Do Tones Have Features?

Unless explicitly concerned with developing a system of features per se, most studies of tonal phonology refer to contrasting high, mid, low and contour tones as H, M, L, HL, LH (etc.), or with integers, rather than with features such as [±UPPER], [±RAISED]. Since this practice stands in marked contrast to vowel and consonant phonology, where features seem unavoidable, it is natural to ask whether this difference in practice is due simply to convenience or whether tones lend themselves less naturally to a featural interpretation than vowels and consonants. In this paper I suggest that this is indeed the case: while features sometimes facilitate a general and insightful account, there are inconsistencies, indeterminacies, and other reasons to doubt the value of tonal features (and tonal geometry). This then naturally leads to a more general question: Why should tone be different? In Hyman (in press), I provide evidence to suggest that tone is different in its capabilities: tone can do everything that segmental and accentual phonology can do, but the reverse is not true. I start by illustrating some examples to make this point then turn to the question of how this provides insight into the relative unimportance of featural analyses of tone. In this paper I also raise the question of why tone, which might seem like a good bet, is not a linguistic universal (as compared to consonants and vowels). I suggest that it is the relative autonomy and “non-integration ” of tone that accounts for its versatile and unique properties. Since some tonal phenomena have no segmental or stress analogues, I argue that anyone who is interested in the outer limits of what is possible in phonology would be well-served to understand how tone systems work. 1. Introduction: Thre