“…According to this theory, speech perception relies on there being at least a brief period during each speech sound when its short-time spectrum is reliably distinct from those of other speech sounds. For an initial stop in a stressed syllable, for example, this period includes the burst and the first 10 ms. after the onset of voicing (Stevens & Blumstein, 1978). That a listener is nevertheless able to identify speech sounds from which these invariant attributes have been removed is explained by the claim that, in natural speech, they are sometimes missing or distorted, so that the child must learn to make use of secondary, contextconditioned attributes, such as formant transitions, which ordinarily co-occur with the primary, invariant attributes (Cole & Scott, 1974).…”