We live, as I have said before, in a golden age of lexicography. Never before have we had so many people working on so many different dictionaries, and the results are almost all splendid. The Trésor de la langue française continues its marvellous and rapid progress toward completion; in Germany, the re-editing of the Deutsches Wörterbuch -- as Alexander Pope said of the alexandrine line -- "drags its slow length along". Here in Toronto, the Dictionary of Old English, and at home in Ann Arbor, the Middle English Dictionary, make available unprecedented resources for the study of medieval English texts. In Britain, the massive task of revising the entire Oxford English Dictionary is underway. Smaller dictionaries of all kinds continue to appear; I might mention the two excellent volumes published not long since by the University of Toronto Press, the Dictionary of Newfoundland English and the Dictionary of Prince Edward Island English. All this can be celebrated, even without beginning to list the mass-market commercial dictionaries of English and of other languages, every one of which marks some advance on its predecessors.
These dictionaries were all begun with an uneasy mixture of anxiety and hope, and anxiety was an especially large ingredient in the works that pioneered the uses of computers in lexicography. Let me offer one of them as an example.
In March 1969, the Centre for Medieval Studies in Toronto organized a conference that was designed to launch work on the Dictionary of Old English. Re-reading the conference papers gives us some perspective on how we arrived at our present happy condition. Of course one might emphasize how different the world is today from what it was then. Professor Pierre R. Ducretet of the French department had just completed a concordance to Candide, and his was a tale of woe -- long nights spent feeding punch cards into a reader for their conversion to tape, for instance; the lack of any but tailor-made software. "It is not easy," he said. "It sometimes happens that, in running 20,000 cards on to a tape, a job which takes five to ten hours, towards the end of the job, one or two cards will spindle and the whole thing will have to be started over again." Probably there are few scholars now active who have handled a box of punch cards, and fewer still who can imagine the thrill of sympathy that ran though the audience when Professor Ducretet mentioned that dreaded verb spindle.
The 1969 conference was not all dread and gloom, of course. One might rather emphasize the clarity of vision of some of the participants. Professor Jess Bessinger of Cornell had just completed his concordance to Beowulf, and he was more prescient than many. He discussed how really large bodies of natural language text might be processed; he imagined what would later be called distributed processing, with "a single powerful computer, or linkage of computers" making it possible for geographically dispersed scholars to work cooperatively (or even competitively) on the same body of text. He could, perhaps, not have imagined the riches we enjoy today. It seemed a difficult job to create a huge concordance for all the surviving texts of Old English back in 1969, but that work was accomplished and I now have access to it from my workstation at home, while, at the same time, it is in use for the Dictionary of Old English. Equally convenient is the opportunity to use the massive collection of French literature prepared so carefully and over so many years by the workers on the TLF in Besançon and Nancy. This same vision of many workers connected to a massive body of information was also expressed at another conference on lexicography -- this one held in Leiden in 1977 -- by Professor Quemada; the greatest number of inquirers, he said, could work on the same material, with the products of their research in turn stored for future use. This vision has now evolved to the idea of a docuverse, an adroit new word blending document and universe (see Delany); in principle, this docuverse is a virtual world of all the documents -- a world in which, we hope, they will be universally available.
The present we enjoy was created for us by these visionaries of the past, and now it is our turn to speculate and dream about the future. I intend to focus on just one question -- and one that most of those present have already answered: Why should we devote our energy and resources to old dictionaries? What is it about them that makes them an especially high priority for study?
To begin with, old dictionaries contain information that can be found in no other place. Let me offer a rather complicated example. In 1653, that remarkable Scot, Thomas Urquhart, began his translation of the comedies of Rabelais. Now Urquhart was unusually qualified for this difficult job, having passed a portion of his youth on the Grand Tour of France, Spain, and Italy (where he "soon spoke the languages of those countries with such a 'liveliness of the country accent' that he passed 'for a native'"). While abroad, he "seized every opportunity of demonstrating the superiority of Scotland in point of 'valour, learning, and honesty' to any of the nations he visited" (DNB). In his translation of Rabelais, Urquhart renders the word pimpompet as bumdockdousse, a word not otherwise known in English and unrecorded in either the Dictionary of the Older Scottish Tongue or the Scottish National Dictionary. Edmond Huguet's excellent dictionary of the French of Rabelais' century reports merely that pimpompet was a children's game formed, as a word, on the onomatopoeic principle.
