Pupillometry analysis
Pupillary changes were first baseline corrected on a trial-by-trial basis by subtracting the mean change in pupil diameter 1000ms before the beginning of tactile stimulations. Next, to control for inter-individual variability, pupil data were Z-scored for each subject across all conditions (Basile et al., 2021; Rudebeck et al., 2014). In each trial, missing samples due to blinks or loss of the eye-tracking signal during the tactile stimulation period were interpolated via spline interpolation using the nearest valid adjacent samples. Pupil responses were then averaged across trials for each condition. Based on visual inspection of the average response profile, the mean change in pupil diameter was extracted for the time window ranging from 0-4 seconds after stimulus onset (Figure 2b ). Data were analyzed via a 2-way repeated-measures ANOVA with Hand type (Human vs. Artificial) and Touch type (Dynamic vs. Static) as within subject factors. Post-hoc analyses following significant main effects and interactions were performed by running two-tailed pairwise t-tests, and multiple comparisons were corrected using false discovery rate (FDR; Benjamini & Hochberg, 1995). All p values < 0.05 were considered significant.
To test the hypothesis that Dynamic_Human touch alone induced a larger pupil size than Dynamic_Artificial plus Static_Human touch, supralinearity was quantified by contrasting, for each participant, the average pupil size in the Dynamic_Human condition against the sum of the average pupil size in the Dynamic_Artificial plus Static_Human conditions. The effect of Dynamic_Human condition was then compared with the added Dynamic_Artificial and Static_Human condition with a paired-sample t test to determine significance.
Last, we investigated whether the blink rate changed across the four conditions by using nonparametric Wilcoxon Signed-Rank tests.