A rapid. repeating cycle of alternating high and low tones was presented under three conditions. In the "discrete" condition. transitions between tones were abrupt: in the "ramped" condition. successive tones were connected by frequency glides. In the "semiramped" condition. there were partial glides in frequency (as in speech). "Discrete" sequences were most likely to split perceptually into high and low streams, making order discriminations difficult. The "ramped" condition was least likely to split, and order perception was easiest. Results for the "semiramped" condition were intermediate. The discussion relates these findings to the acoustic properties of speech and to the process of auditory stream formation.Not long ago, Warren, Obusek, Fanner, and Warren (I969) reported what they perceived as a remarkable inability of unpracticed human Ss to make a judgment of order. The stimulus was a repeating cycle of four sounds: hiss, buzz, sine tone, and vowel, each lasting 200 msec, Few SS could name the order of the sounds.This struck them as puzzling in view of the already known capacity of the auditory system to discriminate the order of sounds at a much higher rate. For example, in normal English speech, phonemes occur more quickly-Bu-l Ou msec per phoneme (Efron, 1963). Speech can be reported correctly at rates as fast as 30 msec per phoneme (Foulke & Sticht, 1969). Winckel (1967) reports that temporal order of musical notes is resolvable down to about 50 msec per note. Warren et al (1969) found that the order of short repeating sequences of spoken digits was reported more accurately than unrelated sounds presented at the same rate, and concluded that verbal sounds were related in some fashion that permitted more rapid perceptual following. There is some confirmation for this in the work of Thomas, Hill, Carroll, and Garcia (1970), who found that the order of a repeating cycle of four vowel segments spliced together could be correctly identified at 125 msec per segment (but not at 100 msec per segment). Yet, even this rate is not as fast as we can discriminate the order of sounds in speech. Bregman and Campbell (1971) have proposed that the difficulty experienced by Ss in experiments using repeating cycles of sounds is due to the fact that subsets of the sounds group into separate perceptual streams and