At this point, our inquiry turns to Randle Cotgrave's Dictionarie of the French and English Tongues of 1611. In it are found words that are in common use today, for instance botanic and medallion, that do not appear in non-dictionary English until half a century later. Thanks to Cotgrave, we know that they existed earlier. Cotgrave defines pimpompet as "a kind of game wherein three hit each other on the bumme with one of their feet." As this definition shows, Cotgrave was not always entirely clear as an explainer (though few can compete with him in lexicological verve); the object and rules of pimpompet (or bumdockdousse) remain, perhaps fortunately, shrouded in mystery. Bumdockdousse -- which we might gloss as "rump-thwacking" or maybe "butt-kicking" -- probably did not gain wide currency as either a word or a sport in Britain, but thanks to Cotgrave we know a little more about what sort of game it was. The editors of the Oxford English Dictionary recognized the worth of Cotgrave's Dictionaire, and it is quoted there more than 6,000 times. Perhaps the most significant testimony to his value is the OED use of the label "not in Cotgrave" as evidence that some word from French appearing later did not exist in English at the beginning of the seventeenth century. By going back to Cotgrave to elucidate Urquhart's bumdockdousse, Murray added -- however minutely -- to our understanding of sport and of Rabelais.
Just how valuable these early dictionaries are for the study of English was demonstrated on a much larger scale by the late Jürgen Schäfer. (Professor Schäfer's last appearance on the academic scene was in Toronto; just after his return to his family in Augsberg, he died in 1985 at the premature age of 52.) Schäfer studied the precursors of the "hard word" dictionaries of the early seventeenth century, the glossaries often appended to sixteenth-century translations from both the classical and modern languages. These glossarists, he declared, "were not merely indiscriminate copiers but true pioneers in the field of lexicography".
Schäfer's work -- brought to publication by his widow, Loretta Schäfer, and some of his former colleagues and students -- illustrates just how valuable these old word lists can be. Like all the rest of us, Schäfer stood on the shoulders of his precursors: those who formed and catalogued the great research libraries; others in our century who compiled the great bibliographies -- in this instance, the Short-Title Catalog by Pollard and Redgrave; still others who filmed or facsimiled scarce books to make them widely available. Schäfer and his co-workers scanned more than a thousand titles in the search for glossaries; devised an encoding scheme using fixed data fields on punched cards (fortunately for us, a now obsolete technology); and published the results. More than 6,000 quotations, as a result, supplement, amplify, and improve the coverage of English words found in the OED.
Why, one might ask, was the study of these old word lists worth the trouble? If all the additions were on the order of bumdockdousse, the effort could not easily be justified.
One use of Schäfer's work is to re-write the history of English lexicography. His study of these glossaries shows that the seventeenth-century dictionary-makers were more attentive to actual usage than had previously been thought. They did not mainly anglicize Latin or French words to swell the size of their books but gathered evidence of borrowed words already used in English. Schäfer's attention to minute detail also illuminates the history of the Oxford English Dictionary, a matter previously treated in his Documentation in the OED. Studying what might have been found -- but wasn't -- could be seen as an indictment of the Dictionary; in fact, the rather meagre harvest of all these labours only increases one's admiration for Murray and his successors, particularly when we now know that the collecting was often haphazard and whole sections of the alphabet gathered before Murray became editor were simply lost.
