Volume Two

Item Maps

ANGUS McINTOSH, University of Edinburgh
M.L. SAMUELS, University of Glasgow
MICHAEL BENSKIN, University of Oslo
with the assistance of Margaret Laing and Keith Williamson
University of Edinburgh



1. General

The maps in this volume display the geographical ranges of the variants recorded for each of sixty items on the survey questionnaire. For present purposes, an item is a Modern English word, and the variant forms are what correspond to it in Middle English. This material is taken directly from the linguistic profiles (LPs) in volume III: all of the information there presented for an item or sub-item appears in this volume at its appropriate place on the map corresponding. The area covered is mainland Britain between the Forth and the English Channel, excluding the south-west of Scotland; for each place at which information is entered, all of the variants recorded from that place are entered directly on the maps, in their mediaeval spellings and with an indication of their relative frequencies in local usage. The conventions used here are the same as in the other volumes of the Atlas, except that the renderings of closely similar forms are condensed by means of square brackets (see below, p. xix); otherwise no symbols have been used, and the entries can be interpreted without recourse to a key.

For such a mass of detail to be reasonably legible, it is obvious that the maps must be printed at a fairly large scale. It has proved necessary to spread each item map over six pages of the Atlas; the geographical division is constant, and a key will be found on pp. 380–88. The same key maps can be used to identify the LP from which the entries at any survey location are drawn, whether for the item maps in this volume or for the dot maps in volume I. Identification may be assisted by use of the National Grid: references are entered in the summary descriptions at the head of the LPs from which the maps are constituted, and the grid itself is printed along the frame that encloses each item map (see further, p. xix). The county boundaries are printed for the most part as they were before the local government reforms of 1974 and 1975 (see below, pp. 380–81). They provide an additional key and, besides facilitating the description of geographical distributions, they link the cartographic displays with the appropriate sections of volumes III (Linguistic Profiles) and IV (County Dictionary).

No isoglosses appear on the item maps, and nor does any other interpretational device. Monochrome severely restricts the number of isoglosses that can well be attempted on a single sheet, as may be seen from the maps presented by Bloomfield and Newmark 1964 (p. 221), Strang 1970 (p. 420), Jones 1972 (p. 213) and Berndt 1982 (p. 36)––all of which are based on materials collected for this present Atlas. When very many distributions are to be highlighted, as is commonly the case here, even colour is inadequate: the course of any given isogloss all too easily disappears in a tangled skein. In these circumstances, if isoglosses are to be used at all, they are best printed separately on transparent overlays, in the style of the older fascicules of the Deutscher Sprachatlas; but such publication is bulky, and expensive. The solution adopted here is to print a separate series of small-scale dot maps (volume I) to accompany the item maps. For each item map, the associated dot maps are diagnostic: a dot map displays the distribution of some specified variant or set of variants, and on any dot map only one such distribution appears. The interpretation they offer is less subjective than that of hand-drawn isoglosses, and they can be multiplied at small cost to meet the complexities of distribution inherent in any item map. They lack aesthetic appeal, but offer an adequate key to the material. Their characteristics are treated more fully in the introductory notes that precede them in volume I. This is not the place to dwell further on the problems that attend linguistic cartography, but it may be useful to refer the reader to the excellent account in W. Nelson Francis’s recent book, Dialectology: an introduction (1983, pp. 110–44).

The sixty items for which maps appear in this volume are listed on p.2, together with such notes as are necessary to explain the scope and content of particular collections. A general account of the procedures that underlie the compilation will be found in the introduction to the LPs in volume III, and in chapter 3 of the General Introduction in volume I. The criteria by which the sixty items were chosen may be briefly stated. Since the number of item maps had to be so restricted, it was decided to print only maps for the whole area of survey; items collected for only the northern (NOR) corpus, or for only the southern (SOU) were rejected from the series. A high frequency of attestation was required: maps which would have contained a large proportion of null entries were excluded, unless the items in question were of overriding philological interest. Maps on which distributions proved to be uninformative were likewise rejected from the series, as also were those for which dot maps and the County Dictionary were an adequate substitute.

2. Representative character of the sources

It is at times difficult to assess how far an individual writer’s usage is representative of the language generally current in his own neighbourhood. If that neighbourhood is only sparsely documented, or not at all, then assessment turns on the likely nature of the local language, not on its known character. With the ‘fit’-technique, assessment of this kind is implicit: previously unlocalised material is assigned to a gap in the configuration just in so far as it matches a type of language implied to exist by the attested and localised combinations which delimit the gap.

In making such an interpolation, it is possible to discount eccentricities providing that they are recognised for what they are. For example, the spelling system of the Ormulum is unusually individual by comparison with other writings in early Middle English, regardless of their place of origin. When Orm’s idiosyncrasies are set aside, however, the whole assemblage of his forms can be localised with some confidence, in the Stamford Crowland Bourne area of south Lincolnshire (McIntosh 1963, p. 11; Laing 1978, i, p. 21; Parkes 1983, pp. 125–7). It must be left for some future editor to decide how far Orm’s spellings can well be entered, in their raw state, on the maps of a linguistic atlas for the twelfth century; the decision would no doubt be controlled in part by the availability or lack of other writings from Orm’s area. For the present Atlas, by contrast, the availability and spread of material in most counties enable a representative portrayal to be attempted; writings in language of a type that appears to be atypical for its area have in general been excluded from the maps (cf., however, 5.3.2 of the General Introduction to volume I, on certain of the Cornish material).

What counts as atypical depends here on the degree to which usages are unsupported by other writers from the same area; there is no question of comparison with some putative national norm. So, for example, by comparison with writings from other parts of the country, much East Anglian orthography is bizarre; but its oddity characterises a whole region, and it is in terms of such regional usage that the individual practice must be assessed. In East Anglia (especially Norfolk), and also in Lincolnshire, an individual writer’s range of variant spellings for a single word is generally greater than in most other counties, as immediately appears from the sizes of the text blocks on the maps; the contrast with the more disciplined orthographies of the West Midlands is particularly striking. Regional differences in the tradition and circumstances of vernacular literacy are perhaps to be inferred, and call for the attention of social historians no less than of philologists.

