Article Figures & Data
Figures
- Figure 1.
Experimental design. A, Stimuli were created by convolving 600 unique dry speech samples (50% male and 50% female voices) with one of 30 reverberant IRs. B, Reverberant IRs comprised real-world reverberations and their synthetic (“fake”) variants (ecological, linear decay, time-reversed, flat- and inverted-spectral dependence). C, The task consisted of listening to a reverberant sound lasting 2,000 ms, followed by a 500 ms blank before an untimed response display appeared. Subjects judged whether the reverberation was real or synthetic (“fake”) by clicking on the “R” or “F,” whose locations varied pseudorandomly trial by trial to prevent motor response preparation before the response cue. Responses were followed by a jittered interval varying between 700 and 1,500 ms. D, EEG data analysis pipeline: MVPA using retrospective sliding window.
- Figure 2.
Behavioral and neural signatures of reverberant perception. A, Group-level performance (mean accuracy ± SEM across participants) for real and fake trials overall and broken down by synthetic variants. B, Real versus fake average decoding time courses across subjects. C, Decoding time courses of real versus fake variants averaged across subjects. The vertical dashed lines at zero and 2 s indicate the stimulus onset and offset, respectively; the horizontal dashed line indicates the decoding percentage at the chance (50%); and the horizontal colored bars in the x-axis indicate significance. D, Time course of brain–behavior correlation (Spearman’s correlation, rho) relating pairwise decoding of real and fake variants (panel C) to each participant's behavioral performance (panel A). E, Decoding accuracy time courses for the first 5 blocks of trials (black) versus the last 5 blocks of trials (violet). F, Decoding time courses for trials labeled by response button location (black), speaker gender (Male vs Female; cyan) and Impulse Response Time (red). For all statistics, N = 20; t test against 50%; cluster-definition threshold, p < 0.05; 1,000 permutations. *p < 0.05; **p < 0.01; ***p < 0.001.
- Figure 3.
Neural signatures of reverberation perception. A, Decoding accuracy time courses for real versus fake trials labeled by physical (black line) and perceived (pink line) categories. The vertical dashed lines at 0 and 2 s indicate the stimulus onset and offset; the horizontal dashed line indicates chance (50%), and color-coded bars in the x-axis indicate significance. The violet area denotes the time when the physical and perceived decoding differed. B, Real versus fake decoding accuracy for correct (orange) and incorrect (gray) trials, compared against physical (black) adjusted to match trial count for correct condition. For all statistics, N = 20; t test against 50%, cluster-definition threshold, p < 0.05; 1,000 permutations. For incorrect response trials, N = 18 because two subjects did not have sufficient real incorrect trials.
- Figure 4.
TGM of real versus fake decoding according to physical stimuli, perceptual report, and correct response. A, Physical real versus physical fake. B, Perceived real versus perceived fake. C, Real correct versus fake correct. White contours indicate significant clusters across subjects and dashed lines at 0 and 2 s indicate the stimulus onset and offset. For all statistics, N = 20; t test against 50%, cluster-definition threshold, p < 0.05; 1,000 permutations.
- Figure 5.
Temporal generalization matrices from two subsets of electrodes. The temporal cluster, pooled across electrodes and marked in pink in the layout, comprised the following electrodes: FC5, FC6, FT9, FT10, C3, C4, CP5, CP6, T7, T8, TP9, and TP10. The centroparietal cluster, marked in gray in the layout, included Cz, Pz, P3, P4, CP1, and CP2. A, Real versus fake (physical stimuli). B, Real correct versus fake correct (correct). The white regions in the TGMs indicate significance. For all statistics, N = 20, t test against 50%, cluster-definition threshold, p < 0.05, 1,000 permutations.
In this issue