But the history of lexicography is a narrow field. If the study of these word lists contributed only to that, it would hardly be worth the effort. Yet much larger questions are invited by Schäfer's work. For one thing, he shows that Shakespeare's lexical originality has been exaggerated. Many words are given in the OED as Shakespearean coinages but they are not. This exaggeration of his importance is partly the result of the attention given to him as the great national poet of England. Some of these "Shakespearean" words -- attractive in Hamlet, for instance -- were, as Schäfer discovered, already in use by others -- as one might expect from a popular playwright aiming at pleasing, rather than puzzling, the audience. From my own viewpoint, as a historian of English, I find the "inkhorn" controversy much illuminated by Schäfer's collection. "Inkhorn" terms were the undigested borrowings from other languages -- particularly Greek and Latin -- thought not to match the "genius" of English but taken straight from the classical ink pot. In the sixteenth century, this battle of ancient and modern worked itself out over such words as the "Saxon" scypman vs. the "Latin" mariner. Schäfer gives us a useful insight into this dispute by showing how touch-line competed with tangent or wisdom-teeth with teeth of sapience. Even more significant, from my viewpoint, are his discoveries about borrowed words from the Americas. In the English Renaissance, anglophone patriots were already beginning to see the "treasure of our tongue" enriched by borrowings from distant languages. In fact, most of the American exotica -- like coca or sassafras -- came from a no more distant a spot than Spain.
We can argue, as I have just done, that old dictionaries are the repositories of knowledge found nowhere else, but this claim alone cannot suffice to justify the expense of turning them into databases. After all, scientific works -- say the Middle English translation of Guy de Chauliac's Cyrurgie, or the splendidly-titled Renaissance work, Robert Record's Urinal of Physick -- have an equal claim to our attention if we ask for information that cannot be found elsewhere -- in the case of these texts information about the practice of medicine.
Because they sketch the landscape of language, dictionaries are uniquely valuable for historical inquiry. To see what is now possible, we need to return to the ultimate origins of English lexicography, the classified word lists found several Old English glossaries. Here lists of Latin and Old English words were compiled for such categories as birds, metals, social classes, and parts of the body. But what can have been the uses of these lists? They were hardly helpful to translators, and only a little more valuable to those learning Latin -- presumably the reason for their compilation in the first place. The list of colour terms, for instance, was as a sketch of the "landscape", the viewpoint offered by the modern thesaurus or dictionary of synonyms. The lists, in short, express one aspect of the lexical imagination. What else can we create with the same application of lexical imagination?
In 1975, I built a database founded on C.T. Onions's Shakespeare Glossary, first published in 1911. This work is not a complete dictionary of Shakespeare's language; the scope of it he described in his preface.
The aim of the Shakespeare glossary now presented to the reader is to supply definitions and illustrations of words or senses of words now obsolete or surviving only in provincial or archaic use, together with explanations of others involving allusions not generally familiar, and of proper names carrying with them some connotative signification or offering special interest or difficulty in the passages in which they occur.
Creating a concordance from this glossary would, I thought, provide a useful map of Shakespeare's language, as well as offering insight into lexicographical method.
On the whole, this project was not very successful. True, I could create lists of Shakespeare's colour terms -- or at least enumerate those terms in which Onions had used the word colour in the definition. Similarly, I could identify aspects of Onions's lexicographical practice -- for instance, his discrimination of epithets as contemptuous, complimentary, or conventional. In all important respects, this method of sorting a dictionary was identical to that used by Laurence Urdang (working on the first Random House Dictionary) in the early 1960s to pull out entries with field labels (like astronomy) for specialist review or to check consistency of practice in definitions.
This approach turned out to be the right idea insufficiently pressed to conclusion. Part of the problem was that Onions did not treat the complete Shakespeare lexicon, only a somewhat eccentric part of it -- mainly the subsequently obsolete and provincial. The concordance did not give the whole range of Shakespeare's musical terms, for instance. There wasn't enough to say what I wanted to say, and it was not possible for me to do a Boolean search -- except with paper and pencil -- in order to build inferences, for example, about the intersection of sets A, B, and C. I had plenty of ideas about the lay of the land, but the map my concordance provided only identified isolated features and those not necessarily the ones of greatest interest for the student of Shakespeare's English.
