Consider the relations of the nuclei of the low diphthongs /y, w/ to the low vowel //, in the above figures. /y/ is raised and to the front of //, while /w/ is raised and to the back, for all these speakers. How shall we interpret these relations? It appears that there are three distinct phonetic specifications for the three vowel nuclei. This is unlike the pattern of Reference American, where the nuclei of all three vowels are phonologically identical, that is, all have the same low, back, unrounded nucleus, and where no phonetic rules are postulated to distinguish between them.
An attempt may be made to attribute these differences to natural phonetic effects of the offglides, thereby allowing the same unitary treatment of the three nuclei as in Reference American. For example, one might suppose that there is a process of nucleus-glide coarticulation or assimilation, by which the opening gesture associated with the nucleus is coarticulated with the closing gesture associated with the offglide. This would explain why /y/ is raised to the front and /w/ is raised to the back: /-y/ is high and front and /-w/ is high and back, relative to the // nucleus.
However, if this supposition were true, it would seem to weaken explanations of vocalic sound changes which rely on an explanatory process of nucleus-glide differentiation. If the process of nucleus-glide differentiation ``explains'' some kinds of variation, and nucleus-glide assimilation ``explains'' others, it would seem that there remains only description, and no explanation at all. On the other hand, if the two processes were restricted to complementary environments, we could maintain the explanatory value of both, within their restricted contexts of occurrence. For example, if assimilation were typical of unstressed environments, while differentiation was typical of stressed environments, then appealing to both processes in different cases remains possible and has explanatory value.
Unfortunately, the effects of stress in the current case do not bear out this explanation. If the three nuclei overlapped when stressed, and separated only when unstressed, then the explanation would be supported. Instead, /y, / rise and come together when unstressed, while /w/ fronts insignificantly (cf. Figure ). The facts are opposite to the prediction, so the explanation of nucleus-glide assimilation must be abandoned, until some other explanation can be found for these signicant phonetic differences. I therefore I conclude that the three vowels /y, , w/ indeed have three different phonetic nuclei specified in the phonetic grammar.