Although there is a high degree of variability here, analysis of Wood’s (1987: 125ff) and my own data reveals that realizations are typically in the region of [ɐu] and [Λu] , for all speakers; other realizations observed include [əu], [çu] , [ɒu] fairly frequently, and less often [ou], [au]. There is also some evidence for a Canadian Raising-type distribution, with back onsets tending to occur in pre-fortis environments and non-back onsets elsewhere. Despite the fact that Wood records some cases of offset weakening, e.g. [ɔ^ʊ] , [o^ʊ] , and Hastings (1979, quoted in Wood 1987: 111) maintains that diphthong weakening is typical of CFE, the available data would suggest that in fact such weakening occurs especially (a) before /l/ – where, in common with SAfE, onset quality is also affected, typically yielding [ɒ'] (e.g. ) – and (b) in unstressed position. In fact, as with MOUTH, offsets of GOAT are typically markedly peripheral (i.e. strongly backed and rounded) rather than weakened. Yet again, Lanham (1982: 343) maintains that this ‘high diphthongal glide’ is characteristic of Afrikaans-influenced English generally, and used even by well-educated speakers – as confirmed in Wood’s (1987: 137–138) and my own data. It is particularly noticeable in word-final position. In hiatus (as in the subset GOING), the offset is typically realized as [w] (e.g. ['gɐuwiŋ]). Wood (1987: 137–138) observes that both onset lowering and markedly backed and rounded offsets, are typical across the social scale.