Pre-choice pupil dilation reflects the value of the upcoming choice.

(A, upper panel): Beta coefficients accounting for choice-related pupil dilation in the learning phase. Larger value differences between options (blue dashed line) elicited larger choice-related pupil dilations (black solid line) prior to choice (at t = 0). After choice, this relationship reversed, as smaller value differences elicited larger post-choice pupil dilations. (A, lower panel): Average choice-related pupil dilation for AB, CD and EF pairs. (B,upper panel): Beta coefficients of chosen and unchosen value regressors accounting for pupil size fluctuations in the pre-choice decision interval of the learning (left) and transfer phase (right). (B, lower panel): Beta coefficients of chosen and unchosen value regressors split by learning phase pairs, showing that pre-choice pupil size is modulated by values of the to-be chosen stimulus, irrespective of uncertainty. (C, upper panel): Differences in learned value beliefs between the chosen and unchosen option did not modulate choice-related pupil dilation prior to choice (at t = 0). However, smaller value belief differences elicited stronger pupil dilation after a value-based choice. (C, lower panel): We defined value differences between options presented in the transfer phase using the experimentally defined reward probabilities (see also Fig 1A). Value differences ranged from 10% to 60% and were subsequently divided into three categories to describe Small (10-20%), Medium (30-40%) or Large (50-60%) value differences between presented options. Small value differences between presented options elicited largest post-choice pupil dilation, in line with findings in C, (upper panel), suggesting that choice conflict drove pupil size after a value-based choice. Lines and (shaded) error bars of represent mean ± s.e.m of within-subject modulations. Horizontal significance designators indicate time points where regression coefficients significantly differentiate from zero (PP P<.01, repeated measures ANOVA.