Fig. 1: Reverse correlation results (study 1).

a Dynamic prosodic representations. Normalized kernels derived from the reverse correlation analyses in both tasks (top: certainty, blue; bottom: honesty, green) across the three acoustical dimensions (pitch, loudness, and duration). Filter amplitudes (a.u., arbitrary units) correspond to the values obtained for each participant, task, acoustic dimension, and segment by subtracting the average (pitch, loudness, and duration) values obtained for stimuli judged as certain/honest from the values averaged for the unchosen stimuli and normalizing these values for each participant by dividing them by the sum of their absolute values. Data show group averages, with shaded areas showing the SEMs. Significant deviations from zero (one-sample two-sided t tests) are indicated at the corresponding segment positions by circles, with increasing sizes corresponding to p < 0.1; p < 0.05; p < 0.01, and p < 0.001; certainty task (p values per segment for pitch: 0.86, 0.69, 0.91, 0.64, 0.77, 0.49, 0.11, 0.11, 0.13, 0.14, 0.01, 0.004; loudness: 0.0005, 0.51, 0.37, 0.007, 0.38, 0.18, 0.0001, 0.0001, 0.44, 0.22, 0.12, 0.16; duration: 0.6, 0.03, 0.07, 0.04, 0.94); honesty task (pitch: 0.33, 0.29, 0.44, 0.34, 0.14, 0.09, 0.03, 0.06, 0.08, 0.30, 0.62, 0.30; loudness: 0.29, 0.24, 0.07, 0.01, 0.002, 0.53, 0.96, 0.17, 0.42, 0.5, 0.098, 0.88; duration: 0.98, 0.24, 0.30, 0.048, 0.94). Kernels were computed for 5 time points for duration (corresponding to the initial values of the audio transformations) and in 12 time points for pitch and duration (corresponding to post-transformation acoustic analysis of the stimuli, see “Methods”). Individual raw (i.e., non-normalized) kernels are shown in Fig. SII.a.b Sensitivity to mean features. To assess the extent to which mean pitch, loudness, and duration affected participants’ judgments at a static level, we constructed for each participant and task psychometric functions relating sensory evidence (computed for each trial as the area under the curve corresponding to the difference between the dynamic profiles of the first minus second stimuli) to participant’s choices (i.e., the probability to choose the first stimulus). Bar plots show the slopes averaged over the group separately in each task, with error bars showing the SEM. Dots show individual data. The white asterisk shows the result of one-sample Wilcoxon signed-rank test with p < 0.05; pitch (0.33/0.19), loudness (0.4/0.84), duration (0.012/0.053). c Sensitivity to feature variability. For each trial, the standard deviation of the pitch, loudness, and duration for the stimuli judged as more reliable (honest, certain) were subtracted from the stimuli judged as less reliable (lying, doubtful; Δ: difference). Bar plots show the slopes averaged over the group separately in each task, with error bars showing the SEM. Dots show individual data. White asterisks show the result of one-sample t test against chance with p < 0.05; **p < 0.01; ***p < 0.001; pitch (certainty p = 0.017/honesty p = 0.0002); loudness variability (0.7/0.4); duration variability (0.009/0.0007). Source data are provided as a Source data file.