Figure 4.

The energization signal in pupil rate is independent of value and choice difficulty, relates only to effort (not reward), and is only evident in phasic (not tonic) arousal. A–C, Control analyses replicating the pupil-rate effects reported in Figure 3 while regressing out (one at a time) the effect of reward (A), RT (B), and choice difficulty (denoted dSV; C). These analyses establish that the effort-by-choice pupil-rate effects are not confounded by reward and choice difficulty. Panels D, E show selectivity analyses to rule out reward effects on pupil rate (D) and effort-by-choice interaction effects on tonic arousal (pupil baseline; E). The absence of effects in the selectivity analysis confirms the specificity of our findings for effort-by-choice signals and phasic arousal as measured in pupil rate. First column, Dots with error bars represent means ± 1 SEM. Lines are linear fits of the means [using the MATLAB polyfit(x,y,1) function]. Middle column, Boxplots display the median (central line), 25th and 75th percentiles (bottom and top edges), and non-outlier low and high extreme values (bottom and top error bars). Blue lines show subjects whose effort slope is higher in effortful choice than in non-effortful choice, gray lines show subjects who show the opposite effect. Symbols indicate significance levels between two conditions: star for p < 0.001, triangle for p < 0.01, diamond for p < 0.05, ns for p > 0.05. Right column, Each data point represents a subject; p values represent significance level from robust regressions. rew=reward; dSV=absolute delta subjective value; resid= residual.