Enduring non-verbal memory for spectral timbre.

Abstract

It is often assumed that auditory sensory memory disappears within a few seconds or is so smudged as not to permit fine discriminations between successive sounds beyond a few seconds. But studies have been confounded by proactive interference, task difficulty, and verbal labeling. In Experiment 1, participants heard two complex tones separated by a silent retention interval of 2, 4, 8, 16, or 32 s. Tones varied widely in pitch from trial to trial (so no standard memory could be formed) and were non-speech-like (difficult to verbalize), and very long inter-trial intervals were used (32 s) to reduce interference from prior sounds. After the tone-pair presentation, listeners made a same-different response based on slight differences in spectral timbre. Discrimination performance expressed as d′, at each retention interval revealed that the persistence of memory for spectral timbre surpassed the usually quoted lifetime of some few seconds and for certain participants it endured for 32 s. In Experiment 2, overall stimulus levels were roved within trials to rule out loudness cues in the discrimination. It is proposed that reasonably robust non-verbal memory traces for spectral timbre persist for several seconds allowing listeners to make fine discriminations of changes to a spectrum.