ABSTRACT
In the early days of distinctive feature theory, features were regarded both as attributes of phonemes and as properties referring to a single acoustic scale (Jakobson, Fant & Halle, 1951). Later work continued this notion of regarding features as a bridge between abstract phonological entities and physical facts (Chomsky & Halle, 1968a). Each feature was defined in articulatory terms, but it was claimed that this was merely a matter of expository convenience, and the definitions could equally well have been made in acoustic terms. In these distinctive-feature theories, each feature was regarded as being, at the same time, an element in phonological theory and a single physical scale. This made it possible to describe contrasts in binary terms and phonetic differences between languages as different values along a particular physical parameter. Some phonologists, for example McCawley (1968) and Postal (1968), recognized the need to incorporate feature interpretation rules as part of the grammar, but they were extremely vague when it came to working out the details and specifying actual values on any of these physical scales. Ladefoged (1975) is the only concrete proposal of a complete set of feature values for a number of segments in a systematic phonetic transcription (and, as we will show, this proposal is wrong). Just as phonologists were vague in their phonetic specifications, so equally, phoneticians who devised their own feature systems (e.g. Fant, 1971, 1983), did not make any attempt to show how specifications in terms of these features could be related to phonological theories of a language.
