The perceptual grouping of sequential sounds can be studied using sequences of alternating ‘A' and ‘B' tones that differ in one or more acoustic dimension. A and B tones may be grouped together as a single perceptual stream (‘integration') or segregate into two perceptual streams. We examined the factors that influence segregation when tones subsets differ in interaural timing cues.

In Experiment 1, listeners (n=12) heard sequences comprising 5×ABA– repetitions (‘–‘ = extended silence), and reported whether integration or segregation was heard at sequence offset (2AFC). A & B were 80-ms pure tones, with a 40-ms silent inter-stimulus interval (ISI) between each tone. The extended silence (“–“) was 160 ms. The amplitude envelope shape was varied across trials, but was common for A and B tones within a trial. One envelope comprised 10-ms ramps at onset and offset, and was constant for the intervening 60-ms. The asymmetrical fast attack-slow release (FA-SR) envelope comprised a 10-ms onset ramp and 70-ms offset ramp. The slow attack-fast release (SA-FR) envelope was the reverse of this configuration. In a first task, A and B differed in frequency (∆f), and envelope shape did not affect segregation. In a second task, A and B differed in interaural phase difference (∆IPD), and significantly more segregation was heard in the FA-SR envelope conditions.

Experiment 2 (n=12) tested only the FA-SR envelope, and the ISI was varied from 0-60 ms. In the ∆f-only task, segregation decreased with increasing ISI, but in the ∆IPD-only task, segregation increased with ISI. The influence of envelope shape and ISI on ∆IPD streaming suggest that the temporal integration of binaural cues across A and B tones may reduce stream segregation. The data imply an integration window of between 40-200 ms, within the range of previous estimates. In a third task, A and B tones differed in either frequency, IPD, or both. Contrasting with the ∆IPD-only task, ∆IPD-based segregation was not observed in this mixed-cue task. This suggests ∆IPD segregation is highly sensitive to across-trial contextual effects, potentially accounting for differing findings in previous studies.

In Experiment 3, listeners attempted to hear integration to detect a delay imposed on the final four B tones of an 8×ABA- sequence (an adaptive 2I-2AFC procedure). FA-SR tones were tested (ISI = 60-ms, maximum B delay = 40-ms). Preliminary data (n=6) suggest performance worsened with increasing ∆f, but ∆IPD had only a marginal impact on overall performance. These findings are consistent with previous studies. However, ∆IPD elevated thresholds substantially on the first repeat of each condition, suggesting that ∆IPD may have initially promoted segregation, but listeners learned to ‘ignore' the cue on later repeats.

These experiments suggest that whilst ∆f may be a more robust cue for segregation than ∆IPD overall, each cue appears differently influenced by additional factors, such as ISI or amplitude envelope (Experiments 1 & 2), contextual effects (Experiment 2), and potentially, the time-course of learning effects (Experiment 3). Therefore, stimuli that are optimized for ∆f-based segregation may be suboptimal to observe ∆IPD-based segregation.