Hearing out repeating elements in randomly varying multitone sequences: a case of streaming?: From Sensory Processing to Perception

One of the most essential functions of perception is to detect and track signals of interest amid other, potentially distracting stimuli. Understanding the mechanisms behind this function is important not just from a theoretical perspective, but also because of its potential applications in artificial scene analysis. An important aspect of auditory scene analysis is the organization of sounds into streams (Bregman 1990). While the vast majority of studies on auditory streaming have used repeating sound sequences, perhaps one of the most informative findings regarding the potential usefulness of streaming in everyday life comes from a study that involved randomly varying tones. Kidd, Mason, Deliwala, Woods, and Colburn (1994) measured detection thresholds for a fixed-frequency target tone presented simultaneously with multiple other spectral components, the frequencies of which varied unpredictably across trials; a situation known to produce large amounts of informational masking (Neff and Green 1987). Kidd et al. found that if the target-plus-masker bursts were repeated identically several times, so that they formed an unchanging sequence, large informational masking effects remained. However, if the masker component frequencies were allowed to vary from one burst to the next within the course of the stimulus sequence, as illustrated schematically in Fig. 1, thresholds were substantially reduced. Kidd et al. explained this outcome as being due to the repeating target tones being perceptually organized into a coherent auditory stream, which stood out from the randomly varying background. From this point of view, Kidd et al.’s results appear to demonstrate that stream segregation can greatly alleviate informational masking effects, which might otherwise Fig. 1. Schematic spectro gram showing a sequence of regularly repeating tones amid random masker tones. Time F re qu en cy