Introduction (continued)

The translation of the phonetic system

The system of phonetic transcription devised by Sir James Murray for use in the first edition and followed, for the sake of consistency, in the Supplement is a subtle and flexible means of recording English pronunciation. But many of the effects for which Murray strove in the design of his system were realized, not long afterwards, in the International Phonetic Alphabet (IPA). It seems very possible that, if the IPA had already achieved full development and widespread acceptance at the time that Murray was beginning to work on the Dictionary, he would have adopted it instead of a system of his own. It was therefore logical to consider replacing Murray's system with IPA throughout the Dictionary. IPA has the advantage that it is very widely accepted and understood, and can be used to represent the sounds as well of regional and dialect English and foreign languages as of standard English. Indeed its introduction was regarded by many whom the project team consulted as among the highest priorities. It was decided that the change should be made for the present edition, rather than left until a future revision phase. But for reasons of historical interest, the Murray transcriptions have been retained in the electronic version alongside the IPA ones, the latter only having been printed.

A notable feature of the Dictionary is that obsolete main words, derivatives, and certain variant forms and combinations are not given phonetic transcriptions but have their stress-pattern indicated: the stress-dots (which are placed after the accented vowel, just as they are within the transcriptions) are printed within the body of the word or form. Naturally consistency required that the stress-dots within these forms should also be altered to IPA stress-marks (which are placed before the beginning of the accented syllable).

The short time available meant that there had to be rather strict limitations on the extent of the changes made. For native English words the variety of pronunciation represented is, broadly speaking, educated standard southern British, or 'Received Pronunciation' (RP). There could be no question, at this stage, of systematically registering non-RP (i.e. other British and non-British) pronunciations; although of course those already included in the Dictionary have remained and have been augmented by a few analogous cases. Again, pronunciation variants within RP, apart from those already registered in the Dictionary, could not be researched and documented, except as the evident need to include them arose in the course of the other editing. The adding of up-to-date pronunciations, except when prescribed by the Supplement, had to be kept to a minimum, for the same reasons. Essentially, a straightforward literal translation from the Murray system to IPA has been attempted, accompanied by correction of the errors inevitably arising from that process.

The method by which the translation of the Murray system to IPA was performed can be briefly summarized. As much of the translation as possible was carried out automatically by computer. The computer had to operate on two kinds of material: sequences of phonetic characters, and normal English words containing dots indicating the position of the stress accent. First, the program identified the strings of characters which it was required to process. It used the mark-up tags to do this. Then the actual translation was run. The program referred to three tables, one giving IPA equivalents for Murray's phonetic symbols (or groups of them); one giving rules that showed whether each IPA symbol is a vowel, a consonant, or a consonant cluster; and one giving similar rules for ordinary English spelling. The second and third table enabled the computer to deduce the correct position for an IPA stress-mark from the existing stress-dot in the text. Certain characters and groups were known in advance to have more than one possible IPA equivalent. These were printed out on a special report, as were all pronunciations in which the conversion failed; and from these, corrections were made at the keyboard. Altogether, 137,152 phonetic transcriptions and 137,274 stress-marked words were automatically translated.

A considerable number of phonetic transcriptions are found in other parts of an entry than the pronunciation key. These were not originally identified by the computer, and so a separate program was run in order to register them on a list for editorial transcription.

The conversion of Murray's phonetic system to IPA was not entirely straightforward. The peculiar nature of the system, the way in which it is applied both in general and in particular cases, and the historical variety of English speech which it was employed to record, all presented obstacles to the smooth application of the scheme of translation.

The peculiarities of Murray's phonetic alphabet can be ascribed to the time and circumstances of its devising.[1] It must here suffice to enumerate a few of its characteristics. It is used indiscriminately both for phonetic and phonemic transcription: that is, it is a set of symbols employed to represent both the members of the particular set of sounds that constitute the phonology of English, and the much larger set of infinitesimally differing sounds from all other phonologies to which reference is made in the Dictionary. As a means of representing standard English pronunciation, Murray's system is sensitive and generally lucid. It is less well-adapted for the transcription of dialectal and foreign words. Shift had often to be made with the limited range of typographical shapes that Murray had devised at the outset in order to transcribe unexpected foreign sounds. Symbols not listed in the original pronunciation key appeared during the progress of the Dictionary. Uncommon foreign sounds were handled differently on widely separated occasions.

