Mechanism and paradigm.

a phenomenology results from external (solid arrow) and internal contributions (dashed arrow). The integrated brightness should affect pupil size: Light (dark) synesthetic colors should cause constrictions (dilations) at equal physical luminance in synesthetes, but not in controls where externally and internally generated brightnesses align. b We expected synesthetes’ pupils to be larger for reported lower brightness and smaller for reported higher brightness. c Paradigm. Block 1: a digit was presented. Participants (except passive controls) subsequently indicated the color that most closely corresponded to the digit in their opinion. This was followed by an interstimulus interval (ISI). Block 2 (synesthetes only): a disk was presented, colored according to the synesthete’s average indicated color for that digit. At its center sat a gray patch matching the luminance and pixel area of the original digit from Block 1, together allowing assessment of externally triggered light responses.

a Reported colors per grapheme on all trials for synesthetes (left) and controls (right). b Synesthetes showed (near) perfect grapheme-color consistency and moderate to very strong grapheme-color couplings (rainbow circles), while controls reported none to moderate coupling and varied in consistency (grey circles). Note that higher consistency is reflected in lower color distance, hence lower values [17]. Larger dots indicate group means. c,d (HS)Lightness of color reports per synesthete (c) and control (d). Blacked dashed line represents lightness being 0.5. See Supplementary Figure 1 for color reports on the hue and saturation axes.

Pupil size change to graphemes, median-split by reported color lightness (dark gray = low lightness; light gray = high lightness).

Top row: pupil responses to graphemes in controls. Mid row: pupil responses to graphemes in synesthetes. Bottom row: pupil responses to colored discs in synesthetes (Block 2). a, c, e Depict average, baseline-corrected and within-participant demeaned pupil responses. Shaded error bands: ±1 SE across participant means. b, d, f depict mean pupil size (800–4000 ms) for dark vs. bright colors. Dots show individual participants; squares denote grand means with 95% CIs as whiskers. Dot luminance corresponds to the participant’s average synesthetic color lightness per bin, dot size to the number of trials. **: p < .01, ***: p < 0.001 based on within samples and one sample t-tests. Significance relative to zero for lightness bins (left, right) and between bins (center). Participants with less than 25 trials per bin excluded for visualization (controls: n = 3, synesthetes: n = 4, see a, c and Supplementary Figure 4 for visualization without demeaning and Supplementary Figure 3 f9or visualization without data exclusion).

Results of per-time-point linear mixed effects model (LME) predicting pupil size in synesthetes while presented with graphemes.

Covariates for the individual graphemes and intercept are not visualized here. a depicts t-values of the LME over time. Horizontal lines denote significance threshold (p = 0.05 dashed, p = 0.01 dot-dashed, p = 0.001 dotted). Higher lightness was associated with smaller pupil size (red), this effect was stronger for stronger reported grapheme-color couplings (orange), with a trend for higher PA scores (purple). Furthermore, higher lightness constricted the pupil more for stronger grapheme-color couplings in synesthetes with higher PA scores (gray, three-way interaction). b-d visualize interactions for the LME run on the average pupil size between 800 ms and 4000 ms. Dotted de-notes low, dashed high of median splits. b Interaction of grapheme-color coupling strength with lightness: lightness affected the pupil more when grapheme color couplings were reported higher. c Interaction of PA scores with lightness: lightness affected the pupil more for synesthetes with higher PA scores, but note that this effect only reached borderline significance for a short interval. d three-way interaction of lightness, coupling strength, and PA score. See Supplementary Figure 6 for the same model in controls.

Average pupil responses to graphemes from baseline, split by group: controls picking a color forced-choice (’active’, gray), controls passively viewing the graphemes (’passive’, black), and synesthetes (purple).

a Pupils dilated more for active controls than both synesthetes and passive controls. Shaded error bands represent 95% confidence intervals across participant means. Horizontal black line represents average pupil size during baseline. b Mean pupil size (0.8 s–4 s interval) per group and participant. Dots show individual participants; squares denote grand means with 95% CIs. c as a, but for the velocity of pupil size changes (first derivative, filtered). d as b, but for the velocity of pupil size changes and the 0.7 s-2 s interval. p < .01: **, p < .001: ***, based on two-sided independent sample t-tests.