This is a very slightly adapted, annotated and shortened copy of a contribution to the April 1993 Volume 47/2 issue of the English Language Teaching Journal published by the Oxford University Press
The Longman Pronunciation Dictionary (LPD) is a completely new work, avowedly EFL oriented, different in many ways from the EPD (the Daniel Jones et al English Pronouncing Dictionary). It is larger and heavier, offering 75,000 entries against EPD's 'more than 60,000 headwords'. A hundred or so of its over 800 pages are devoted to incorporating in the alphabetical sequence 'Spelling-to-sound guidelines' and excellent explanations of the terminology etc used in the dictionary.
Although not much concerned with Scottish, Irish, or Commonwealth usages, LPD deals extensively with General American as well as British pronunciation for which it not merely shows the originally and still mainly southeastern forms given in EPD but also usages common to the majority of well-educated speakers over the rest of England, like bath rhyming with Kath. These last items are identified by a dagger (exchanged for § in LPD2). Wells's unfortunately negative way of referring to them as 'non-RP' serves to highlight the regrettable insensitivity of the old Jonesian-Victorian term 'Received Pronunciation' to which he seems as wedded as most British phoneticians, when one might have hoped to see him give a lead away from that 'less than happy' term (his own words). These features are in fact almost all simultaneously Northern and Midland and so would be in nearly all cases far better labelled 'N' or 'NM'.
In referring to the 'model of British English pronunciation recorded in LPD, Wells comments that 'RP is not a local accent associated with any particular city or region' (p. xii); yet his 'modernized version' of it admits to RP the very markedly London (or at least southeastern) usage which eg distinguishes wholly and holy by using a diphthong [ɒʊ] with a much opener beginning for the former than his notation [oʊ] for the latter. This seems to represent a shift (at p. xix) from his view in Accents of English § 4.2.6.
LPD has various family resemblances to LDCE (the Longman Dictionary of Contemporary English), including American forms unimaginatively signalled by a preceding double bar ||. A 'possible stress shift' is indicated by a wedge sign. Those who can grasp the point that all really strong syllables in English can trigger this rhythmic treatment hardly need the sign. The device was introduced when Gordon Walsh was the pronunciation editor of the LDCE first edition in which he also first used the weak vowel symbols /i/ and /u/. Wells fortunately espouses them (for eg the final sounds of happy and thank you. He views them as conveying neutralisations of the oppositions normally found between the vowels of leap and lip and suit and soot. Whether one shares this view or not the more a learner might succeed in matching the final vowel to the first in a word like silly (as the older symbolisation seemed to invite), the more the result would suggest an elderly-posh or strongly Northern style. Many people may find the last sound of easy exaggeratedly clear if it is made exactly like the first. No one is very likely to react adversely to a weak sound there. Thank heavens, by the way, that the combinations of other letters on top of schwa that disfigure the LDCE transcriptions are avoided.
The adoption of /i/ has brought a few awkward complications in its wake. Surely it is not appropriate for any but very unusual and very markedly localised versions of the middle syllables of eg hurriedly, Derbyshire, or pumice stone, which should better have been treated more like those of eg funnily, multiply and uniform.
As in EPD, but much more comprehensively, many words from 51 (40 more than EPD) living foreign languages which may tend to receive less than fully naturalised pronunciations from some English speakers are shown in (over 100) IPA symbols as pronounced by native speakers of those languages. These should be fascinating for anyone with a little knowledge of the International Phonetic Alphabet.
Unlike EPD, which has practically ignored unhyphenable compound words, LPD includes in paragraphs after main entries '9000 important' compound words and phrases. These are not really enough and their placing is not in harmony with the general approach by which LPD scores over EPD by being more generous in providing their own separate alphabetical lines for many derived words like the adverbs which EPD squeezes uncomfortably and not infrequently misleadingly into the same paragraphs as their corresponding adjectives. American variants of compounds get little attention. To illustrate both these points, not only is there no entry for such a common expression as olive oil but also its omission precludes reference to its being earlier stressed in "Gen Am".
