A) Choice task. Monkeys began the trial by fixating a central target. Next, the central target disappeared and two peripheral targets appeared 15° to the left and right of center. If the monkey shifted gaze to one of these targets, juice was delivered. If he chose the left target (contralateral to the recording site), immediately following juice delivery an image appeared for 500 ms centered on the site of the chosen target. For non-image choices (right), 500 ms of blank screen followed juice delivery. Ten to 20% of trials were single target trials, which were identical except that only one target appeared and only movements to that target were rewarded. The lower panel illustrates outcome contingency changes over three example image blocks. B) Estimating social image value. For all choice trials, the frequency of choosing the image was plotted against the relative juice volume delivered for that choice. These data were fit with cumulative normal functions, and horizontal shifts in those functions were used to estimate image value (). Example psychometric functions from four image blocks of one behavioral session are shown. C) Monkeys differentially valued image classes (data pooled across all sessions: one way ANOVA, F_{(3, 355)} = 2.67, p < .05; individually: monkey Ot [closed bars] and Os [open bars]: p = .23, p < .03, respectively). D) Viewing duration varied with social image class. For image target choices, the average distance of gaze from the center of the image is plotted as a function of time. Vertical lines depict image onset and offset. The horizontal line is placed at half the height (or width) of the image (3.5°). Shading represents SEM. Due to a data collection error, looking time data is restricted to 36 of 71 behavioral sessions (12 sessions from monkey Os). E,F) Response times were inversely related to juice value and directly related to image valence. Response times separated by juice outcome and image class (E), or juice outcome and image value quartiles (F). The image value quartile boundaries were −8.47, −0.09, and 8.20 ms open juice valve time (, see also ).

A) Coronal slice MRI of monkey Os with recording chamber and grid attached. The plane of the slice is at the center of the recording chamber (~22 mm anterior to interaural 0). The red square highlights the area of recording. B) Recording locations overlaid on an outline of the striatum [adapted from ]. When multiple units were recorded in the same session, locations have been staggered slightly (< 0.5 mm) for visualization. C) Example neurons showing differential firing rates for social image category or juice value when the monkey shifted gaze to the image target. Average normalized firing rates separated by image category (left panels) or juice value (right panels) are plotted for 0.5 s or 1 s epochs aligned to acquisition of fixation target, saccade onset and juice or image onset. Gray shading indicates ti mes of significant firing rate differences by sliding-window ANOVA analysis (). Insets illustrate recording location of each example neuron. D, E) Heterogeneous population coding of social image class and juice value. For each neuron that significantly differentiated image category or juice level during any epoch, the average firing rate for each image category or juice level during all significant time windows was calculated and ordered from least to greatest. The number of cells with each possible ordering is plotted (see also ).

A) Pseudocolor plot depicting the modulation index (strength of the influence of social image category relative to that of juice volume on neuronal firing rates), aligned to three trial events when monkeys shifted gaze to the image target. Neurons are ordered from medial to lateral recording position along the y-axis. B) Modulation indices regressed against medial-lateral recording position (linear regression) when monkeys shifted gaze to the image target. Filled circles represent neurons from monkey Os, open circles represent neurons from monkey Ot. All regressions were significant at p < 0.005 when performed for individual monkeys. C) Average modulation indices for each trial event, separated by striatal region (as depicted in , open bars: monkey Os, closed bars: monkey Ot) when the monkeys shifted gaze to the image target. One way ANOVAs (striatal region × modulation index) collapsing across monkeys: all p values < 10⁻⁷. * = p < 0.01, ** = p < 10⁻⁴ (post hoc two-tailed t-test). Results were similar when data were separated by monkey. D, E and F) Conventions as in A, B, and C with analyses performed on data from trials in which the monkeys shifted gaze to the blank target (see also ).

A) Scatterplots and histograms of the variance in firing rate explained by image category for each cell when the monkey shifted gaze to the image target (y-axis) or the blank target (x-axis, ). * indicates the mean of the distribution is significantly above the diagonal (p < 0.001, t-test against 0). B) Scatterplots and histograms of the variance in firing rate for each cell explained by juice outcome when the monkey shifted gaze to the image target or the blank target. None of the means of the distributions were different from zero (p > 0.4).