ABSTRACT
Any invariance that exists in speech production targets in normal speech appears to be governed to a large extent by perceptual criteria applied during language acquisition, based on the properties of the acoustic signal as analyzed by the auditory system. For this to be possible there must clearly be some form of perceptual equivalence of men, women and children producing the same phonetic quality, in spite of the manifestly different acoustic signals in these cases. I assume that there is some fairly systematic transformation mapping phonetically equivalent acoustic descriptions between speaker types, and that this transformation might be found cither by studying analyses of real speech, or by applying experimental transformations to parametric synthesis data and making subjective judgments about their adequacy.
