Fig. 1: Experimental task, pain ratings, and visualization of interaction effects.

a Time-course of one experimental trial. Participants either chose the upcoming stimulus intensity (controllable runs) or confirmed the color of the circles (predictable, unpredictable runs) by selecting the correct button displayed on the lower part of the screen. Number of remaining stimuli of each intensity level were shown inside the circles in the controllable and predictable condition. In the unpredictable condition a zero was shown inside the circles for the duration of the entire run. To guarantee ratings for all intensity levels in controllable runs, the limitation was imposed that participants had to select each intensity level five times in one run. After the button press, the circle was highlighted, which corresponded to the intensity in size that was either self-chosen (controllable) or script-determined (predictable). In the unpredictable condition, all circles were highlighted for 3 s. After a jittered anticipation phase (2–5 s) that was added to the information period, a white fixation cross turned red and a 4 s long painful heat stimulus of either low (VAS 30) medium (VAS 50), or high (VAS 70) intensity was applied. After stimulus offset the cross turned white again (2–5 s jitter) before the rating scale was shown. Each condition was repeated once, resulting in 6 runs and 90 trials in total. b Pain ratings of the fMRI sample (n = 59 participants; 590 ratings per condition x intensity combination) in the controllable (C, dark green), predictable (P, light green), and unpredictable (U, yellow) condition. Large black points in the boxplots indicate mean values, boxes show the 25th and 75th percentile around the median line, whiskers extend to ± 1.5 × IQR, outliers are shown as small black points. c Posterior distribution of interaction parameters and fitted lines from the linear model. The grey curves (posteriors) show the estimated probability densities of the parameter value. A positive value of the interaction parameter (beta) indicates a slope increase over intensities for the indicated condition pair, i.e., when moving from the unpredictable to the predictable (U to P), from the unpredictable to the controllable condition (U to C) and from the predictable to the controllable condition (P to C), ratings show a stronger increase from the low to the high intensity level. Posteriors of the interaction parameters are above zero, implying an effect on ratings. The schematic plot on the right shows the simplified fitted lines derived from model fitting, which illustrate the changes in slope between the controllable, predictable, and unpredictable condition.