In the assessment of the representative or other character of a potential source for mapping, difficulties arise not so much from the established peculiarities of regional spelling, nor yet from obvious idiosyncrasies: rather, they derive from language that, in all save one or two of its features, conforms to what is expected from its apparent area of origin. It is the status of those few features that casts this whole language, even its authenticity, into question. When the source of such language is not an original text, but a copy, then the suspicion must always arise that the unconformable features are not part of the copyist’s own language at all, but relict forms perpetuated from an exemplar. (They could equally, of course, be sporadic intrusions of a near-literatim copyist’s own forms into what is otherwise the language of his exemplar: the problem of linguistic integrity is the same.) It is in the nature of the case, however, that when very few such unconformities are discerned, contamination remains doubtful: one or two anomalies do not amount to such a peculiar combination of extraneous forms as may point decisively to a Mischsprache (cf. 3.5 of the General Introduction in volume I). As may be expected, a good many such examples have come to notice in the course of work on the Atlas. By no means all of them, however, remain so uncertain as when they were first examined. Forms at first questionable as part of an otherwise localisable état de langue, some of them present in the host text only as rare by-forms, have proved in many cases to be integral, their local currency authenticated by appearance in other writings that were later found to belong to the same area. (For one such example, see Benskin 1981b, pp. 10–11, and cf. McIntosh 1983; more generally, McIntosh and Samuels 1968.) In similar fashion, forms which, regardless of their local origins, looked on first acquaintance to be aberrant or even spelling mistakes, have frequently turned out to be common currency as additional texts came under scrutiny. In cases such as these, evidence provided by texts which are independent of each other is an important safeguard: if confirmation of usage rests only on additional copies of a single text, then the perpetuation of a form, whether in isolation or in a given type of language, may be no more than just an accident of copying.

So far, literary manuscripts have been much in view, but the same reservations apply to local documents. It is not to be supposed that, because a given document is associated explicitly with some place named in its text, it can be taken on trust to represent the unalloyed language of the place so named. Men travelled, and their familiar habits of written language travelled with them. So, for all that he earned his living in Ireland, as registrar to the archbishop of Armagh, the notary William Somerwell continued to write the forms of Bristol English. Even had there survived no biographical evidence of the ordinary sort, his origins could still be traced to that town, because his language is at odds with all other known specimens of Hiberno-English, and can be localised as an assemblage only in Bristol or its near neighbourhood (Benskin 1977, p. 509). That William Partrik was native neither to Durham, where he was a monk, nor to Lancashire, where he was notoriously prior of Lytham, was established beyond reasonable doubt from his own written language; but for his earlier biography no materials were available. Partrik’s language can be localised, at Hemingbrough in the East Riding of Yorkshire. That a family of so uncommon a name was indeed established at Hemingbrough emerged only after the linguistic evidence had narrowed the search for biographical information to the relevant local historical records (Benskin 1982b, pp. 50–2).

The writings of Somerwell and Partrik present striking mismatches between the linguistic provenances of their authors and the places from which the documents issued. Local movements of people, by contrast, of the order of fifteen miles or so, may well remain undetected: there are, no doubt, such small-scale distortions in the configurations that appear on the Atlas maps, and as micro-dialectal analysis proceeds, it is to be hoped that they will be duly identified. Here, much will depend on the density and quality of documentation for the area in question; sometimes it may provide linguistic evidence strong enough to override an ascription based on the dating clause of an individual document, even though the adjustment involved is but a few miles.

So, for example, the researches of Professor Carter Revard have shown that the scribe of British Library MS Harley 2253 was active for much of his life at Ludlow, in south Shropshire; there, during the period 1314–49, he produced thirty-four documents in Latin (Revard 1979). Without other evidence, this might well be taken as a guarantee that his written English belongs also to Ludlow, that it is indeed firmly localised. Certainly it conforms more or less with what might be expected from that neighbourhood. Yet detailed linguistic comparison, between the lyrics in this scribe’s hand, and other writings from the area around Ludlow, indicates origins rather in Leominster, some nine miles south into Herefordshire (Samuels 1984, pp. 39–47). We cannot be sure that the Harley scribe would have replicated the forms of the lyrics had he ever produced legal documents in English, that the lyrics represent his own, spontaneous, usage; but suppose it to be so. A large series of such documents, all dated from Ludlow, would then be prima facie evidence of a most persuasive sort for the local language of that town; and were there no independent writings having equally good claims as securely localised sources belonging to Ludlow and its environs, then the linguistic provenance of the Harley lyrics could scarcely have been called into question. The case so stated is hypothetical, of course, though but slightly removed from what in fact obtains: the vernacular writings are not themselves local documents, but are linked to Ludlow by localised Latin documents in the same distinctive hand. All this indicates the need for caution in the use of local documents as evidence for local dialect; this matter is treated from a rather different point of view, and more extensively, in the introduction to the Index of Sources in volume I.

In the same place will be found some account of the fifteenth-century written standard, its London origins, and its effects on the local administrative and legal vernaculars. Here, it is enough to note that recognition of elements from the London standard in the language of a local document normally excludes it from the maps. Problematic, however, are those documents in substantially local language, but containing some few forms which are not otherwise known from the area in question, and which in principle might have been adopted from London usage. The problem is much the same as that noted above in respect of literary manuscripts in which linguistic contamination through copying is suspected; the chief difference is that the source of the suspected contamination, and its linguistic character, are here identified in advance. Evaluation is complicated to the extent that the local dialect in any case shares features with the London dialects; roughly speaking, the further from London the local dialect belongs, the fewer are the imponderables associated with a dialectal common core.

London presents problems of a special kind. The linguistic heterogeneity of the capital cannot well be indicated on a two-dimensional map, particularly given the relatively very rapid rate of change during the period of survey. The London material on the maps is contained within the administrative county of modern times; its disposition is perforce schematic. The later London language known as ‘Chancery Standard’ (‘Type IV’ of Samuels 1963) has been excluded: partly because it does not appear until the last decades of the period of survey; partly because, as the incipient national standard, it is already to some extent a non-local form of language. The disposition of the Type II texts (pre-1360) reflects their Essex affiliations. The Type III texts, belonging to the generations of Chaucer and Hoccleve, have been taken to represent the inner London usage.

The preoccupations and findings of urban sociolinguistics have of late brought register and class-dialect into focus as essential categories for the description of present linguistic change in complex societies. Application to the study of linguistic change in the past is beginning to produce interesting results, and even though the reconstructions are at times somewhat speculative, the new insights and perspectives are historically worthwhile (cf. Milroy 1981; Milroy and Milroy 1985, esp. pp. 375–80). That late mediaeval English society was complex is not in question, but in spite of the antidotes offered by the Paston letters and the Cely papers, to name only two of the best known, the prevailing image of that society has until recent years been one of order and static hierarchy. The extent of social and geographical mobility in those times, however, has begun to command the historian’s attention, and a very different view of provincial society is emerging. The genesis of literary works may be set in a new cultural context: so, for example, in Dr Michael Bennett’s account of Sir Gawain and the Green Knight, as the product of a society having intimate connection with the Ricardian court (Bennett 1979 and 1983). Inevitably, the view of the state of late mediaeval English will acquire depth as the social milieu is more clearly understood. It is to be hoped that the publication of the present work will stimulate further historical research of this kind, as well as the linguistically-oriented studies that can be expected to draw upon it.

