Some Problems with Binary Features for Tone

0. A fundamental assumption made in much modern linguistic research is that phonological contrasts are to be described by binary feature systems, an assumption which has encountered some of its greatest difficulties in the description of tonal systems. This article will be restricted to data from level-tone systems and will argue that the phonological representation of tone in binary features is inadequate for either synchronic or diachronic description. The discussion will begin with a review of the major feature frameworks for tone in use today, showing how each makes incorrect claims for some languages, while correctly accounting for phenomena in others. Then data from Igede and ProtoMixtecan will be discussed, showing binary features to b:e formally unworkable. Two criteria are necessary, though not sufficient, for any universal phonetic framework for describing the sounds of natural language. First, a phonetic theory of 1 An earlier version of this article was read at the Sixth Conference on African Linguistics, held at Columbus, Ohio in March 1975. I am grateful to John Goldsmith for his helpful comments.