The LPD word-list is far more up to date than EPD's, notably going in for coverage of commercial products and pop-culture items. Obscenities are not eschewed. It is a longer list too, though there is a law of diminishing returns that operates with increase of size so that no such dictionary could ever completely satisfy the demands of even one user.
LPD has more than a dozen unfamiliar symbols besides the phonetic ones including various signs such as the asterisk used to draw attention to 'unpredictable and important' British-American differences, the bracketed italic exclamation mark to indicate (very occasionally) pronunciations 'unexpected for this spelling', and the equals sign to draw (occasional) attention to different spellings with the same sound values eg = fair at fare (though not vice versa). It does have the degree-style small circle to show 'tertiary' stress which is another small contrast with EPD (abandoned in LPD2).
The traffic-warning type of triangle for 'not considered standard' yields plenty of evidence of how difficult such things are to decide and how different individual impressions can be. It was quite surprising not to see it at the version of presumptuous that sounds as if it is being taken to be spelt presumptious and to see it used for the version of meteorological which suggests the spelling metrological. That LPD preferred model is not the one I've usually heard from BBC announcers: the wamed-against one is.
LPD takes a little getting used to. Take for example temporarily, which these days is mostly indistinguishable from temporally. When one has peered at the transcription and eliminated the "optional but not recommended" raised minuscule schwa after the /p/, and the two later optional but recommended italic schwas, one comes out with a shortest version which differs from a proscribed version only by having two successive /r/ letters rather than one. This must mean these two versions differ only by having a syllabic /r/ rather than an unsyllabic /r/, a tenuous contrast in such an unstressed situation. Not that the version recommended is not more realistically representative of usage than those currently offered in either LDCE or ALD. One fears that the ordinary EFL student will be daunted by the LPD's complexity. Perhaps most EFL students who understand the importance of constant checking with a pronouncing dictionary would find a slighter work quicker and easier to handle.
A new theory of English syllabification makes plain the special (in EPD invisible) syllabic structures of words like teapot (first syllable shorter than in teacake), peacock, bedroom, beetroot, rectangle, and so on by introducing spaces at all syllable divisions within words. I find myself intuitively uneasy about the visual results. I do not care for the publisher's decision to have all recommended forms coloured blue. I have repeatedly been irritated by mistaking an uncoloured American variant for 'recommended' through forgetting this convention. It makes ordinary photocopies highly misleading too.
Yet another innovation is the insertion of small loops below the line of print between successive syllables to indicate the possibility of their being compressed into a single syllable, thus neatly accounting without respelling for words like diagram, lenient and maddening being alternatively reduced from three to two syllables. Various other compressions that seem very common but are not represented include reductions of borrowing and worrying to disyllabic forms containing respectively the semivowels /w/ and /j/. A similar form of following is given but is oddly daggered as "not RP".
A matter that has been ignored in other pronouncing dictionaries is the existence of assimilated variants 'derived by automatic rule'. These are in LPD indicated by a horizontal right-pointing arrow showing eg that the first consonant of input can become an /m/. The trouble with such a practice is not that one minds their generous inclusion (though it might be argued that they are clutter and off-putting to some learners), but that one often wonders whether others one knows of have been omitted deliberately or not. Is one really to believe that /b/ is a common substitution for /d/ in cadmium but not for example in admin?
In the same way, one can think of many elisions that are perfectly commonplace and not accounted for in LPD such as previously rhyming with grievously, or thousand with no /d/. I found it amazing that a weakform of only was not acknowledged. The form with no /l/ is daggered as non-RP: in my observation, it is commonplace in all types of English and overwhelmingly the norm in unstressed situations. With such a work, involving so many thousands of decisions, it is no surprise that any one reader is likely to have many minor differences of opinion with it. There are not many really major differences of opinion between LPD and EPD.
In short LPD is a brilliant epoch-making achievement and obviously by far the best buy if one can afford only one such book. No EFL library is really fully equipped without it.