Phonetic typesetting has always been a difficult undertaking. As long as typesetting was the business of professional typesetters, the ordinary transcriptionist did not have to worry very much about this. Today, typesetting techniques are downgraded by a variety of other textprocessing machines, ranging from electrostatic printers, or other non-impact printers that use digitized font information, to impact printers, such as print-wheel and selectric typewriters (see Hsu (1982)). It is the output of this kind of printer that can be used for camera-ready copy. Moreover, the area of influence of the author is extended since he himself has access to such textprocessing machines (see the TEX user group movement, explained below). Some years ago, most of the text-processing machines were quite inflexible; they could not cope with additional symbols or additional fonts. Early attempts to make phonological and phonetic data representable in such textprocessing systems had to accept existing symbols, and mappings between existing standard symbols (the ASCII code, for example) had to be established, sometimes on the basis of similarity (& for epsilon, for example) (see Millar/Oasa (1981), Wothke (1981), Roach (personal communication, 1981). Now that it is possible to define character sets for matrix representation (Pfeiffer-Rupp (1981a), (1982a), Neuhaus (1983)), both for printer representation and for representation on a visual display (Pfeiffer-Rupp (1981b)), it is no longer necessary to operate with such makeshift devices—which demand an additional level of coding within the same alphabet and often lead to the idea that the ASCII code is adequate for everybody's purposes, which is what the graphics industry would like to hear from us.

Like many vocalic variables which have been described for other varieties of English, Durham (eː) is correlated with a range of factors, some linguistic, others non-linguistic; in this paper, some of these correlations are presented, (eː) is one of three ‘long’ or ‘tense’ vowels which are variably realized as pure monophthongs or as opening or centring diphthongs; it is represented by words of the lexical set gate, wait, day, etc. The other two vowels are (oː), which occurs in words such as bone, goat and toe, and (ɔː), which occurs in torn, shore, etc., and variably in such items as law, walk and all (in which it alternates with /ɑː/). These three vowels probably show rather similar patterns of variation; however, in this study only (eː) will be considered.

It is well known that Hong Kong has a floating population consisting mostly of fishermen. These fisherfolk call themselves the ‘boat people’ (shuishang ren literally, ‘people on water’), and upon being asked their native county, they would usually reply, “We're boat people” instead of naming a particular county in the Guangdong Province as is the usual custom. The boat people, of course, are not a homogeneous group as far as speech is concerned. The informants interviewed for the present study came from the Pu Tai Island (a small island to the southeast of Hong Kong). The children went to school on Pu Tai, while the grownups went out to sea to work for their living. About eight years ago the family gave up their boats and settled on land. The speech of this family and that of some friends who were also formerly boat people exhibit certain phonetic characteristics which are interesting departures from Modern Standard Cantonese (hereafter referred to as MSC), the dialect spoken by the majority of people in Guangzhou, Hong Kong and Macau.

Until comparatively recently, the diphthong systems of southern English accents have appeared virtually symmetrical, and such an ideally symmetrical system for the closing and centring diphthongs is shown in Figures 1 and 2. (The diphthong /ɔI/, which of course does upset the symmetry, is omitted for the sake of exposition. Historically too this omission is well-motivated, as this diphthong was a borrowing from Norman French (Gimson, 1962: 132), and still has a limited distribution (Gimson, 1962: 148).)