Now, of course, that's all changed. I can use the astonishing power of the machine to consult all of Shakespeare's vocabulary, not just a part of it. But for me to do so effectively I need the interpretive side as well, and for that purpose I can turn to the map hidden in the multi-volume landscape of the new OED.
My current work involves nineteenth-century English, and the map of the OED provides an unparalleled view. Early in the century the medical community began to give unprecedented attention to classical suffix -itis to describe the inflammation of various body parts: tonsillitis (1801), hysteritis (1803), gastritis (1806), mastitis (1842), prostatitis (1844), colitis (1860), appendicitis (1886). As surgery advanced during the century, the suffix -tomy was freely applied to the practice of chopping out the offending organ: hysterectomy (1886), gastroectomy (1886), colostomy (1888), prostatectomy (1890), appendectomy (1895), tonsillectomy (1899). Now the ideology of English at the time was that it was a bad practice in forming words to combine elements from different languages, an idea we might describe as "etymological harmony". Bronchitis (1808) and laparotomy (1878) thus were both well-formed since the "Greek" suffixes were attached to words of Greek origin.Vaginitis (1846) and ovariectomy (1889) were not because their roots are from Latin. Some learned observers wasted their breath lamenting this practice of mixing word parts from different sources, suggesting, to no avail, that the word digamy was better than bigamy and dicycle better than bicycle because bigamy and bicycle violated the principle of etymological harmony. The on-line OED allows me to track the futility of their efforts.
A second example of the insight made possible by the use of the OED in database form draws our attention to the competition in nineteenth-century English between words formed with the prefix auto- and those made with the prefix self-. Early in the century, auto- and self- offered essentially synonymous prefixes for the creation of new words. Only a few new words were created using auto-, however -- among them autobiography (1809), and automatic (applied to machinery in 1802). Most of them were either derivatives of words already in existence -- for instance, autocracy (1655) yielded autocrat (1803), autocratic (1813), and autocratically (1860) -- or applied in limited technical contexts -- for example, autophony "self-examination of one's voice" (1862) or autolaryngoscopy "self-examination of one's larynx" (1870). Self- compounds, according to Henry Bradley's account in the OED, had achieved a particular vogue in the seventeenth century, especially in philosophical and theological writings, but the nineteenth century revived the practice of forming words with self-, using the prefix with "unlimited application". Some of these were in competition with their auto- counterparts -- for instance, self-acting (1824) with automatic, self-fertilization (1877) with autogamy (1880), and, of persons, self-reliance (1827) with autonomy (1803). What made these words created with self- "unlimited" was that the prefix could be attached to virtually any word class: self-abandonment (1818), self-awareness (1880), self-directed (1808), self-effacement (1866), self-loathing (1899), self-sacrificial (1855). For those who debated the propriety of new English words, these new formations with self- counted as ones properly belonging to the language since most of them were made "from its own roots and stems". This argument about the propriety of borrowings, and the dispute over etymological harmony, revived the old Renaissance dispute, adding yet another chapter to the history of the inkhorn controversy.
From the perspective of old dictionaries, these examples offer what is to me a powerful insight into the relation between linguistic ideology and linguistic practice. Of course since we are dealing with prefixes in this set of examples, it would have been possible for me to gather the same information by the use of the printed books, but it would have been far more tedious to do so, and I could not be sure of having collected all the examples. If I want to extend the inquiry, however, the computer is essential; unselfpitying, for instance, is a modern coinage by William Faulkner, a usage buried deep in the lengthy entry for the prefix un- and thus not easily found by conventional means.
