Thoughts on the Second Edition of the Oxford English Dictionary

The second edition of the Oxford English Dictionary* merges the original Oxford English Dictionary (OED1), published between 1884 and 1928, with the recent (1972-1986) four-volume Supplement, to produce an extremely handsome and typographically seamless whole. The editors of OED1 set out to give a full historical account of all the words ever used in the English language; the editor of the Supplement, R.W. Burchfield, aimed to update OED with comprehensive evidence on 20th-century words and senses. So OED2, which combines these two dictionaries, makes a substantial claim to lexical authority. OED2 contains over half a million definitions, including 5000 which entered the lexicographers’ files after the relevant volume of the Supplement appeared. The new edition also switches to the International Phonetic Alphabet for representing pronunciation. How successful has the project been, and how useful is this publication?

The first question is difficult to answer. The riches stored in the two previous dictionaries are now much more accessible, and reviewers have already reported how delightful it is to browse through one of the 20 handsome volumes of OED2, and ponder, marvel or cavil at individual words and definitions among the extraordinary wealth of examples recorded on its pages. But what can such sample browsing tell us about the project as a whole, which purports to provide a systematic, reliable and comprehensive history of the English vocabulary? To make an assessment of the dictionary in its entirety, we need to judge its editorial premises and methodology. This boils down to three separate factors, crucial to any lexicographical enterprise: the nature of the sources consulted to supply evidence on word usage, the thoroughness and accuracy with which these sources were read, and the use made by the lexicographers of the evidence which the source study provided. Unfortunately, it is surprisingly difficult for the average OED2 user to form an accurate opinion on any of these three factors.

The obvious place to look is the Introduction. OED2 reprints the Historical Introduction to the OED1 by C.T. Onions, which tells how the editors, headed by James Murray, relied on armies of volunteer readers. These readers worked through lists of titles provided for them by the editors, and recorded on slips of paper the words and senses (in their original context) they thought worthy of inclusion in the dictionary. The slips were returned to the editors, who analysed them to draw up their categories of the various senses of words, and choose which quotations should illustrate which usage.

But Onions leaves us hungry for more information, especially on the three major lexicographical factors I have identified above: source material, reader proficiency, and the processing of the readers’ evidence. How did the editors decide on which sources to read? Was there a policy of balancing literary with non-literary sources? Did the editors carry out any vetting process, or did they accept all those who volunteered to be readers? How efficient and accurate were the readers, and how was this tested? What sorts of implications did the change-over of editors have for editorial policy, and how does this affect the dictionary?

The Introductions to OED1 and the Supplement do not answer these questions. James Murray’s grand-daughter gives more information in her book Caught in the Web of Words, which paints an extraordinary picture of the difficulties and problems which made Murray’s achievement even more heroic, but also more flawed, than is usually presented. Her account is not reassuring about the quality of the readers and the editorial processing of their material. Numerous problems arose from relying heavily on voluntary (and therefore unskilled) labour. Murray describes the readers’ ‘most pernicious and deceptive practice’ of reading books for particular letters only, and their tendency to pass over illustrations of the usage of common words. Murray also worried ceaselessly about the scholarly consequences of the pressures placed upon him by his publishers to complete the work. For example, he lamented in a letter to a friend, the distinguished Medievalist W.W. Skeat: ‘I have always said that the letter [E] ought to be done again.’ Such a statement, from the person who had best cause to know the relationship between the evidence and the dictionary’s treatment of it, is extremely disquieting. Was Murray merely a worrying, busybody perfectionist, or are some of his concerns well grounded? Should we trust the evidence on words beginning with E as reliable, or not? OED2 provides no answers, since it makes no reference to Murray’s reservations.

So far as I know, only one study of OED1 makes any attempt to subject it to thorough-going methodological examination. This is the little-known but revelatory book by Jürgen Schäfer on OED documentation.† In a paragraph that ought to be placed at the head of the introductory matter to the second OED, with its import prominently displayed in the self-lauding publicity material, Schäfer soberly remarks:

The increasing discrepancy between the methods used at that time [i.e., when OED was compiled] and those used now for evaluation calls for a detailed analysis of the nature and reliability of the OED documentation itself. Insted of providing an unquestioned basis for further research, the OED has to become its object. If we are ignorant of the premises of the OED documentation, we cannot properly evaluate it; and it is indeed remarkable how little is known about the application of the principles set forth in the introduction of the dictionary. Selection and distribution of source texts, lemmatisation policy, consistency of approach under various editors, to mention only a few of the major questions, have only recently begun to be studied in detail, despite the fact that all of these may have far-reaching consequences for the nature of the OED documentation and, of course, for any conclusions based upon it.

