Abstract Accurate detection of sound onsets and offsets is thought to be vital for speech perception. Sound‐onset event‐related potentials (ERPs) have been well characterised in EEG studies, but the characteristics of sound‐offset ERPs have often been obscured by temporal confounds in experimental design. Here two EEG studies were conducted in human listeners performing an interval duration discrimination task with (i) noise intervals in silence or (ii) silent intervals (gaps) in noise. Stimuli used in the active task were also presented under passive listening conditions with no participant response required. This design enabled us to investigate whether features of the offset ERP could be masked by task demands, and whether the usual temporal precedence of the onset event in a sound influences the relative magnitude and shape of onset and offset ERPs. The morphology of the offset ERP was distinct from that of the onset ERP in both noise and silent interval duration discrimination tasks, even though the functional roles of onsets and offsets as start‐interval versus end‐interval cues were reversed between the two tasks. This observation corroborates evidence from animal studies that there are fundamental differences in the brain mechanisms of onset and offset perception. In all experimental conditions, the amplitude of the offset ERP was one‐third to one‐half that of the onset ERP. Differences between active and passive listening conditions were largely explained by enhancement of ERPs for whichever transient marked the end of the discriminated intervals. However, this context dependence emerged in earlier ERP waves for offset than onset responses. image Key points Onset and offset event‐related potentials (ERPs) differ in morphology and amplitude. For both onset and offset ERPs, P200 and later waves depend on behavioural context. For offset ERPs only, P100/N100 waves also depend on behavioural context. Engagement in duration discrimination alters ERPs primarily at the end of intervals.
Ali et al. (2026) studied this question.