Hearing discrete elements of a temporal sequence: Learning and generalization

Abstract

The ability to hear discrete elements of a temporal sequence was assessed. Subjects judged whether the first element of a three- or four-burst sequence of 50% duty cycle noise that alternated in level from low to high (or high to low) began with a low-level (60 dB SPL) or a high-level (80 dB SPL) burst. The rate was varied from trial-to-trial [selected values between 5 to 40 pulses per second (pps)] to prevent absolute duration and loudness cues. Initially, listeners were at chance performance for rates above roughly 20 pps. However, after 6 h of practice with feedback performance was near 90% for the highest tested rate (45 pps). After training with the stimulus that alternated in level, subjects were presented with a new condition; the three- or four-pulse sequences began with either a high-level or a low-level burst of noise, as before, but the level of each subsequent burst in the sequence was selected randomly between 60 and 80 dB SPL. Performance remained high for this condition, suggesting that listeners did not learn patterns of sequences but, rather, were able to hear discretely the first element in the sequence. [Work supported by NIH.]