Evidence Integration in Natural Acoustic Textures during Active and Passive Listening

Reaction times and response rates suggest improved performance for stimulus exposure and larger changes in stimulus statistics. A, Reaction times decreased significantly as a function of change time and change size. Within each change size, reaction times did not reach statistical significance for individual change sizes. Reaction times decreased significantly as a function of change size (mixing coefficient, across colours). B, Response rates increased significantly with change time for all change sizes >0. As expected, the response rates did not increase in catch trials (light blue) with change time, indicating that subjects closely listened for changes, rather than responding as a function of time within the trial. Response rates also increased significantly as a function of change size (different colors). In addition, a two-way ANOVA was performed, which also indicated a significant effect of both factors. Error bars represent 1 SEM. Details on statistical testing given in the main text.

Onset responses are similar across levels of task engagement. A–C, The scalp distribution of the P2 component (200 ms after stimulus onset) is centered with a slight asymmetry toward the front. Since the different stimulus conditions are all matched at the onset, the data were averaged over all stimuli. The scalp topography in the passive–aware conditions mirrors the active potential (B). In the passive-naive conditions, the topography was still centered, but slightly broadened toward frontal locations, probably due to uncontrolled blinks (C). Overall, the topographies were not statistically different (see text for details). Subject involvement was supported by their relative level of α-band activity (Extended Data Fig. 3-2). D, The central potential, typically associated with auditory-cortex activity (Nie et al., 2014), exhibited a classical response sequence after stimulus onset, with a negative component (N1, 150 ms) followed by a positive component (P2, 242 ms), both slightly delayed compared to classical latencies for pure tones (left, average over electrodes Cz, C1 and C2). The P2 peak size decays slightly but nonsignificantly from active to passive-aware and passive-naive (right, height averaged in [0.2-0.3]s). E, The onset-elicited response in the PO region also displayed a negative deviation around [0.2,0.3]s (left, electrodes Pz, POz), which likely reflects the negative part of the auditory cortex dipole (Extended Data Fig. 3-1) This activity did not differ significantly across conditions. Error bars in all plots indicate ±1 SEM. Effect size is given in parentheses after the p value.

PO potential depends on stimulus evidence and integration time during task engagement. A, Scalp topography of change-elicited activity at 0.7 s post-change exhibits a single peaked positivity over the PO regions for correct detection (hit) trials. Only trials with the lowest difficulty, i.e., MC = 0.6 were included for clarity of display. B, In the central electrodes (auditory cortex associated, as in Fig. 3), the change in statistics leads to a delayed ERP only for the longest change times and the largest mixing coefficients. Data are aligned to change time and averaged over hit trials only. Black lines at the bottom indicate window selected for the analysis of height (300–350 ms) and slope (200–330 ms). The responses for different change times are shifted vertically to avoid crowding. Gray trials indicate conditions with <10% of response percentage and correspondingly more variability. If all trials are included, variability is comparable between conditions (Extended Data Fig. 4-1). C, The PO potential (electrodes as in Fig. 3) first turns slightly negative and then becomes strongly positive, lasting until the end of the sampling period of the second texture. The potential's height (600- to 1000-ms window) and slope (400- to 600-ms window) depended significantly on change time and size. Colors as in B. The subsequent panels show the dependence of the slope and height of the respective potentials for all stimulus parameters (left column, color bar represents slope or height) or in relation to change size (middle) and change time (right) only. D–F, Height of the central potential did not depend significantly on change size or change time (see B for measurement range). G-I, The slope of the central potential depended significantly on change size, while it did not depend on change time (see B for measurement range). J, The PO potential's height also depends systematically on change time and size (see C for measurement range). K, The height of PO potential varied significantly with change size both if catch trials are included or not. L, The PO potential's height dependence on change time was also significant. M, The slope of the PO potential increased with both change time and change size. N, Change-related build-up in PO potential is significantly faster with increasing change size. O, The PO slope dependence on change time was highly significant. All error bars indicate ±1 SEM, and p values of the two-way ANOVA for the factor on the x-axis are denoted in the figure, all tests are described in the main text. p⁺ is the significance when leaving out the no-change condition. Effect size is given in parentheses after the p value.

Same analysis as in Figure 4 but for misses. Nomenclature identical to Figure 4. No dependencies were found, except for a central potential, slight, but inverted dependency for height (with a borderline p value). Gray curves indicate conditions containing <10% of possible trials.

The PO potential is reduced but stays stimulus dependent for passive-aware subjects. A, The post-change scalp topography (600–700 ms) is dominated by posterior components, but the overall potential size is much smaller (only trials for MC = 0.6 included for display); however, note the scale difference in comparison to Figure 4. B, In the central electrodes, the change in statistics elicited a small ERP complex (positive peak ∼320 ms) for the longest change times and largest change sizes (colors as in Fig. 3, see legend). C, In the PO electrodes, the change in statistics also led to a slow negative-positive potential, reaching a plateau around 700 ms. D–F, The height of the central potential neither depended on change size nor time. G–I, The slope of the central potential showed a borderline significant dependence on change size, but not change time (measurement range: 200–330 ms). J, The PO potential's height varied systematically with change time and size. K, The PO potential was significantly larger for bigger change sizes; however, the size of the potential was much smaller than in the active condition, corresponding to the change in topography. L, The PO potential also increased slightly but significantly in height with change time (see Results for interpretation). M, The slope of the PO potential exhibited a similar dependence on change size and time as in the active condition (compare to Fig. 4G). N, The PO slope is significantly steeper as a for larger change sizes. O, The PO slope displayed a nonsignificant tendency to be faster for longer change times. Conventions as in Figure 4.

In the passive-naive group the PO potential exhibits much weaker but partially significant dependence on change size and time. A, The post-change scalp topography (600–700 ms) exhibits a more diverse shape than in the active or passive-aware conditions (only trials for mixing coefficient equal 0.6 included for display). The scale bar was kept the same as in Figure 5, to emphasize the difference in potential size. B, In the central electrodes, the change in statistics again elicited a small ERP complex (positive peak around 320 ms), as in the other conditions (colors as in Fig. 3, see legend). C, In the PO electrodes, the change in statistics lead to residual potentials with much more diversity between stimulus properties than in the other conditions. D–F, The height of the central potential depended significantly on change size but not on change time. Measurement range (300–350 ms post-change) is shown in B. G–I, The slope of the central potential also depended significantly on change size but not on change time. J–L, The PO potential's height depends significantly on change time, however, with a nonmonotonic progression. M–O, The PO potential's slope depends significantly on change size; however, this effect is driven by the last condition with MC = 0.6. Conventions as in Figure 4.

Task involvement scales change-related potentials A–C, Change-time aligned, central potentials exhibited different profiles across task involvement. The initial P2-like response (peak at 313 ms) did not differ significantly across involvement levels [height (B), slope (C)]. However, the potentials diverged strongly thereafter, with increasingly negative slopes for more active involvement. Data averaged over all change times and the two largest mixing coefficients (0.3,0.6). D–F, The PO potentials exhibited clear differences in slope (E) and height (F) across different levels of task involvement. For the passive-naive condition, the potential appears to be a combination between a lasting negative potential in combination with a more transient positive potential added. All error bars indicate ±1 SEM, and p values are denoted in the figure, all tests are described in the main text. Effect size is given in parentheses after the p value. Same electrodes as in Figure 3.