ABSTRACT
At this point, we might ask what exactly the problem of invariance in speech is. It seems to be that linguistic units for the description of sound are in· variant under vastly more powerful transformations than are the parameters we measure in physical descriptions of speecl1. Units like the phoneme are so abstract that stating their invariance transformations in physical terms seems almost impossible. Indeed perhaps the only linguists ever to try to do so arc Jakobson, Fant aud Halle, (1952). Instead, linguists have attempted to provide an inventory of perceptually defined units to which they tied their phonological units. What seems to be needed is a description of speech sounds that ( 1) provides explanatory principles for the properties of the linguistic units where possible; and (2) is abstract enough that the defining traits of a phonetic object can be satisfied by patterns in a space of articulatory parameters, acoustic parameters, or whatever other kinds of parameters seem useful.
