This thesis presents a variationist account of phonological variation and change in the vowel system of Tyneside English. The distributions of the phonetic exponents of five vowel variables are assessed with respect to the social variables sex, age and social class. Using a corpus of conversational and word-list material, for which 32 speakers of Tyneside English were recorded, between 30 and 40 tokens per speaker of the variables (i), (u), (e), (o) and (3) were transcribed impressionistically and subclassified by following phonological context. The results of this analysis are significant on several counts. First, the speakers sampled appear to differentiate themselves within the speech community through the variable use of certain socially marked phonetic variants, which can be correlated with the sex, age and class variables. Secondly, the speakers style shift to a greater or lesser degree according to combinations of the three social factors, such that surface variability is reduced as a function of increased formality. Third, the overall pattern among the sample population seems to be one of increasing uniformity or convergence: it is speculated that social mobility among upper working- and lower-middle class groups may lead to accent levelling, whereby local speech forms are supplanted by supra-local or innovative intermediate ones. That is, the patterns observed here may be indicative of change in progress. Last, a comparison of the results for the (phonologically) paired variables (i u) and (e o) shows a strong tendency for Tyneside speakers to use these 'symmetrically', in that choice of variant in one variable predicts choice of variant in the other. It is suggested that the symmetry in the system is exploited by Tyneside speakers for the purposes of indicating social affiliation and identity, and is in this sense an extra sociolinguistic resource upon which speakers can draw. In addition, the variants of (3) are discussed with reference to the reported merger of this variable with (a); it is suggested that the apparent 'unmerging' of these two classes is unproblematic from a structural point of view, as the putative (3)—(o) merger appears never to have been completed.