Using the OED as a database allows me to turn the dictionary inside out for various purposes. I can, for instance, convert it to a thesaurus, revealing, for instance, the possibilities in English for naming the tasty sea-slug: bêche-de-mer, sea-cucumber, sea-swallow, and trepang. This may seem a somewhat exotic example, but it is obvious, I hope, what question this list answers -- that is, the question of the English encounter with the mysterious East with the borrowings here from Malay (trepang) and French (bêche-de-mer), the loan translation from Dutch (sea-swallow), and the homespun metaphor from English (sea-cucumber).
It is just as easy to transform the OED into an etymological dictionary. By selecting all the words derived from the Peruvian language Quechua, I can see differences between nineteenth-century imports and those words borrowed in the Renaissance -- for instance, quinoa (lentil-like grain, 1625), quinquina (the cinchona tree and its bark, 1656), quipu (the mnemonic device, 1704), vicuna (the cousin of the llama, 1622), viscacha (chinchilla-like animal, 1604). The nineteenth-century borrowings are more exotic than these but they also reveal, once again, the well-known fact that borrowed words are not always selected to fill gaps in the vocabulary but, as in tambo (1830), are often used to ornament travellers' tales with a been-to sophistication. A tambo is simply an inn. Making the OED into an etymological dictionary is not difficult, though one needs to know something of the practices of the editors -- for instance, William A. Craigie (who edited Q and V, among other parts of the alphabet) spelled Quichua with an i while his colleagues followed the more usual practice of spelling it with an e. It also helps to know that James Murray began the Dictionary with the idea that this language might best be called "Peruvian" or, more finely, "native Peruvian" (as in the entry for alpaca ). Those labouring on the complete revision of the OED have plenty of challenges like these to face; the computer enables them to impose consistency on the idiosyncratic -- and for us to discover inconsistency when they have not managed to do so.
What pleases me, of course, is that I can bend the Dictionary to my purposes -- without awaiting the revision that will appear in the next century. Having this particular old dictionary at hand enables me to do my work as a historian of English right now -- and fortunately I know enough of its quirks and foibles to compel it to yield the information I need.
Not everyone wants to make a list of the synonyms for sea-slug, nor is it to just anyone's taste to compile an etymological dictionary illuminating the impact of Quechua on English. Yet the point of building reference works (and databases) is to allow inquirers to ask unanticipated questions. Much as I honour them for their labours, those who converted the OED into a database seem not to have been especially imaginative in anticipating questions -- at least the examples I have seen mainly involve such strategies as counting how often Shakespeare is quoted. Now I certainly have no quarrel with using the database to investigate Victorian reading tastes or Shakespeare's contribution to English. But there is much more there to be discovered than that. The database gives us a map of the language and the means to investigate both the centre and the remote districts (see the essays by Taylor and Tompa). As I hope is obvious, I believe that any old dictionary allows us to see the landscape in new ways -- and that the sooner we have more old dictionaries to explore the better.
A generation ago a metaphor from physics became popular among those who study language -- just as the physicist sees light from the viewpoints of particle, wave, and field, so too linguists recognized that language varies depending upon one's perspective (Pike 1972). Words like pimpompet and bumdockdousse are exemplary "particles"; domains like the synonyms for sea-slug or the reflexes of Quechua are "waves". I want now to conclude with the "field" perspective.
James Murray wrapped up his Romanes lectures of 1900 by claiming that "it can be maintained that in the Oxford Dictionary, permeated as it is through and through with the scientific method of the century, lexicography has for the present reached its supreme development". In this, as in so many other things, Murray was entirely correct, and surely at the distance of nearly a century we can begin to see the "field" of the OED, the way in which is precisely mirrors the ideas of nineteenth-century linguistic science. So far, we have been reluctant to examine this question, and as a result, our conception of what a dictionary is and might be is still constrained by the magnificent OED.
Let me list some of the cultural "fields" that open themselves thanks to the availability of the OED as a database. With this powerful tool, we can see just what Murray meant by the "scientific method" of his century.
