Memory for repeated auditory textures

Even though memory plays a pervasive role in perception, the nature of the memory traces left by past sounds is still largely mysterious. Here, we probed the memory for natural auditory textures. For such stochastic sounds, two types of representations have been put forward: a representation based on temporally-local features, or a representation based on time-averaged summary statistics. We synthesized naturalistic sound textures and used them in a “memory for noise” paradigm based on repetition. On the one hand, if the brain represented texture exemplars as time-averaged summary statistics, then repetition detection and learning should all but vanish for longer exemplar durations. On the other hand, if the learning of noise and textures relied on similar processes, presumably involving temporally-local feature sets, then results should be similar for white noise and natural textures. Results replicated previous findings, showing an effect of duration on baseline repetition detection performance but little to no effect on learning afforded by repeated exposure. Thus, naturalistic texture exemplars are amenable to learning when repeated exposure is available. This finding is consistent with two interpretations: the existence of a special processing mode when acoustic repetition is involved, to which textures are not immune; or a convergence of the feature set versus summary statistics description of sound representations, if a continuum of time scales is considered.