Published online by Cambridge University Press: 22 April 2026
The analysis of tone phonemes, or tonemes, like that of segmental phonemes, necessarily involves considerations of phonetic similarity—necessarily, because there is no other basis upon which tones occurring in noncontrastive distribution can be phonemically identified with one another. But how is ‘phonetic similarity’ to be understood in the case of tones? Not, certainly, in the sense in which it is understood in the case of segmental phones, where it denotes some objective similarity in the articulatory or acoustic characteristics of the phones themselves. To be sure, even in this case phonetic similarity is a rather elastic criterion; but at least it can be shown, for example, that two vocalic phones have a common tongue position, or two consonantal phones certain common distinctive features. No such objective similarity is demonstrable in the case of tones—at least, of level tones; for, since it is relative rather than absolute pitch that is involved in tonemic systems, a given toneme may occur at virtually any pitch of which the voice is capable. Variants of a man's high tonemes may be acoustically similar to those of a woman's low tonemes; and in the speech of a single speaker a given level of pitch in two utterances or even within one utterance may represent variants of several different tonemes. The possibilities of phonemic overlapping of level tonemes are virtually unlimited.
1 The tonemic identification of the initial tones of language B poses a problem. While noninitial tones share either an echo or a change relation to the preceding tonal environment, initial tones exhibit neither relation. How, then, are initial tones to be phonemically dentified with noninitial tones? Are initial low tone and noninitial echo tone to be considered variants of one toneme, and initial high tone and noninitial change tone of another, or vice versa? Phonology can offer no answer to this question. Methodologically, either solution is arbitrary; tactically, either is satisfactory. If language B were a real language, one would choose the solution that promised the simplest morphological statement.
2 Initial variants of language C can still be identified with noninitial variants on the basis of the fixed contrastive relation, since this relation is implicit in the other.
3 This statement is something of a simplification. Actually, if tone a (or a series of tones beginning with a and not including z) is followed by tone z, and this z is in turn followed by another a, the pitch of the second a is lower than that of the first. Tone z, in such a sequence as a-z-a, operates as a kind of register-shifter, such that any tones that follow it are, as it were, in a lower key. The contrastive possibilities, however, are unaffected by the change of key, and the second a in the sequence a-z-a is identifiable as such by virtue of being the only tone other than z that occurs in its environment.
4 Other tone sequences : Mede me nnamfo no baa me fi ‘I brought my friends to my house‘
(z -z -a -a-z -z -z -a -a -z-a -z)
Me-de-me-n-na-m-fo-no-ba-a-me-fi.
Da bi Fida mekəə Nkran ‘One Friday I went to Accra‘
(a -a -b -b -z -a -z-z -z -z z)
Da-bi-Fi-da-me-kə-ə-N -kr-a-n.
Akutu yi ahe ahe? ‘How much each are these oranges?‘
(z -z -a -a -a-b -b-c)
A -ku-tu-yi-a-he-a-he?
Dənkoro apa ho du ‘It's ten past one‘
(a -a-b -b -b-c -c -d)
Də-n-ko-ro-a-pa-ho-du.