3. Location of points

The geographical placings of the literary material are mostly and perforce approximations, arrived at by way of the ‘fit’technique described in the General Introduction in volume I (2.3). The local origins even of the linguistic material in the documents may not be precisely known: as outlined above, and considered more fully in the introduction to the Index of Sources in volume I, the dating-clauses of local documents commonly define a geographical focus for the language in question, rather than identify the particular village of origin. There is accordingly an inescapably schematic element in the maps as they are presented here.

Those who believe that, for the study of dialect, preciselylocated forms of language are a sine qua non, will no doubt entertain various misgivings about such a presentation. It will be observed, not for the first time, that the placings worked out via the ‘fit’-technique may not even coincide with human settlement. It will be found, for example, that in topographical terms LP 314 (London, British Library, Sloane 3160: hand M, ff. 166r–170r) is set near the summit of Kinder Scout, alias The Peak, a desolation in north-west Derbyshire some 2088 feet above sea-level and five hundred feet above the highest village in England. This placing, however, is not to be understood as an essay in peat-hag philology: we do not imagine that Middle English was regularly spoken there; still less do we suppose that it was written, or that LP 314 characterises ‘the dialect of Kinder Scout’. Rather, the position is this. The Peak lies roughly at the middle of the area to which the language represented by LP 314 is likely to belong, and that area is fairly narrowly delimited: within a five-mile radius there are several villages eligible as the historical place of origin, but there is at present no means of choosing between them. It is of course possible that the real place of origin, if we did but know it, lies somewhat beyond, but the disposition and linguistic character of the other material from north-west Derbyshire, and from the counties adjacent, fix the likely place of origin within so small an area.

Whether such indeterminacy is good reason for omitting LP 314 from the map is not wholly a matter of judgement, as will appear from the later paragraphs of this section; but matters of judgement may usefully be considered. It is implicit in the present work that such approximations, many of which must be fairly close ones, are worthwhile; and especially are they worthwhile if the alternative is no map at all. Accordingly, LP 314 is entered on the map; its place in the dialectal continuum is reasonably secure; and its topographical placing, on the wastes of The Peak, is a schematic element within historical limits that are reasonably narrow. Since the micro-dialectology of the Peak District is not itself the object of inquiry, it matters little whether the language so represented belongs precisely to Glossop or New Mills, Castleton or Chapel-en-le-Frith. Should the materials necessary for such a study ever emerge, the present placing may well have to be altered; and if the experience of codifying the Atlas materials be any guide, then the existing framework is likely to prove an indispensable aid in identifying and evaluating them.

Naturally, the extent of indeterminacy varies from one area to another, depending upon how close is the matrix of ‘anchored’ material––chiefly the local documents––and how packed are the interstices of the matrix with welldifferentiated ‘fitted’ material. In Devon, for example, the initial matrix is relatively sparse, and the resulting maps are hence not very finely focused. Devonian origins, except for one or two specimens entered near the border with Somerset, are little in doubt; but the anchored material permits only the definition of broad sectors within this county, inside which sectors placing is relatively unconstrained.

Such interpolations, within bounded areas from which no other information is available, can scarcely be objected to from a traditionally philological stance. Indeed, if for area we substitute time-span, then the procedures involved have been long accepted. A series of texts, individually undated, may yet be ranked in chronological order. That their absolute dates cannot be fixed need not invalidate the utility or good sense of the ordering.

The comparison may be sustained further. For all the confidence reposed in them, writings of known date can themselves be an uncertain guide. Normally, three generations co-exist, and on occasion even children may write an adult hand. If nothing were known of the writer of the Ayenbite of Inwyt, beyond the date of his work’s explicit, ‘Kentish’ of the mid-fourteenth century would seem variously archaic by comparison with other dialects; and so, on occasion, it has been described. But if Emden’s identification of Dan Michel be correct, and if the manuscript indeed be holograph, then the language of the Ayenbite may have been old-fashioned before the manuscript was begun, and may typify the usages of forty years earlier: in 1340, the date of the explicit, Dan Michel was probably in his mid-seventies (cf. Gradon 1979, p. 12). For philological purposes, the less certain ascription of the language entails no great loss: even a scholar who insisted that his sources must be of known date, and took the explicit simply at face value, would hardly maintain that the linguistic evidence of the Ayenbite held good only for 1340, and not for 1339 or 1341. In practice, of course, the time-span for which a dated sample serves as representative is usually determined by what other material happens to be available, and by the aim and scope of the inquiry.

Reflection on the character of dialect atlases for living languages will show that, even in the dimension of space, bounded indeterminacies are nothing new for linguistic geography: they are inescapable in so far as a survey’s coverage is not complete. The informants are never the whole population, but a sample, perhaps fewer than one in fifty thousand. For the area of survey, their dwellings are in turn only a sample of the inhabited places: the rest of the map is so much blank space. Even supposing that, within the terms of the investigation, the information presented for the points of survey be complete, the dialectal character of the settlements lying between them remains uncertain in detail––and strictly, it is unknown, though it may still be the subject of controlled hypothesis. It is worth glancing at some figures to set this indeterminacy in perspective.

The Leeds (Orton-Dieth) Survey of English Dialects covers England by a network of informants from three hundred and eleven representative locations (Orton 1962, p. 15). Each location represents on average, therefore, a gross area of about one hundred and sixty square miles; from this must be deducted an allowance for uninhabited parts, and for the areas of urban speech deliberately excluded from the Survey. Now within even one hundred square miles, there may be a good many different settlements. As illustration, from the fox-hunting country of north-east Leicestershire, which is not densely populated by the standards of agricultural England, sample counts average thirty or so villages and hamlets for such an area. In parts of Worcestershire and Norfolk the figure rises to forty, and no doubt higher incidences could be found elsewhere. If it could be assumed that the speech-forms of all the settlements within such an arbitrarily-chosen hundred square miles were the same, it would not matter in the least that only one of the settlements figured in a survey: the settlement chosen could perfectly well stand proxy for all of the rest. But the facts of language are otherwise: the dialects of even closely neighbouring villages may differ in detail, and be perceived to do so by their speakers. The area for which each survey point is truly representative cannot therefore be delineated, and herein lies the indeterminacy.

It may or may not, in particular cases, be reasonable to suppose that the usages recorded at a survey point are current to about the halfway line between that point and the next. On that basis, and to take from the Leeds survey an example at random, the village of Blagdon in Somerset is the informational focus for an area of roughly one hundred and eighty square miles (cf. Orton 1962, p. 31). The network in which it stands is somewhat uneven, with the result that Blagdon is not at the centre of its domain; were geometry the criterion, Felton four miles to the north would be a better choice. Consider this now in terms of the locations established for Middle English dialectology by means of the ‘fit’-technique. What practical difference would it make to a user of the Leeds corpus if the information collected at Blagdon had been collected from Felton instead? Or from Chew Magna, or Winford, or Flax Bourton, or even Wraxall? So far as this information is representative for the whole of the Blagdon sector, from a methodological standpoint it matters not in the slightest which of the settlements is its proper source––provided always that the material has been collected according to the Survey’s standard criteria, and so is directly comparable with the rest of the corpus. To return to the mediaeval configurations: suppose that the Blagdon material were one of the LPs derived from a Middle English text, and that application of the ‘fit’technique showed it to belong inside the Blagdon sector. Would it matter that the village to which it really belonged could not be identified, and that a schematic placing at Felton were the result? To what extent would the map for the middle ages be deficient by the standards of the modern survey?