Schäfer chooses to examine just one aspect of OED policy: the accuracy of the dates given for the first use of a word. He reread a selection of the sources read by OED to discover that ‘at least one-third of the potential first citations in any corpus examined for the OED was normally overlooked.’ Of these, 30 per cent would shift OED dates by over fifty years, 7 per cent by over one hundred years.

One cause of such striking inaccuracy is that some authors were much better known and loved by the OED readers than others. The works of Shakespeare and Milton were combed for OED quotations with far greater thoroughness than those of Malory and Wyatt, which means that the former are disproportionately well-represented in the dictionary. Shakespeare is listed as the first user of 1904 words in OED, but 50 of these can be ante-dated from the works of Nashe alone, an author also read by OED for the same period.

This favouritism extended beyond authors to periods. ‘The chronological distribution of OED resources,’ says Schäfer, ‘is extremely uneven’: for example, OED readers examined nearly twenty times more works from the Shakespearean period than from the decades around 1500. Unsurprisingly, OED records more new words and senses from the better-read years, and this has undoubtedly contributed to the (not necessarily erroneous) belief that the late 16th century was a time of exceptional lexical productivity.

A similar explanation may account for the apparent paucity of new words from the 18th century. This may be due, not to the sterilising influence of the Augustan age, but instead to the loss of 18th-century slips, which had been assigned to American readers, and never reached Murray’s scriptorium. The 18th-century decline, therefore, may have more to do with perceptions than with reality: the point being, once again, that OED evidence is only as good as the quality of the lexicography providing that evidence.

Schäfer believes that these ‘astonishing divergencies in authorial reliability and in period coverage’ were largely unperceived by the editors. He also suggests further reasons (other than literary taste and relative familiarity) why readers should have read their texts with such varying accuracy. Early readers had been asked to look out for unusual words only, and even when Murray corrected this instruction, there would be a natural tendency for the eye to slide over familiar words and pick up unfamilar ones. Similarly, words of conspicuous morphology will stand out more than others, so that it might be very difficult, for example, to detect when a familiar noun started to be used as a verb, with no corresponding morphological change to alert the eye. (The verb mirror is used by Nashe, but OED dates its first occurrence 227 years later, in Keats’s ‘Lamia’.)

Schäfer’s evidence on the unreliability of readers is corroborated by one of the most indefatigable contributors to the Supplement, Marghanita Laski, who commented in a letter to the TLS in 1972: ‘As every dictionary-reader knows, two people can read the same book and record almost non-identical lists of words to be found in it. One reader can read a book twice and come up with a different lists of words each time. In addition, and little as it becomes me to denigrate my predecessors, many of the OED’s original readers were simply inept.’ Not very surprising, but crucially important for any assessment of the reliablity of OED. Why then do the lexicographers themselves not acknowledge it? Laski goes on to say: ‘I cannot speak for the earlier material, but I know that all the literature after, say, 1600 needs to be read again. The amount that has been missed in even the most famous works never ceases to astound.’ Let us hope that this is a wild exaggeration. But why do the lexicographers not recognise the problems that it raises?

The information I have presented so far largely refers to two of the three factors I identified as crucial to an appraisal of OED1: the efficiency and accuracy of the readers, and the use made by the lexicographers of their material. Laski makes an important observation on the third factor, the selection of source material for OED1: ‘anyone reading the OED would get the impression that it was the giants of literature who formed our language. Any reading in trivia shows this impression to be wrong ... it is clear that extended reading in trivia of past centuries could be as valuable to a revision of the OED as the reading of contemporary trivia has been to the new Supplement.’ OED2 prints a bibliography of works used by the readers, but no detailed information on the criteria for inclusion, nor the relative weight given by the editors to the ‘giants of literature’ as against more mundane sources. Yet close attention to literary sources is regarded by both lexicographers and dictionary users as one of the main jobs of OED.

Curiously enough, the lexicographical habit of turning to creative writers as sources to elucidate definitions of words has been little remarked. But great writers often go to notable lengths to use language in original, startling and imaginative ways, and this is one of the main reasons why we value them. We should not expect Shakespeare to represent general usage (of however small a class) of the late 16th or early 17th century, any more than we should expect Keats to represent that of the early 19th century, or James Joyce, Dylan Thomas, Ted Hughes, Anthony Burgess, Martin Amis (all OED2 authors) that of today. The opposite is true: we should expect the language of these writers to stand out in a contrasting way from current usage, although this will obviously vary from writer to writer and indeed from word to word. But this raises a major lexicographical problem for OED1. Any word selected at random from its pages will frequently have the bulk of its illustrative quotations drawn from literary sources, often poets or writers like Tennyson and Walter Scott who often employ archaic usage. Will this not tend to produce a biased account of the history of its usage? Murray’s comment on Browning illustrates the problems posed by poets for lexicographers: ‘Browning constantly uses words without regard to their proper meaning. He has greatly added to the difficulties of the Dictionary.’

