Subjects were required to determine temporal order in a continuously repeated sequence of four vowels (/i/, /ɛ/, /a/, /u/). All vowel samples were spliced from the enunciations of one speaker and were controlled for pitch (125 Hz) and true rms intensity. The duration of each vowel segment, T, was varied from 300 to 75 msec; vowel order was randomized for each value of T. In identifying temporal order, listeners achieved near perfect scores for T⩾125 msec; for T⩽100 msec, scores were near chance. Results are compared with those for a similar experiment employing nonspeech sounds.