The third need is for better theory. We should expect that general phonologic theory should be as adequate for tone as for consonants and vowels, but it has not been. This can only be for one of two reasons: either the two are quite different and will require totally different theory (and hence techniques), or our existing theories are insufficiently general. If, as I suspect, the problem is largely of the second sort, then development of a theory better able to handle tone will result automatically in better theory for all phonologic subsystems.

One issue that must be faced is the relative difficulty of analysis of different phonologic subsystems. Since tone systems typically comprise fewer units than either consonant or vowel systems, we might expect that they would be the easiest part of a phonologic analysis. Actual practice does not often work out this way. Tone systems are certainly more complex than the number of units would suggest, and often analytically more difficult than much larger consonantal systems.

Welmers has suggested one explanation. Tone languages use for linguistic contrasts speech parameters which also function heavily in nonlinguistic use. This may both divert the attention of the uninitiate and cause confusion for the more knowledgeable. The problem is to disentangle the linguistic features of pitch from the co-occurring nonlinguistic features. Of course, something of the same sort occurs with other sectors of the phonology: consonantal articulations have both a linguistic and an individual component. But in general the individual variation is a small thing added onto basic linguistic features of greater magnitude. With tone, individual differences may be greater than the linguistic contrasts which are superimposed on them. Pitch differences from one speaker to another, or from one emotional state to another, may far exceed the small differences between tones. However, any such suggestion accounts for only some of the difficulties in hearing tone, or in developing a realistic attitude about tone, but not for the analytic difficulties that occur even when tone is meticulously recorded.