First it is obvious that the Dictionary mirrors nineteenth-century modesty about bodily functions, particularly sexual ones. Not until Robert Burchfield began his supplementation project did the very interesting and very frequent words from this domain receive proper treatment in an English dictionary on historical principles. One of the reviewers of the re-issue of the OED had already lamented this deficiency; in 1934, A.S.C. Ross declared: "it certainly seems regrettable that the perpetuation of a Victorian prudery (unacceptable in philology beyond all other subjects) should have been allowed to lead to the omission of some of the commonest words in the English language". Burchfield did much to repair the omissions, though it is clear that much more remains to be done if this domain of the English vocabulary is to be fully treated. Of course, the original OED did contain a few "dirty words", ones noticed promptly by senior-common room wits: ballock, merkin, and twat. The editors were obliged to include them for reasons that overcame squeamishness: ballock to explain the folk names for the plants ballock('s) grass and ballock-wort; merkin inexplicably appears, perhaps because it was both literary and obsolete; twat because of Browning's gaffe in thinking "it denoted some part of a nun's attire". Yet these entries are mere accidents, ones that pale in comparison to the rich hoard of such vocabulary in Farmer and Henley's seven-volume compilation published between 1890 and 1904.
A second, and equally well-known fact about the Dictionary is its view that southern England was the centre of the language and all else periphery. (Scotland was a privileged domain on the periphery, thanks in part to the patriotism of Murray and Craigie.) In an unpublished letter now in the National Library of Scotland, Murray wrote in 1900:
It is a cheap pleasure evidently which we have given our "philological time-fellows", and one that they need never be without, since every newspaper contains South African Dutch, Malay, Patagonian, Alaskan, Samoan, or Chinese words "new to Murray", who confesses ignorance of all these far-off languages, and is merely a Little Englander in lexicography. The bolder notion of an Imperial Dictionary or Pan-Lexicon, to include all the languages which Englishmen have & have had dealings, belongs to "the new Imperialism" no doubt. It will be a big order; we find the language of Little England enough for us. (MS 3219. f158)
As he had vastly expanded attention to the vulgar portion of the vocabulary, Burchfield was similarly pioneering in extending the scope of the OED to the United States and the Commonwealth, but with the database we can find out more than ever before about just what part of "Little England" Murray and his colleagues set out to cover and what parts of the larger English-speaking world they ignored. A comparison of the original OED with The Australian National Dictionary shows that coverage of the emerging English that region was very sparsely treated.
Nowadays, we view prescriptive dictionaries with suspicion, presuming that they are compiled by elitist charlatans bent on celebrating their own, or their paymasters's, usage. Since the OED is a great dictionary, it must not be prescriptive. Thus, in her recent Guide to the OED, Donna Lee Berg declares:
Contrary to the popular view of the OED as an authority on the "correct" usage of the language, the Dictionary is intended to be descriptive not prescriptive; it records, non-judgmentally, the history of the language as mirrored in the written words of a democratic mix of novelists, playwrights, journalists, scholars, scientists, legislators, politicians, diarists, saints, and philosophers.
This is an attractive statement, and it is certainly true that Murray's Dictionary was far more inclusive than the one foreseen in 1857 by Richard Chenevix Trench and far more tolerant than any other by that time produced.
But the word "non-judgmentally" simply will not stand the close examination of the OED that the database allows. Consider the nearly two thousand entries where the label "erroneous" appears, for instance; I include here just two lexical examples with seasonal Canadian content: badger, Murray declares, is "erroneously applied to beaver and otter"; brocket "sometimes applied incorrectly [to] a deer in its third year" (rather than its second). As for pronunciation, under the prefix pn-, Murray writes: "It is to be desired that it were sounded in English also [as in French and German], at least in scientific and learned words; since the reduction of pneo- to neo-, pneu- to new-, and pnyx- to nix-, is a loss to etymology and intelligibility, and a weakening of the resources of the language." Such statements as these are by no means "descriptive" or "non-judgmental". Surveying them as a group, we might notice that there is a strong reluctance to admit the anglicization of foreign borrowings -- particularly ones from the classical languages. In this respect, the OED swims against the tide of the language.
