Limits of temporal discrimination for sequences of speech sounds

Earlier studies from a number of laboratories found that the order of vowels in iterated sequences of three or four items could not be identified below durations of about 100 ms/item. The present study required untrained subjects to distinguish between iterated sequences of three vowels (/i/, /a/, /u/) presented in the same or in permuted order. As would be anticipated, distinguishing between permuted orders could be accomplished through naming of individual items at durations greater than 100 ms, but contrary to some theories, metathesis resulting in perceptual equivalency of different orders did not occur at item durations below 100 ms. Rather, permuted orders were readily distinguishable down to durations corresponding to single glottal pulses, although component vowels could not be recognized. These results are consistent with theories considering that comprehension of speech does not require the ability to detect component speech sounds and their orders, but is based instead on recognition of unresolved temporal patterns. [Work supported by NIH and NSF.]