In The Sound Pattern of English, Chomsky and Halle's approach to structuring the 3X2 grid began with the observation that, in preparation for speaking, as opposed to breathing or eating, the vocal tract assumes a certain ``neutral position.'' They point out previous claims that this position, the ``base of articulation'' of Sievers (1901), varies across languages, and claim that for English, it is mid and front, as in . SPE proceeds to define the 3X2 vowel grid on the basis of movements of the tongue body away from this position: up [high], down [low] and back [back]. These dimensions have been considered to be fundamental and universal phonological dimensions for vowel quality, despite the clear implication that the neutral position, and thus the movements of the tongue away from the neutral position, is language-particular.
Not much research has followed upon these observations. However, a major empirical result of the phonetic investigations in later chapters would seem to be quite relevant, namely, the pattern of phonetic effects of phrasal stress on vowel quality. In most dialects, where a clear pattern is evident, the effect of stress is a graded shift of the phonetic quality of the nucleus of unstressed vowels from the quality of stressed nuclei towards a single ``reduction target'', which differs from dialect to dialect. This reduction target may be identical to the average location of //, the vowel which is phonologically unspecified for quality. These facts are consistent with the identification of SPE's ``neutral position'', Sievers' ``base of articulation'', the phonetic realization of // (which is, in Reference American, the unstressed allophone of //), and the ``reduction target''.
Phonological vowel quality features may be derived, as in SPE, by their opposition to the qualities of the reduction target. Thus the phonologically unspecified or unmarked height is that phonetic height which unstressed vowels reduce towards, so that in dialects where unstressed vowels shift towards high-central, the unspecified or default vowel height should be phonetically high (and the unspecified level on the front/back scale should be central). The three American English dialects studied here, Chicago White English, Alabama English, and L.A. Chicano English have a high (central or front) reduction target. (A mid-central reduction target occurs only in Jamaican Creole.) These facts suggest that the neutral position for English dialects is not mid-front.
Either the phonetic foundation for the vowel-quality features, [high], [low], and [back], is to be abandoned, and they are to be stipulated without this justification, or a different set of features may be established - indeed, feature sets, which differ from language to language. For example, if a language had a high-back neutral position as the base of articulation from which sounds deviate, then the marked features [front], and [mid] and [low] or some equivalent formulation, would be indicated.