had two independent noisy measures of each neuron’s optimism (reversal point and asymmetry; Fig. 2), we used the mean of the two measures (after z-scoring them), which we call ‘neuron optimism’. Neuron optimism is plotted against the normalized AP locations within ACC. The normalization ensures that, for example, the most anterior portion of the ACC in one animal corresponds to that in the other. Each point denotes a neuron. See Extended Data Fig. 6 for further analyses.