Statistical change detection requires predicting stimulus statistics rather than sequence events

Evidence

Detecting relevant changes in auditory textures is hence complicated due to the lack of deterministic spectrotemporal predictability, and instead relies on the prediction of stimulus statistics (and their dynamics). The aim of this study is to determine how changes in stimulus statistics are detected under continuous presentation of stimulus. Concretely, we assessed how first-order non-uniform sound statistics are integrated and how listeners disentangle a change in these from the intrinsic stochastic variations of the stimulus. We designed a broadband stimulus, composed of tone-pips, whose occurrence is only constrained by a frequency marginal probability. — Dijk, Pim van, Physiology, Psychoacoustics and…, 3286

Categorization

• DEFINE: Introduces a minimal paradigm for studying statistical hearing: tone-pips constrained only by a frequency marginal probability, clean of other acoustic structure.
• QUESTION: Worth chasing: how does the brain carry ‘marginal probability’ estimates over continuous listening, and what neural time-constant governs the integration window they report?