Evaluation clearly depends on the use to be made of the survey corpus. For purely linguistic purposes, until such time as information were collected from other places within the Blagdon sector, the putative mediaeval map would be a research-tool no less reliable than the modern one. In neither configuration does the Blagdon material contrast with material from elsewhere in the same sector; contrast is possible only with material from the other locations in the survey network. The geographical ordering––ranking over space––is the same in both configurations, which therefore express the same dialectal taxonomy (cf. Benskin 1977, pp. 512, 503–4). When the issue is presented in these terms, the importance of exact location turns out to be largely illusory (cf. Sundby 1970, p. 507); and to that extent, the species of map aimed at in the Atlas is not inherently defective by comparison with what modern surveys actually produce. Its real deficiencies emerge when comparison is attempted with distributions established independently of the survey corpus. In a linguistic context, place-names provide an obvious example; but the patterns of dialectal variation may be correlated with many distributions of a more generally historical or geographical sort, like the incidences of Scandinavian settlement in England, or changes in the character of the landscape as they affect population densities and local contact. Even so, it may be wondered what kind of study could usefully demand such intricate correlations as to invalidate maps of the present type. In so far as these proved inadequate, so would the maps produced by most other dialect surveys, because the precise details of a distribution can hardly be established from the densities of coverage normally achieved in such work. Moreover, language is not a fixity that can be plotted with the same cartographic precision as burial mounds or settlement-types. In practical terms, there are limits to the kind of comparison it is sensible to undertake, and dialect maps in which the schematic element is unavoidable need not impose any crippling disadvantage.

There are, it is true, occasions when the local origin of a given form of language needs to be known in order that other kinds of inquiry may proceed. A recent and notorious example is the attempt to trace the murderer known as ‘The Yorkshire Ripper’ from a tape-recording of what was purportedly his voice. (The tape was a hoax, and the expertise of Stanley Ellis was applied in vain.) In the course of the Atlas project, intending editors and literary scholars have made countless inquiries about the dialectal provenances of the particular manuscripts that concern them, with a view to identifying likely scriptoria, the circulation of texts, and so on. Even allowing that dialectal origins need not tally with the place where the manuscript was produced, it is obviously desirable that they should be precisely stated. Approximations here are understandably irksome, and the limitations of the ‘fit’-technique may be regretted. But without it, or some closely analogous device, not even the approximations could be made. The choice lies between an imperfect discovery procedure and none at all.

4. Postscript on method

To conclude the Introduction proper, it may be useful to set the compilation of the Atlas maps in a broader methodological context. Analogous techniques have long been used in other disciplines, though (once again) with time rather than space as the familiar dimension.

Relative chronologies, and dating of material by typological interpolation, have been at the very foundations of some science. So, for example, in archaeology, Thomsen’s threeage system (Stone, Bronze, Iron), and later, Montelius’ assemblages of artifacts and typological classification; indeed, it was Montelius who gave ‘typology’ its modern meaning. Such techniques have of course been supplemented, notably during the last thirty-five years: radio-carbon dating, corrected against tree-ring chronology, enables historical dates to be fixed within fairly narrow limits for organic materials and the artifacts associated with them. Even so, the effects have been to refine the older chronologies, not to overturn them; and it may be noted that fixed dates do not of themselves provide an evolutionary schema, local or otherwise. Somewhat similar developments have taken place in geology: relative chronologies, based on stratigraphical sequences and fossil assemblages, can be assigned historical dates within ranges delimited by the decay of radio-active minerals. In geology, however, analogues with procedures for localising the language of Middle English texts can be found also in the dimension of space: fossil assemblages link not only strata of like date, but establish ancient geographical continuities since breached by erosion, or, more dramatically, by continental drift. In this perspective, the ‘fit’-technique is nothing new; it is merely a species of typological interpolation, an organised procedure for assigning a particular assemblage of forms to its proper part of the matrix, where it does not interrupt but reinforces the continuities already established. (Cf. 3.5.5 of the General Introduction to the Atlas in volume I.)

In the first instance, such interpolations are discovery procedures in which the material to be interpolated is the focus of interest. The initial problem is to classify a given assemblage, whether of fossils or artifacts or Middle English spellings. It is essential, however, that unless the evidence afforded by that assemblage is in some way impaired, it should then be made part of the classificatory schema. Otherwise, all that is achieved is atomistic labelling: the schema in terms of which the assemblage is to be understood is neither augmented nor revised––it may not even be tested. The part is interpreted by the whole to which it belongs, but the whole is no less a construct from its parts. Consider, for example, the likely state of biological classification if newly-identified species, fossil or modern, were at no stage permitted to count towards knowledge of their respective taxa, and so to affect subsequent classifications. Admittedly it has been alleged that the familiar taxonomic procedures are in fact circular, and particularly in so far as they bear on evolution, but it can be shown that such claims are largely misconceived. Petitio principii is not at issue here: it is rather a matter of successive approximations and (in Hennig’s apt phrase) reciprocal illumination (Hull 1969; Hennig 1979, pp. 21–2). The incorporation of ‘fitted’ material in the Atlas maps is closely analogous. The subject of classification is here not a single organism, but an assemblage, and the classification is spatial. Once the assemblage is classified––assigned to its proper place––it serves firstly as evidence for the dialect of the area to which it has been assigned, and secondly as part of the schema within which further additions to the corpus are to be classified.

On a small scale, and in time rather than space once again, configurations established by such means have long been familiar in English philology. In some cases, the intervals between dated (cf. localised) sources are narrow, and rather precise intercalation is possible. A notably intricate example appears in Professor Norman Davis’ study of the letters of Edmond Paston (Davis 1952). The identification of these letters as the work of a single hand, and the sequence of dates to which they are ascribed, are to a large extent mutually dependent. The letters are associated chiefly by their distinctive combinations of linguistic features, but from one group of letters to another the particular character of the assemblage shifts. These groups, however, can be ordered as a series in which there is a single direction of change, a record, as it appears, of one man’s progression from usage solidly provincial to semi-standard. The language of the earliest letters differs so greatly from that of the latest that if only these two groups were known they could hardly be associated as the work of one writer; it is only linkage through the intermediate groups that shows them to be the ends of a cline. To identify, by linguistic criteria, a letter as Edmond Paston’s, is accordingly to place it in a relative chronology; and because the dated letters in the series establish intervals of at most a few years, the relative datings are historically confined.

