Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Spectral continuity determines perceptual timing precision in complex sound sequences
Poster Session C, Thursday, October 1, 10:45 am - 12:45 pm, Wangari Maathai
Charlotte M. Mock1,2,3, Leon Ilge1,4, Yulia Oganian1,2,3; 1Centre for Integrative Neuroscience Tübingen, 2Graduate Training Centre of Neuroscience Tübingen, 3International Max Planck Research School for The Mechanisms of Mental Function and Dysfunction, 4University of Tübingen
In everyday listening, temporal structure allows listeners to predict the timing of upcoming sounds. In speech, predicting the timing of acoustic or linguistic units (e.g., syllables) may support the analysis of temporal variation relevant for segmentation, prosody, and emphasis. To use temporal information effectively, listeners must evaluate onset timing relative to preceding temporal structures with high perceptual timing precision (PTP). Previous research reports seemingly conflicting results: On one hand, PTP is better for simple (short, sharp-onset) sounds than for complex (longer, slow-rising) sounds. This has been suggested to reflect a reduced ability to evaluate the timing of temporally extended complex sounds. On the other hand, high timing precision has been reported for speech, which also falls in the latter category. This suggests that target sound acoustics alone cannot explain PTP. Here, we hypothesized that PTP depends not only on the acoustics of individual sounds, but also on the temporal extent and spectrotemporal structure of preceding sound contexts. Across five experiments, participants heard an isochronous cueing sequence and iteratively adjusted a target sound’s timing relative to it until reaching perceptual isochrony. We estimated PTP as the variance in timepoints perceived as isochronous across repeated adjustments of identical cue-target combinations. Exp.1 (n=37) tested whether PTP following complex cues improves with the temporal extent of the cue sequence (i.e., number of preceding cues). A short click target was preceded either by four clicks or four to seven slow-rising harmonic tones. As expected, PTP was lower in harmonic-tone contexts than in click contexts. However, it did not improve with cue sequence length. Next, we tested in Exp. 2 (n = 42) whether reduced PTP following slow-rising cues might be driven by spectrotemporal dissimilarity between cue and target sounds. We used the same sounds (click vs. slow-rising harmonic tone) as cues and targets, yielding matched and mismatched cue–target pairings. Surprisingly, PTP was best for identical cue-target combinations, regardless of their acoustics. To test whether spectral or temporal envelope similarity underlies this effect, Exp.3 (n=32) orthogonally manipulated both factors. It revealed that the cue-target similarity effect was dominated by spectral similarity, whereas effects of temporal envelope similarity emerged only with musical aptitude. Exp.4 (n=31) further showed that PTP improves with spectral overlap between cue and target sounds, regardless of harmonicity. Exp.5 (n=31) replicated this finding in vowel sounds, where spectral overlap is naturally high. As expected, PTP remained high across pitch and vowel shifts between cues and target. Finally, we developed a spectrally resolved oscillatory entrainment model to capture these effects. Oscillatory entrainment reflecting cue and target spectra alongside spectrally-specific onset and offset dynamics successfully captured all observed behavioral effects. Together, our findings reveal and characterize a previously unknown constraint on perceptual sensitivity to rhythmic sequences: effective temporal prediction depends not only on rhythmic structure, but also on shared spectrotemporal properties of successive sounds. At the neural level, this may reflect frequency-specific entrainment in auditory cortices. Our results shed light on mechanisms underlying listeners’ ability to maintain precise temporal expectations in naturalistic environments, including speech.
Topic Areas: Speech Perception, Computational Approaches