Murray's lecture was titled The Evolution of English Lexicography, and the "evolutionary" metaphor was as powerful for him as for any nineteenth-century scientist. In the OED, words "evolve", and the ones that survived into nineteenth-century English are more copiously treated than the ones that became extinct at some earlier stage of the language. Thus the OED is a historical dictionary as the language was seen from the perspective of the century, and Murray was discomfited by Thomas Hardy's revival of dead words which then had to be treated as contemporary usage rather than as obsolete or archaic. The fact that the Dictionary itself played a role in the revival was doubly troubling, since it was not playing by the rules of natural selection for Hardy to take words out of the Dictionary and install them anew in literature. With the database, it is easy to find the label revival (or revived) in close proximity to obsolete in order to discover this process of revivifying dead words -- particularly, according to the Dictionary, the preoccupation of Walter Scott. This process results in many now familiar words: jape, obsess, ruffle, scant, and veritable, for instance. Dais, we are told, "died out in English about 1600", but came back into use "due to historical and antiquarian writers". Chivalrous had been described by Johnson in 1755 as "now out of use", but not long after "writers on the romances of chivalry" made it again part of the language. From the viewpoint of the OED, usage was not supposed to be discontinuous, and words like these are treated as exceptional in some way. When species "evolve", some die and some survive. They are not supposed to come and go at the whim of Nature. Murray, and his successors, were not able to escape the "evolutionary" metaphor so compelling to science and to see the history of words in a different way.
In his 1954 Plan and Bibliography, my late colleague Hans Kurath described his expectations for the Middle English Dictionary; "The Editors of the MED," he wrote, "do not share the easy optimism of nineteenth century scholarship with regard to historical semantics." This sentence occurs with no subsequent elaboration, and I very much regret that I never asked him to explain it. Still, it takes no great leap of imagination to see what Kurath was criticizing as "easy optimism". He must have had in mind the alternating schemes of "historical" and "logical" ordering of senses that is one of the hallmarks of the OED. "Historical" ordering, of course, means presenting the sense development of words so the earliest usages come first and the subsequent ones follow. "Logical" ordering ignores the historical record -- presuming that gaps in it are accidents of collecting (see Zgusta). Usually in a "logically" ordered entry, the narrow senses come first followed by the broadened ones, or the literal is seen as prior to the figurative.
Murray's concept of nineteenth-century science was that there is regularity and order hidden within the welter of confusion and chaos. Sense-divisions have sharp boundaries rather than fuzzy ones; grammatical distinctions (say, between noun and verb) are prior to semantic ones. We can see how the vocabulary ought to have evolved -- even when the evidence suggests that it didn't follow a "natural" development. With the Dictionary database, we can discover the uses of the word natural in proximity to development to explore the idea of sense evolution. Thus, for formula, Murray declares: "Carlyle's use of formula, however, though suggested by a mistake, is in itself a very natural development from the ordinary sense." Murray's semantics is further revealed in such comments as "The subsequent development assumed in the arrangement of the verb is quite natural, though not actually established." (s.v. cant sb. 3). In case one had been impressed by the empiricism of the OED, such comments give an idea of just how much theory is imposed on data rather than derived from it.
These inquiries into the history and structure of the OED show just how much information is made accessible when old dictionaries are turned into new databases. Most of these observations simply could not be made from the published volumes. The power of the computer helps us wrest new knowledge from these old dictionaries. The more we can know through this work, the better our understanding of language -- and language culture -- will be.
 In Cameron et al. 1970: 12.
 In Cameron et al. 1970: 8.
 In Pijnenburg et al. 1980: 112.
 Schäfer 1989: vol. 1, 8.
 In Bailey 1991: 41.
 Onions 1966: iii.
 Murray 1900: 49.
 Quoted in Burchfield 1973: 84.
 Berg 1993: 5.
 Kurath 1954: 3.