There are, obviously, mutual dependencies here, but they are not mutual justifications: the one is not a proof of the other, nor is it presented as such. Rather, a historical reconstruction is developed reciprocally, and so brings formerly diverse elements into coherent relation with one another. The validity of the reconstruction is a separate issue. Inner coherence offers one criterion for choosing between alternative accounts, and an important one, but in principle there are external tests: for example, the degree of coherence with what is known (perhaps more reliably) of contiguous matters, or the extent to which the reconstruction promotes new observations that do not depend upon it for their own validity.

From this it will be seen that the compilation of the Atlas maps represents no great departure from a methodology that is in other contexts well established; indeed, the previous paragraph could stand unaltered as commentary on the Atlas reconstructions. It may, however, be worth adding a further note on the ‘fit’-technique, and the assumptions that underlie its application. The technique has its precursors, in the neogrammarian mechanisms for ‘establishing the dialect of a text’. The phonological categories of Old English or West Germanic were the customary taxonomic framework and by progressive elimination the dialect of a Middle English text could be discovered to descend from this or that quarter of the Old English dialect pattern. In this form, the limitations of such a procedure are sufficiently obvious: the Old English dialect pattern is known merely in crude outline, and the possible regional discriminations are correspondingly coarse. But progressive elimination is the key to the whole problem, a narrowing down to some one area which is a plausible origin for the whole assemblage of forms to be localised.

Given a densely crowded matrix for a base map, instead of the fragmentary Old English distributions, the question then is how far the principle can be driven: at what level is closer discrimination impossible? the county? the hundred? the parish? In the present compilation, the answer varies from one part of the country to another, but it is only with the relatively undifferentiated material from the north of England that placings to within a county have regularly eluded confidence.

The mechanics of the ‘fit’-technique have been described in the General Introduction in volume I (2.3.3–7). It remains to note the criteria for its reliability. The character of the maps in this Atlas is in principle not relevant, for a procedure may be perfectly valid yet misapplied; and preconceived notions as to what the mediaeval distributions ought to look like are no test at all. Internal tests depend on repeatability. Different attempts to localise a given sample in the same configuration should produce the same result. Likewise, dialectally similar samples from diverse sources should be recognised as such by their ascriptions to the same part of the matrix. On both counts, the technique has proved sound; indeed, it has on various occasions provided the means of recognising diverse manuscripts as the work of but a single hand (cf. General Introduction 4.1.2; Benskin 1977, p. 513). Such tests show that it is a reliable indexing device for a given matrix, but the configuration is largely built up by the technique. The overriding question, if the maps reflect any historical reality, is the relation of such a matrix to ‘natural’ dialect maps. The issue therefore turns on the properties of dialect maps for living languages, and the results of experiment with the technique in conditions that allow independent assessment of the results. To take the experiments first, work with materials from the Linguistic Survey of Scotland lay at the very beginnings of the Middle English project: only after the technique had proved itself was it applied to the mediaeval material (McIntosh 1963, p. 6). For different purposes, the ‘fit’-technique and certain analogous methods of reconstruction were later tested on data from the Deutscher Sprachatlas; intensive work led to the same conclusions as before (Benskin 1972–3, unpubl.). It is hardly surprising that this should be so. For it is a property of regional dialect that it changes, for the most part, in an orderly way over space. Distributions unordered in respect of space can certainly be found, as witness those in Middle English for ‘betwixt’ and ‘between’; but continuity of distribution is relatively much the commoner, and a historical reconstruction in which it is maximised is therefore unlikely to be very far wrong. The sheer complexity of the material, combined with a large number of independently localised sources, sets its own limits on the extent to which alternative reconstructions are possible. When a dense configuration has stabilised, and excludes no genuine sample of a dialect that belongs to that continuum, then its historical inaccuracies are unlikely to be extensive.

By these criteria, the present configuration cannot be judged as a whole: some parts are much more densely documented than others, and there are areas from which, although material is available, it is not well differentiated and cannot be localised within acceptable limits. Comparison of the maps for northern England, which are perhaps unduly cautious reconstructions, with those for the Midlands and most of the south, will serve to illustrate the point; the linguistic profiles for northern material of unknown local origin, printed in volume III, are a measure of what has yet to be worked out. The position now attained is in some ways comparable with what had been achieved in geology and archaeology before techniques of radiometric dating were developed; the methods of reconstruction are closely analogous to those on which these sciences first depended. It remains to be seen whether techniques of a different sort will enable further advance with the Middle English material. The Atlas is work on a large scale, and the labour of many years, but it remains provisional, not definitive.

5. Production of the maps

Established methods of cartography have been prohibitively expensive at almost every stage of the project’s history. In 1963, the cost of bringing out the Atlas, including the preparation of over a hundred maps, was estimated at ca. £7,000––in terms of 1985 values, roughly £45,000 (McIntosh 1963, p. 10). There was no great hope that this sum would be secured. By 1982, the position was so far discouraging that, for the preparation of final copy per single item map, a major cartographic publisher informally estimated a cost of over £1,600. Even by the mid-1970s, however, it was sufficiently clear that automation, whether partial or complete, offered the only practical way out of the problem (cf. Benskin 1981a, pp. xxxv–vi).

Both the item maps and the dot maps are formatted by computer program, and printed by a phototypesetter (Oxford University Computer Service’s Lasercomp). Programmed formatting of dot maps is a fairly standard routine; before phototypesetters were available, graphplotters and line-printers were the usual output devices. (Displays on a Tektronix screen, which could be photographed with a Polaroid camera for lasting record, were much used for experimental work on the Atlas in 1972–4.) Maps on which more complex information is represented by a range of symbols are likewise straightforward to generate by computer; the first to be published for dialectal distributions are apparently those in Professor W. Nelson Francis’s paper on modal daren’t and dursn’t, a contribution to the Festschrift for Harold Orton in 1969. Automated production for such a series as the item maps, by contrast, is something new: the present system, in both design and implementation, was called into being by the Atlas itself. A brief memoir of the practical difficulties encountered, and of the means adopted to resolve them, will serve to explain the particular character of the final formatting.

In principle, it is easy enough to assign strings of textual information to predetermined points of origin on a page by means of a computer program, and then print the result as a ‘map’. So, for example, the following set of instructions to a typesetter: ‘move to the point of origin specified at the start of the entry, print a locational dot, move four points to the right, print the string of text rightwards across the page, proceed to next entry...’ But in practice this will not do. As may be guessed, the chief hazard is overprinting. Most strings must be broken into shorter lines and stacked, compacted into text blocks instead of running into each other across half a county or more. That reduces the lateral (east-west) scope for overprinting, and improves the look of the map, but only at the cost of increased scope for vertical (north-south) collision. The risk remains that some text blocks will overrun others, and for practical purposes that risk is a certainty. Moreover, in some parts of the map the points for which information is to be entered are so crowded together that, even when all of the available space is exploited, some of the text blocks are extremely difficult to fit in at all. Variable placing of the blocks in relation to their points of origin is here essential: a simple, once-for-all instruction to print the text blocks (say) rightwards and downwards from their respective dots, could be made to work only if the print were very small in relation to the map––a choice, in effect, between wide empty spaces and illegibility, because the density of documentation is so variable from one area to another as to rule out any useful compromise.

