Synergistic yet dissociable roles of temporal and spectral predictions in auditory detection

Predictive processing enables the brain to anticipate both when and what sensory events will occur. Although temporal and feature-based predictions are both known to facilitate perception, their functional contributions and interaction remain poorly understood, particularly under natural listening conditions where predictions must be inferred from implicit statistical regularities rather than explicit cues or rhythms. Here, we orthogonalised temporal and spectral predictability in a non-rhythmic auditory detection task and used a signal-detection framework to dissociate perceptual and decisional processes. Temporal predictions primarily increased response readiness: predictable target timing led to faster responses and more hits but also more false alarms, consistent with a liberal shift in response criterion. Spectral predictions, in contrast, selectively enhanced perceptual sensitivity, mainly by reducing false alarms, without a comparable speeding of responses. Crucially, when both predictive dimensions were available, they interacted synergistically to maximise perceptual sensitivity, revealing how distinct predictive mechanisms combine to optimise behaviour. Furthermore, temporal and spectral variability were internalised differently: performance adapted strongly to the distribution of foreperiod durations but remained largely stable across the spectral distribution, suggesting that the brain encodes temporal and spectral statistics in fundamentally different ways when forming predictions under uncertainty. Together, these findings demonstrate that temporal and spectral predictions influence distinct computational components of auditory decision-making and are integrated synergistically to improve perception under uncertainty. The results provide a mechanistic framework of how multidimensional predictions jointly shape sensory processing in natural listening environments.