The state of lexicographical theory at the end of the 19th century and the financial constraints under which Murray operated made many of OED1’s deficiencies inevitable. Despite its weaknesses, it is both a research tool of unparalleled value and a general source of delight. But its limitations are significant, and cannot be wished away or ignored. Reprinting OED1 in OED2, with its material virtually unchanged, merely reproduces these limitations. For the dictionary to be truly useful to an academic researcher, the user has constantly to exercise judgment, scepticism and imagination, wherever possible consulting other sources. But search as one may through the introductory matter to OED2, one comes across no guidance to the user as to how possible shortcomings may be identified, measured or circumvented: indeed, one has to search hard to discover any editorial acknowledgment of the problem, although a single phrase at the beginning of the Introduction to OED2, and a few lines at its end, briefly allude to OED1’s limitations. The comments of almost all the newspaper reviewers of the Supplement and OED2 testify to a touching faith in the OED enterprise: there is virtually no suspicion reported or doubts voiced of OED’s ‘supreme authority’.

What of the other major components of OED2, the 1972-1986 Supplement? In his Introduction to Volume I of the Supplement, Burchfield clearly states that the new dictionary was not designed to go back over OED1 material. This means that the Supplement, and consequently OED2, make no attempt to incorporate the vast amount of evidence on OED antedatings and corrigenda published in academic journals and scholarly editions over the last hundred years (although the Supplement’s pages do carry a very small number of pre-19th-century entries). Burchfield’s brief was, in fact, comparatively limited: to bring the dictionary up to date by recording new words, and new senses of existing words, that had entered the language since OED1.

In the Prefaces to the four volumes and in articles published elsewhere, Burchfield gives more information on editorial procedure than is available for OED1. Unfortunately, though, he gives no account of the problem that evidently bedevilled OED and renders it so liable to suspicion: reader reliability. Nor does he disclose details of the criteria governing the selection of source material, and the processing carried out by his assistants. Not least because of this lack of information, there are a number of problems with his dictionary, whether assessed independently of OED or in its role as supplement to the parent dictionary. These problems only become apparent after some extended use of the Supplement, and the casual user is given no guidance as to how to extract the maximum benefit from its pages.

Burchfield introduced some major changes in OED policy: for example, the inclusion of many non-English words, and expansion of the range of sources to include technical, scientific and popular works, notably journalism. These changes seem admirable, but have been implemented with some inconsistency. Reviewers have shown that certain areas of vocabulary are, for no apparent reason other than whim, treated far more fully than others (among words relating to hobbies and sports, there is virtually 100 per cent coverage for surfing terms compared with nearer 20 per cent for philately). This suggests that either source selection or reader reliability was not all that it should have been. More worrying is the failure by the Supplement to include many words from half a dozen recent American dictionaries, some of which are even cited in its bibliography. This exposes the third factor, editorial processing, to some considerable doubt. One of the advisers to OED1, Dean Trench, pointed out in 1857 that trivial and (possibly) transient words (like many of those the Supplement omits from these sources, e.g. unflapped, shock-rock, wargasm – or indeed service-break, as in tennis, or triad, as in the US strategic nuclear force) are nevertheless important as a record of a historical stage in our language. If the OED passes them over, how will historians two hundred years hence make sense of many aspects of our popular culture?

As newspaper reviewers noted with approval and delight, the Supplement seems to make a better job of recording the eccentric and once-off usages of a different cultural stratum, those of writers like T.S. Eliot, Virginia Woolf, W.H. Auden, Dylan Thomas and James Joyce (except for most of Finnegans Wake); though virtually no recognition is offered of the problem of using literary sources as evidence for current usage. Unfortunately, the Supplement is not beyond reproach in this respect too. I have done some research on source selection, reader reliability and editorial processing in relation to its treatment of Auden (a fair example, I thought, since the editor several times claims that the Supplement made a special point of recording this poet’s vocabulary). My findings strongly suggest that the treatment of Auden’s vocabulary was inadequate.