The system used here was devised by Professor Benskin. The necessary programs were written by Mr. Adrian Doyle, who implemented the system on the Edinburgh Regional Computing Centre’s Digitron in the summer of 1982. Mr. Robert Cowham was responsible for various additional features, incorporated in 1983, and for the essential datascanning programs. A report of this work, with a more detailed account of the formatting problems than is given here, appeared in the Association for Literary and Linguistic Computing Journal (Benskin, Cowham and Doyle 1985). In a post-script to that report, it was noted that various problems had been encountered with large-scale production runs on the Digitron, and that efforts were in progress to transfer the system to the Oxford University Computing Service’s Monotype Lasercomp. It was not expected that the transfer would be straightforward, and nor did it prove so, but it has been effected. For this we are greatly indebted to Mr. Doyle, who undertook the preliminary work, and to Mr. Hamish Dewar.

The principles of production are as follows. For each point on the map at which information from a LP is to be entered, an eligible printing area must be defined. This area can be regarded as a box. It must be big enough to contain the largest amount of textual information ever to be entered at the point to which it belongs, and it must not overlap any other box. Provided then that the information attaching to a point is confined always within its box, automatic formatting without overprinting is possible.

This base format is designed to meet the worst possible case, that is, the appearance on one and the same map of each LP’s largest entry. In practice, such a map does not need to be printed, because the item that yields the largest entry is not the same for all profiles: a composite map, in which some points display the forms for THROUGH, others for TOGETHER, yet others for WHILE, and so on, is of course no use in itself. It does, however, afford an economical and once-for-all specification of a practical base format. The alternative is to design a separate format for each item map, so losing many of the advantages of automatic production; and the density of print for some items in some parts of the intended maps is in any case not far off the theoretical limit. (A base format in which the boxes were drawn to meet some notionally average density of print would commit the whole operation to hazard: overprinting could be detected only item by item, after the text blocks had been composed, and a single case of it would spoil the map.)

The size of each box has to be calculated individually, an operation variously complex. The LPs are scanned, by means of computer programs, to ascertain for each profile (and hence for each location) the number of characters in the largest text block that must be constructed for it. It is necessary also to establish the longest unbreakable string of characters associated with each point. Suppose, for example, that the biggest entry consists of twenty-eight characters, contributed by four words each seven characters long: these could well be composed as a text block eight characters wide (allowing for the commas) and four lines deep, capacity thirty-two characters. Such a block could not, however, accommodate an entry of twenty-five characters if that entry consisted of five words each of five characters. Even an entry of nine characters would overspill, if it were a single word and not divisible. (Contrast, for example, to-gaddyre, divisible at an editorial hyphen, and togaddyre, which is not.) Especially where the map is crowded, the effects of sacrificing box depth for width and vice versa must be known, and these are likewise not straightforward conversions. Here again, computerised scanning and recalculation are indispensable.

The design of the base format is thereafter an intricate exercise in draughtsmanship. The locations for the LPs are entered on a base map, and boxes of the required dimensions are drawn, fitted in as they best may be. The resulting configuration is then treated as a graph, on which the locations of points and boxes alike are registered as x, y co-ordinates. In this form they are entered into the computer, as descriptions of the ‘pages’-within-the-map on which the text blocks are to be confined. The specification is then checked visually, by comparing a computer-printed version of the box map with the original. (For the record, preparation of the base map took well over six weeks, not counting registration of co-ordinates, key-to-disc typing, and subsequent checking.)

Once the base format is fixed, composition of the text blocks is wholly automatic. If a box lies below its locational dot, then it is filled from above; if the box lies above its dot, it is filled from below. If the dot lies at the left or right edge of the box, then the text is aligned with that edge; if the dot is centred, then so is the text. These rules apply no matter how much or how little text is contained on any particular map. In the interests of compact presentation, however, long lines are if possible broken: the third form in an entry of three is assigned to the second line of a text block, box permitting, even though there is space for it on the first.

It can hardly be claimed that the maps produced by such means have any great aesthetic appeal. In almost every case, a better balanced and more pleasing layout could be obtained by traditional methods: the format that proves adequate for all maps is of course ideal for none. Such aesthetic shortcomings, however, are fairly trivial when set against the strengths of the production system. Firstly, the base format provides a once-for-all solution to the problems of layout for crowded areas. For all its flexibility, traditional cartography is far from straightforward when information must be densely packed: repeated trial and error may be necessary, as appears from the manifold erasures on the working draft maps for the Atlas. These, large-scale and written in pencil, represent about five days’ work for each item mapped. In some areas, like the West Midlands and the counties about The Wash, it is commonly difficult to find space of any sort for the required information. Each map is a separate problem: the expedients adopted for one map may well not work on the next. Secondly, proofreading of the maps is much simplified. Locations of entries can be checked once for all: the position assigned to a LP on the base map is the place at which its forms will appear on any item map derived from that base. An error of placing will therefore affect the whole series, not just a single map, and to that extent the system introduces new risk; but visual checking of the base format against the original version can be effected by overlay, so eliminating the eye-movement otherwise necessary from key-sheet to finished map for each and every profile in the corpus. Moreover, the content of the entries does not need to be checked exhaustively against the LPs from which they derive, because the entries are simply a direct reformatting of the profiles themselves. Discrepancies between source profile and map entry are possible only if the reformatting programs are corrupt, or there is machine failure. Errors of that sort are usually easy to detect; whereas odd spelling mistakes or omissions, when the product of pen and ink, may elude the most careful of proof-readers. Thirdly, and to return to the opening theme, the advantages of economy may well be re-stated. Complex and clearly legible maps have been made the subject of routine production. Once the system is operational, the copy for an additional page of item map is no more complicated or expensive to produce than the copy for an ordinary typeset page of this work.

6. Editorial practice

6.1 The representation of mediaeval spellings is in principle diplomatic; but, if comparison with the practice of phonetic transcription be allowed, the renderings are broad rather than narrow.

6.2 Treatment of capital letters.

In the LPs, capitals and minuscules are not usually distinguished: the lower-case printed form implies either or both. Manuscript I ~ J, however, is reported always as I, never as i or j, unless the LP depends on an edition in which the mediaeval usage was incorrectly reported. In some of the SOU LPs, capital H-, and capital T- in the combination Th-, are distinguished.

