The aim of this paper is to show how the investigation of a borderline and apparently minor area of phonetics may help to shed light on certain indistinct areas in phonological theory.

In his book An Introduction to the Pronunciation of English A. C. Gimson (1970) describes the articulation of the English plosives in different phonetic contexts. He demonstrates here (146–54 and 156–8) and elsewhere (Gimson, 1960) the articulatory instability of, in particular, the alveolar plosives /t/ and /d/ in English. As W. S. Allen (1960) points out, the predominance of instability associated with the alveolar plosives, rather than with the bilabials /p/ and /b/ or velars /k/ and /g/ may correlate with the more common occurrence of the alveolars rather than with articulatory or other kinds of constraints specific to consonantal clusters containing /t/ and /d/.

From the time of the invention of alphabetic writing until the end of the nineteenth century, students of language unquestioningly accepted, the view that speech consisted of sequences of discrete sounds, strung together like the beads of a rosary. Speech has been traditionally looked at ‘as consisting of segments, each one of which represents a phoneme, which are put together to build up speech.… Each segment is envisaged as a posture of the vocal organs, and these postures are joined together by means of glides, which take us from one to the next’ (Abercrombie, 1965: 121).

This paper is concerned with a phenomenon that has in previous descriptions been labelled ‘glottal reinforcement’, ‘pre-glottalization’ or ‘glottalization’ (for discussion of ‘glottal replacement’ and the distinction between ‘full’ and ‘pre-’ glottalization see below, 10.1). The essential characteristic of this glottalization is that the oral closure for /p/, /t/ or /k/ is preceded by a glottal closure, but there is much variation from one accent to another in terms both of the distribution of the glottal closure and of its articulatory characteristics (especially the duration of the closure). Of the literature on the subject, a speculative article by Christophersen (1952) followed by a reply from O'Connor (1952), a study by Andrésen eventually published as a book (1968) and a study by Higginbottom (1964) are the best known. I shall use the term ‘glottalization’ in this paper: being a more general label it does not commit one to the possibly misleading implications of the other names.

In a recent paper J. D. McClure (1972) objects to some basic principles of the IPA vowel system. According to him neither the division of the cardinal vowels into ‘primaries’ and ‘secondaries’, nor the mixing of rounded and unrounded in each of these two sets of vowels can be justified. It is obvious that McClure is perfectly right although he argues perhaps slightly more vigorously than needed. As a matter of fact, one wonders if anybody would wish to object to the points he is making.

1. In Lodge (1966), the data presented for the dialect of Stockport, Cheshire, did not show up the possible contrastive opposition in the vowels mentioned by Wells (1970: 238; in Wells' transcription, /e/: /ε̌/). It concerns the reflexes of ME ā, as in name, and ME e + gh (= voiceless palatal or velar fricative), as in eight, which have fallen together in RP. In the North of England, however, including part of the N.W. Midlands, the retension of [ç] and [x] longer than elsewhere in the country has caused differences in phoneme allocation and distribution. In the light of further evidence from Stockport and further experience of dialectal reflexes of Middle English, it seemed worthwhile re-assessing the phonemic interpretation of the Stockport data and comparing a few features of the phonology with older forms of speech from around the same area. We shall restrict our attention mainly to the vowel systems, with particular reference to the distinction mentioned above.

In the specimens of non-tone languages given in the current reprint (1967) of The Principles of the International Phonetic Association, prosodic features are rather inadequately dealt with; stress is usually mentioned, but intonation never. The North Indian languages presented there suffer in this way: stress in Bengali is discussed briefly, and some reading conventions are given, but for Hindustani (Urdu) and Oriya stress is simply described as ‘weak and variable’ and is not marked in the texts. Intonation is not discussed at all. This means, in effect, that a learner without direct assistance from a native speaker will be likely, if he attempts to read these specimens aloud, to use the intonation patterns of his mother tongue—to read with English intonation, or French intonation, and so on. We shall not be the first to observe that any attempt to pronounce a foreign language with unacceptable intonation patterns will not succeed, no matter how well articulated the consonant and vowel segments may be.

The origins of this Revised Phonetic Alphabet lie in the attempts of the Maritimes Dialect Survey to find an alphabet which would allow a computor to handle phonetic data readily and conveniently. This necessitated planning an alphabet which could be adapted to an 88-place ‘golf ball’ which could then be used with, for example, an IBM Selectric typewriter; that is to say, the proposed alphabet had to consist of not more than 88 ‘elements’. In fact, the Revised Phonetic Alphabet described herein consists of 85.

The specimen represents standard Latvian as spoken by educated native speakers in Riga. There are considerable differences between careful speech such as used in dictation, for example, and rapid speech such as used normally in conversation.

The following transcriptions were made from tape recordings of the speech of two natives of Greater Boston, reading ‘The North Wind and the Sun’ followed by a sentence composed to illustrate the distinctive New England distribution of low vowels. In each transcription, [aː] varies from cardinal to []; [ɑ] is about cardinal; [ɒː] varies from cardinal to ; the [o] in ‘o#x0259;, oɹ] ranges from to . Whereas [ɒː] and [oə] are kept quite distinct, as are [aː] and [ɑ], [ɑ] and [ɒː] are close together, overlap, and are not kept rigorously separate, [a] in [aɩ, aɷ] = [] or ; the first speaker uses a more advanced and the second speaker a more retracted variety. Diphthongization of [eɩ], [oɷ] is only slight, and the first element of [oɷ] is near cardinal and well-rounded. [œ] is a lightly-rounded , about like French [ə].