Auden was a voracious dictionary-reader – in 1972, his copy of OED was so worn that he considered buying a new one – and he peppers his poems with archaic and eccentric words probably deriving from his constant browsing through the dictionary (e.g. flosculent, semble, ubity and videnda, none of which have OED quotations later than 1800). Auden may have coined some of these words from their Latin root, or come across them in sources not read by OED readers, but occasionally it seems undeniable that OED was his inspiration: in ‘A Bad Night: A Lexical Exercise’, the word hirple, ‘to move with a gait between walking and crawling’, is to be found in one of OED’s citations for hoast, ‘to cough’, which occurs later in the poem.

I have looked up roughly one hundred and fifty unusual words from Auden’s works, to find that they are recorded with remarkable inconsistency. Some poems are not cited at all, despite the fact that they appear in volumes listed in the Supplement bibliography, and contain many words and usages just as notable as ones which the Supplement does record. In any one poem, some of the unusual words will get into the dictionary and some won’t: flosculent is not recorded, but semble is – they occur nine lines apart in ‘Thanksgiving for a Habitat’; ubity is not recorded, but videnda is – they occur a page apart in the same work. When such words are cited by the Supplement, they are variously and (to my eyes) inconsistently labelled as poet[ic], arch[aic], isolated later example, rare, with no indication how these labels were assigned or what the distinction between them is. And no hint is given to the dictionary-user of the curious lexicographical loop involved in putting back into the Supplement words Auden may well have lifted from the OED in the first place.

Reservations on the quality of the Supplement are confirmed if we turn to another delicate area, that of lexicographical prescriptivism. Most lexicographers these days tell you that their job is simply to describe usage, in much the spirit, if not the letter, of Dr Johnson two hundred years ago, who said his aim was ‘not [to] form, but register the language ... not [to] teach men how they should think, but relate how they have hitherto expressed their thoughts’. Such a policy is not without risk, for most dictionary-users look to the dictionary (and especially OED) for an authoritative statement on what usage should be. That is, they expect dictionaries to prescribe usage. The inevitable resulting clash can sometimes be resounding, as evidenced by the torrent of abuse that greeted the American publication in 1961 of the determinedly descriptivist Webster’s Third International Dictionary, which provoked the fear that we were well on the way to worldwide Communism, since Webster refused to ‘condemn some usages as less equal than others’. The inclusion of words like finalise aroused particular fury, although the sources cited (in the case of finalise, speeches by Eisenhower and Kennedy) provided authority for other, unexceptionable usages.

The trouble with Webster’s documentation of words like finalise, infer to mean imply, and swell, was that it gave no indication to an unwary user of the virulent execration such words could excite. The solution adopted by many dictionaries nowadays is to indicate in some way that a problem word (such as hope fully, or, disinterested to mean the same as uninterested) is objectionable to some users: i.e. to be descriptive about prescriptivism.

The Supplement does not seem quite to take this line, since the warning symbol employed to alert the user, the paragraph-mark (¶), is said to designate ‘erroneous or catachrestic’ usages. To use these terms begs the question. And once again, editorial policy on dealing with these words is inconsistent. The entry on finalise gives us no clue that thirty years ago this word was fiercely resisted in the States, yet such information could be vital to future understanding of the nuances of its use in (say) contemporary fiction. On the other hand, we are told (though without explanation, comment or paragraph-mark) that the use of hopefully as a disjunct is ‘avoided by many writers’. The quotation sources range from the New York Times Book Review in 1932 to the Guardian in 1971, puzzlingly suggesting that the word has been acceptable in otherwise respected publications. No similar warning accompanies the use of parameter to mean ‘a boundary or limit’; I have known this sense to produce paroxysms of rage in conservative speakers. Refute to mean ‘deny’, ‘repudiate’, is sharply ticked off as ‘erroneous’ and stigmatised with the paragraph-mark: yet quotations are drawn from such sources as the Observer and the Daily Mail – i.e., precisely those elsewhere used to substantiate acceptable usage. The comment on pristine to mean ‘unspoilt’ or ‘brand-new’ reads, ‘these transferred uses, though now increasingly common, are regarded with disfavour by many educated speakers’: but no paragraph-mark appears, and although ten quotations are cited, none indicate such disapproval.

A number of questions arise from this swift survey. I shall mention two: first, the wide variation in the range of editorial comment. Sometimes a paragraph-mark is used, sometimes not; sometimes the comment appears in brackets, sometimes not, sometimes there is no comment at all. The variety of the terms used to describe these usages suggests an inconsistency in their treatment: if an editorial policy had been properly formulated and executed, one would surely find some regularity in the descriptions. A second question is prompted by the apparently varying degree of reliance on (quoted) evidence. By what authority are some usages identified as erroneous, some as unacceptable or unaccepted but not erroneous, when the source of evidence is in each case a quality newspaper or journal like the TLS or the Journal of the Royal Society of Arts, i.e. precisely the same sort of source used elsewhere to support unexceptionable usage? It is inconceivable that the judgment of the readers of this journal, for example, will not vary significantly on the degree of impropriety attributable to even the few words I have instanced here. Would it not have been better to warn the user that such usage is controversial, and (preferably) cite quotations that illustrate the variety of views, rather than intrude personal judgment on the evidence? Such anonymous editorial statements appear misleadingly imbued with lexicographical authority, even though this is not borne out by the cited quotations.