H-. Late ME ‘it’ admits forms with and without intitial ‘h-’. In some southerly writings, ‘h-’ forms are preferred for apparently stressed positions, the ‘h-’less forms for unstressed positions. Forms with ‘h-’ may also be preferred at the beginning of a sentence, clause or line of verse, and there written H-; in such positions, H- need not imply a stressed variant. In SOU LPs, therefore, sentence, clause or line initial H- forms are segregated: an entry like it (Hit) implies that the selection of ‘h-’ and ‘h-’less forms is independent of stress; whereas in it (hit, Hit), variation between it and hit may well be stress-conditioned. In a more refined analysis, of course, stressed and unstressed occurrences would be systematically segregated. In NOR LPs, H- and h- forms are recorded indifferently with h-.

Th- is retained in LPs for texts in which capital Þ (or capital Y for ‘Þ’) is not used, Th- appearing instead. In LPs for texts where th- occurs regularly beside Th-, the capital Th- of the manuscript is reported as th-.

In SOU LPs only, F- has been distinguished from f-.

6.3 Treatment of i and j.

In classical and mediaeval tradition, i and j are merely variant forms (figurae) of the same letter (littera): j is ‘i-longa’. In the LPs, however, minuscule i and j are regularly distinguished. It should be noted, however, that in some manuscripts the two forms are not sharply distinguished, and that in particular cases it may be hard to decide whether a form is i rather than j; elements of judgement cannot be excluded altogether from the report. The correspondents of modern capital I and J are not distinguished in the manuscripts, and are reported here as I.

6.4 Treatment of u and v.

As with i and j, these are merely alternative forms of what, in mediaeval tradition, was the same letter. In the LPs, however, u and v are distinguished throughout.

6.5 Treatment of þ and y.

For reasons connected with the development of handwriting in the late 12th and 13th centuries, the letters ‘þ’ and ‘y’ came to be written identically in some modes of script. By the later Middle Ages, insular practice was regionally coherent: south of a line running roughly from the Mersey to the Wash, but excluding much of East Anglia, the distinction was regularly maintained; north of this line, and also over much of East Anglia, ‘þ’ and ‘y’ were represented by the same (usually y-like) symbol. (See further Benskin 1982a). In the LPs, the use of þ implies a systematic distinction between the two letters in the manuscript at issue, regardless of the letter-shapes used to effect it. If the letters are confused, then y is used throughout, regardless of whether the mediaeval symbol is þ-like or y-like. Renderings like mþkþll, corresponding to familiar mykyll, are hence not to be found, but appear with y; and in a manuscript where þ is so used, other þ- spellings are likewise reported as y (so þe ‘the’ appears as ye). The system of transcription, which attends to functional distinction rather than to form, is not ideal; but it is a practical means of reporting, in outline, an important facet of the written language, and it cuts through the taxonomic problems that would otherwise be presented by the many scripts in which þ-like, y-like, and indeterminately þ ~ y-like symbols are used interchangeably (see Benskin 1982a, p. 23).

6.6 Treatment of z.

The letter ‘z’ is reported as z or as ȝ, according to manuscript usage. (Commonly, the ME letters ‘z’ and ‘ȝ’ are not distinguished, but both written ȝ.) When z is written for ‘ȝ’, it is so reported: for example, manuscript zet ‘yet’ is preserved, not altered to ȝet.

It should be noted that word-final z/ƶ/ȝ may have origins other than in the letters ‘z’ and ‘ȝ’: they may derive from the abbreviation for -et found in Latin and French usage. The sign z/ƶ/ȝ was here originally syllabic, but was reinterpreted as a simple consonant, equivalent to t, and written post-vocalically. (So habeȝ from earlier habȝ, for habet. English asset(s) from French assez ‘enough’, is a back-formation.) The usage was adopted in some ME writings where forms like habbeȝ ‘have’ may imply a suffix of the -eþ type (with t from þ), rather than -es inflexion.

6.7 Treatment of n and u.

In most cursive scripts save the formal varieties, the letters ‘n’ and ‘u’ are not distinguished in form. Normally this presents no difficulty, and n or u is printed according to lexical identity and etymology. In some cases, however, the sequence of four minims may be ambiguous: is nn or un intended? So, for example, -aund or -annd in ‘land’ and the suffix of the present participle; -oun- or -onn- in ‘hundred’ and ‘young’. Here, typographical representations are almost bound to be arbitrary, and should be regarded as such in the LPs. In cursive scripts, whether formal or informal, final ‘n’ is commonly written with the second minim recurved upwards over the whole letter; the recurve may be reinforced by being drawn rightward again, into a tilde. Historically, this is a sign of abbreviation; the form may be identical with that form of ‘u’ in which the recurved stroke abbreviates a following m or n. In many, perhaps most late mediaeval scripts, however, the mark of abbreviation is a mere flourish, and modern editors usually ignore it. In general, we have followed suit, but in some scripts the flourish has seemed not to be otiose. Here, expansion to -ne has been preferred to expansion as -nn; and so also -me has been preferred to -mm. In some cases, it is unclear whether the two minims and tilde should be read as -un or -ne (or -nn): so, e.g., soun or sone (or sonn?). Again, representation is in some degree arbitrary; in spite of efforts to interpret these spellings in terms of the script and orthography of the texts in which they occur, it can hardly be claimed that the practice of representation is impeccable.

6.8 Letters superscript.

These are commonly abbreviations, as in þɩdde from þridde ‘third’, qan from quan ‘when’. For most items, such conventional abbreviations are expanded, and the letters implied by the abbreviation are italicised. So fam is represented as fram ‘from’, pei as prei ‘pray’, gow as grow ‘grow’. In some items, however, the superscript is always printed: whic and wch ‘which’, ic ‘I’, þu ‘thou’, wt ‘with’. In cases where the expansion is uncertain, the superscript is likewise preserved.

Superscripts that are not abbreviations, or only arguably so, are retained: ye ‘the’, þei ‘they’, yam ‘them’, boþe ‘both’, not ‘not’. In NOR, usually no attempt is made to distinguish between superscripts placed directly above an on-line letter, and those placed to the right. In SOU, the manuscript positioning is reflected by the typesetting in the LPs volume (III). On the maps, because of the restricted interlinear space, no distinction of positioning is made and all superscripts are set raised and rightward of the on-line letter.

It should be noted that the variation tends to be clinal, and that the binary classification is not always a sure guide.

6.9 Manuscript corrections.

Corrections entered by the writer of the mediaeval text are normally not distinguished in the LPs: insertions and corrected spellings are treated as running text. In the rare cases where a given correction may provide evidence for a copyist’s tolerance of the language of his exemplar, the correction is explicitly marked, either with an appended annotation (‘corrected form’), or (in the case of deleted letters) with the scribal sub punctum.

6.10 Manuscript errors.

In those cases where a manuscript form is clearly erroneous––that is, in cases where the scribe himself could be expected to have corrected the form, had it come to his notice––the fact is noted by appending ‘error’ to the form in question. It should be noted that such annotation is used only sparingly: various forms which appeared on first acquaintance to be aberrant were later confirmed in the usage of other writings subsequently found to be from the same area.

