Back

Repetition Coherence Reveals Computational Principles of Auditory Object Formation

Bastug, B.; Sun, Y.; Adolfi, F.; Schroeger, E.; Poeppel, D.

2026-01-14 neuroscience
10.64898/2026.01.14.699508 bioRxiv
Show abstract

Repetition is a powerful cue for auditory object formation. In natural environments, however, repetitions are rarely exact copies of sound sequences and often overlap with competing sounds. To characterise the temporal integration dynamics and the evidence required for perceptual organisation of acoustic scenes into objects, we developed a repetition-coherence framework inspired by motion-coherence paradigms in vision. Participants listened to dense tone-cloud sequences in which a subset of tones repeated across cycles, with the proportion of repeating tones indicating coherence. Across two psychophysical tasks (repetition detection and sensorimotor synchronization), repetition coherence had strikingly similar effects: it shaped both the likelihood of object formation and the consistency of subsequent tracking. At high coherence, the emergence of the object and the attainment of a stable tracking phase occurred at the same number of cycles. The behavioural trajectory of synchronization performance, used to infer the dynamics of temporal integration, parallels prior neurophysiological findings in non-human primates, establishing a link between perception and neurophysiology. Moreover, we observed threshold shifts due to unit duration in the detection task, which indicates a dissociation between the effect of temporal scale on the likelihood and timing of object emergence: longer units when repeated reduced detection probability, but emergence timing remained constant. These findings demonstrate convergent behavioural dynamics shaped by bottom-up regularities, while intrinsic temporal integration windows constrain the process differently according to task demands. Overall, repetition coherence provides a principled framework for probing the processes underpinning auditory object formation, revealing common behavioural signatures alongside distinct sensitivities to temporal scale.

Matching journals

The top 3 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.