So much for the Supplement portion of OED2 considered on its own terms. How happily does it mesh with the original OED in this new, conjoint edition? Burchfield set out, quite reasonably, to update OED in strictly limited ways. For example, he did not attempt to provide modern examples for words and senses which were recorded in OED1 and provided with at least one 19th-century quotation there, and are still current. If he had provided such modern examples, the Supplement would have had a new entry for virtually every word and sense listed in OED1 – and would probably never have been finished. ‘Our policy depends,’ Burchfield tells the reader (in a paragraph tucked away unobtrusively in the Introduction to the Supplement, Vol. I), ‘upon the realisation by users of the Dictionary that any word or sense not marked obs. or arch. is still part of the current language.’ This sounds acceptable. However, the policy has a number of significant implications both for OED2 and for any future editions.

Leafing through OED2, one comes across many words whose last illustrative quotation is dated pre-1850 – i.e. 150-odd years before the date of publication. So in what sense is the dictionary up-to-date? Murray’s original aim was to supply one quotation in OED for every fifty years of a word’s use. Soon, exigencies of time and space forced him to settle for one quotation every hundred years; and this, surely, is the lowest acceptable minimum. Is it desirable for the new edition of the ‘best dictionary of English in the world’ to feature such an enormous gap in documentation of recent usage? Moreover, is it really conceivable that usage has remained so static? And indeed, in many cases, it has not: the problem is sometimes that current usages are not recorded in OED2, sometimes that this dictionary gives a misleading indication of the current acceptability or frequency of earlier usages. Two examples will illustrate the point. The noun greed is defined as ‘inordinate or insatiate longing, esp. for wealth; avaricious or covetous desire’; seven quotations, from 1609 to 1874, illustrate various aspects of this sense. None of the quotations relate to food – yet wouldn’t most people regard food as one of the most obvious objects of greed? The adverb darkling (as in Keats’s ‘Darkling I listen; and, for many a time/I have been half in love with easeful Death’) is defined as ‘in the dark; in darkness lit. and fig.’ and illustrated with quotations ranging from 1450 to 1859. No indication is given that the word is now (and also in the 19th century, presumably) strictly restricted to poetic use. How will this be useful to the dictionary-user two hundred years hence, when all sense of what is and what is not current general usage will have disappeared? Similar sorts of point are illustrated by the treatment of alimony (sense 1), extinguish (sense 2b), disbloomed and foremention (vb.). I turned up these four examples in ten minutes spent leafing through OED2’s pages. I must assume that a full-length, thorough search, would turn up many more.

Up to now, OED-users consulted first OED1 and then the Supplement: the physical separateness of the two publications made it impossible to overlook the different editorial principles embodied by each. OED2 runs all the evidence together, so that one is not automatically made aware of where Murray (et al) stops and Burchfield starts. The Supplement has not done its job of updating OED adequately, and to merge the two dictionaries together in OED2 compounds the problem, because OED2 makes both implicit and explicit claims for a comprehensiveness the two separate dictionaries do not claim. The new medium tends to gloss over the inevitable, but significantly different limitations of the two separate enterprises.

This leads me to a further criticism of the second edition of OED: namely, to ask why the publishers thought that now was a good time to produce it. They could presumably expect institutions and libraries all over the world to put in an automatic order, and they will presumably make a reasonable return from the sale price of £1500 per set. The bulk of purchasers, I imagine, will already possess copies of OED1 and the Supplement, whether in full-sized format or compact form. In exchange for their considerable further investment, they will receive a very handsome set of books providing them with all the information currently provided in their two earlier dictionaries, and supplying a further 5000 words and senses – that is, a further 1 per cent on their original total. The Introduction to OED2 tells us that the dictionary’s future will be largely in an electronic form, making it extremely easy to update, revise, correct and supplement existing material. Would it not be better for the consumer to sit tight, and buy the dictionary in this new medium?

Such is the power of OED’s name and prestige, however, that many people do not seem to have grasped the provisional nature of this second edition. My old graduate college in Oxford has just sent me a newsletter, listing the purchase of OED2 as one of its appeal aims. I shall write back and tell them no.