With regard to consonants, there are few drawbacks to Murray's system, apart from the dearth of symbols for foreign sounds. Indeed, all but one of the primary consonant symbols correspond exactly to the set employed for English in the IPA. A fundamental feature of the vowel system that does not translate easily into IPA is its analysis into 'ordinary', 'long', and 'obscure' vowels. Broadly, the same list of symbols appears under each heading, but (respectively) without a diacritical accent, accented with a macron (long mark), and accented with a breve (short mark). The implication is that the same underlying vowel manifests itself in three guises; and, further, that these guises are determined at least partly by its relationship to the stress accent of the word. The theory underlying the obscure vowels is that, if they were accorded stress, as they sometimes are in song or very careful enunciation, they would resemble their ordinary equivalents. This principle, though something like it is encountered in phonological theories, is not usually embodied in the standard IPA transcriptions used in dictionaries. If, as has here been done, it is set aside, then most of the obscure vowels are equivalent to the IPA ('schwa'; the sound of a in sofa, particular); while a few are translated into (the sound of e in hatchet). The main peculiarity of the long vowel series is that diphthongs are included in it; but this is an oddity of phonetic theory - the analysis of English diphthongs as long vowels followed by 'glides' - rather than an obstacle to translation.

The major problem arises in the 'ordinary' vowel series, which contains both vowels universally recognized as lacking length - the i of pin, for example - and vowels now normally transcribed as long vowels. These, in fact, are the members of the long series in another guise: a guise apparently taken on in a syllable under low stress. So, for example, the vowel of the second syllable of Matthew is an ordinary vowel, while that of few is long. Most classifications would treat them as identical. This would not pose a major problem of translation, if it were not that the same symbols employed for the second vowel of Matthew and its parallels throughout the vowel system (hero, Psyche, etc.) are also used to represent the principal vowels of the European languages - roughly speaking, the so-called 'cardinal vowels'. The two sets of vowels are not phonetically identical, or even close, and were not in Murray's time. Such symbols as these (along with some others) were straightforwardly ambiguous, and could not be translated correctly in every case by the computer.

Because Murray's system uses many of the same symbols for English and foreign vowel sounds, it was necessary to depart from certain conventions that are common in the transcription adopted by English dictionaries. A notable example is the use of () for the vowel of bed in place of the more usual (e), because the English vowel is, for most English ears, closer to the cardinal open vowel of French faire or German Bär than to the cardinal narrow vowel of French bébé or German Schnee.

The variety of English pronunciation recorded by the Murray transcription is extremely 'precise', conservative, and (in present-day terms) old-fashioned. Most of its peculiar characteristics are systematic (they permeate the phonology) rather than occasional (features of the pronunciation of particular words, or groups of words). They are systematic enough for a phonetician to predict the way in which individual words will be transcribed, but not enough to make it easy for a computer to efface them all automatically. This is one reason why they have largely been retained in this edition. A second reason is that they constitute a useful record of one variety of English pronunciation in a particular period; and a third is that, for the general user, most of them are merely small nuances for which one can make allowance.

The chief phonological features that set the pronunciation represented in the Dictionary apart from most present-day phonetic descriptions of English are the following:

1. The stressed close vowels (i, u) are maintained in words like idea, realize, museum, skua, in contrast with the diphthongs (, ) of dear, rear, secure.
2. The close vowels (i, u) are maintained in unstressed syllables in words like delineate, creation, perpetual, graduate, in contrast with the open vowels (, ) of genius, demonic, circulate.

   It is in such syllables that the Murray transcription uses 'ordinary' vowels, not long ones. It is quite likely, therefore, that Murray was aware of no distinction in duration between these vowels, but it is also clear from the consistency with which the distinction is observed that he was aware of, or believed in, a distinction of vowel quality (specifically, of tension). The conventional IPA transcription, however, observes only a phonemic distinction between open vowels, which are treated as 'short', and narrow vowels, which are accorded symbols of length as well. The IPA transcriptions therefore convey the misleading implication that such vowels have, or had, greater duration.

3. Diphthongs (e.g., ) are maintained in unstressed syllables in words like homographic, protocol, in contrast with the obscure vowel 'schwa' in words like homonym, melody.
4. Unstressed medial obscure vowels are not represented as having been elided, e.g. in veterinary.
5. A distinction is maintained between syllabic consonants and a sequence of obscure vowel and consonant, so that words like principle, and principal are distinguished.
6. A distinction is drawn between a diphthong () in words like glory, boarder, mourning, and a long vowel () in words like saurian, border, and morning. This is not generally maintained in present-day IPA.
7. The nineteenth-century lengthening of the rounded low back vowel of words like soft, cloth, and cross is recorded (in this edition, as a separate variant; in the first edition, by a special symbol indicating the possibility of either pronunciation).