6.11 Doubtful forms.

In the LPs in volume III in cases where interpretation of a manuscript form is doubtful, readings are prefixed by a question mark ‘?’. It may be assumed that all such instances are nonce-occurrences in the writings from which they are reported: a second occurrence normally resolves any doubt.

6.12 Signs of abbreviation.

Abbreviations are conventionally expanded, and italicised. Expansions are conventional: except for the macron or tilde, and for ad hoc uses of the bar of contraction, a given sign of abbreviation is always expanded in the same way. Thus the abbreviation of the noun plural suffix appears always as italicised -es, never as -is or -ys. This involves departure from the traditional renderings of Scots texts, where the expansion is nearly always -is; but it would be wholly misleading to imply a difference between mediaeval English and Scots practice on this point, with -es giving way to -is north of the Border, when all that is at issue is the variant national practices of modern scholars.

The attempt to expand abbreviations to conform with the other spelling practice of the scribe in question we believe to be mistaken. Consider, for example, the writer who employs both þar and þ+abbreviation for ‘there’. It could be held that since the fully written form always has a, then the abbreviated form must be expanded þar: the writer ‘really meant’ þar. But equally it could be held that the writer ‘really meant’ þer and þar by turns; for since there is a historically regular abbreviation for -er, he could always save a little effort by using the abbreviation when he intended þer; whereas there is no distinctive abbreviation for ar, so that when he intended þar he had to write the word fully. The problem can be avoided altogether, however, by representing the form of the abbreviation rather than its supposed significance; and although letter sequences italicised have been used here, a case could be made for a strictly formal representation (e.g. by figures, in the absence of record type).

The bar through h and ll is disregarded, unless, as in some formal scripts, it is clearly an abbreviation and not a mere flourish. Similarly treated is the return-stroke from the ascender of final d.

Recurved final -r is printed ‘-re’; the flourish is generally otiose, but is of some taxonomic relevance in an inventory of ME scribes, and its expansion presents none of the problems that may arise with the other marks of (questionable) abbreviation. Note that its presence is in some degree determined by the mode of script: in textura it is rarely used as a mere flourish, whereas in some varieties of anglicana and secretary it is habitual.

6.13 Word division.

Words containing an unstressed prefix commonly appear as two words in the manuscripts. So bi for ‘before’, a geyn ‘again’, wt out ‘without’. In all such cases, the space between the words is represented by a hyphen (LP to-gader for MS to gader, etc.) In manuscripts the hyphen is almost never so used, save at line-ends; in the LPs, all hyphens are editorial. It should be noted that spacing between words in mediaeval writing is usually clinal, and that the insertion of hyphens may therefore be a matter of judgement. Hyphens are also used to represent the letter-spaces between the separate elements of periphrastic constructions, answering to a single test-word on the survey questionnaire. So manuscript vn to the tyme is reported as vn-to-the-tyme (243 UNTIL), manuscript þeiȝ al as þeiȝ-al (32 THOUGH), and so on. Such renderings are merely editorial convention, designed chiefly to aid in machine-processing; that they are unfailingly explicit has seemed good reason to keep them in the final copy.

6.14 Nouns and verbs: treatment of inflexions.

For most such items, the form of the stem only is the object of present inquiry; in so far as the inflexions are considered, they are treated as separate items, and without regard to their lexical adhesion. (So items 56 Sb pl, the noun plural suffix, to 64 Str ppl, the strong past participle suffix.) In general, therefore, inflexions have been suppressed in the lexical items, and replaced by trailing hyphens. Hence MS þenking ‘thinking’ is represented as þenk-, MS fryndis ‘friends’ as frynd-. (Such stems are accordingly not normally distinguishable in the LPs from similar stems abstracted from compounds: MS fryndschip would likewise yield frynd-). Relative frequencies are expressed in terms of the whole (undifferentiated) sg. and pl. entry.

Words ending in ‘-er’. When an inflexion of the ‘-es’ type is added, ‘e’ of ‘-er’ may either be retained in an ending of the full ‘-eres’ type, or syncopated so that the ending is ‘-res’. So, e.g., fader ~ faderes or fader |PT4||TWI| fadres, oþir ~ oþiris or oþir ~ oþris. Representations of the type fadr- and oþr- should therefore not be taken to imply fadre and oþre in the uninflected stem: they may well answer to stems like fadir and oþur instead.

In some of the most recently compiled LPs, the pl. and gen. sg. forms are assigned to the sub-category ‘pl’, regardless of their form, and the inflexions are preserved. Otherwise, the only regular use of the sub-category is occasioned by such inflected forms of 87 BROTHER and of 100 DAUGHTER as display a distinctively pl. stem vowel: plurals like breþer (sg. broþer) and deghter (sg. doghter) are always so treated.

6.15 Verse texts: forms in rhyme and alliteration.

In general, rhyming and alliterative usages are not reported in the LPs. Scribes who, in course of copying, regularly translate the language of an exemplar into (presumably) their own familiar form of ME, commonly transmit rhyming and alliterative forms unaltered: were these forms likewise translated, then the organising principles of the verse would be variously disrupted. (See further sections 3.3.5–7 of the General Introduction in volume I.) However, there may be reason to suppose that a copyist’s usage differed little from that of the authorial version of his text; or, a copyist’s usage may be in flagrant breach of the verse requirements, implying a degree of carelessness (and hence, perhaps, spontaneity), or plain intolerance. Such forms are obviously to be treated with caution: see General Introduction, 3.3.6. In the LPs, all such instances of rhyme are marked ‘rh’; forms occurring in alliteration have been excluded.

6.16 Punctuation.

No attempt has been made to record manuscript punctuation, except in the case of 158 I (personal pronoun) where a number of instances of the forms I, i, y highlighted by points have been noted; so, e.g. .I, i., .y.


Forms are occasionally commented with an italicised grammatical label, definition or remark following the form.

7. The maps: presentation

Grid references

The figures printed along the frame of each map are the eastings and northings of the National Grid (reproduced by permission of the Director General of the Ordnance Survey). The lines of the Grid, however, are not themselves printed across the maps; otherwise, linguistic information might be obscured. In the Index of Sources (volume I), and the key to LP numbers (this volume, pp. 375–79) grid references are given to six figures, as for example: 328 673, Arthur’s Seat, Edinburgh. The first three figures are the easting, the second three the northing. The combination of the first figure in each triplet identifies a 100 km grid square. In other works using the National Grid, these squares are normally identified by a two-letter prefix to the numerical scale; thus the present formulation 560 280 corresponds to the more familiar rendering TL 600 280. References are hence given to the nearest kilometre, not to the nearest 100 metres.

The numerical grid used here is the same as that printed on the maps of the Ordnance Survey historical series (Roman Britain, Britain in the Dark Ages, Britain before the Norman Conquest, and Monastic Britain), though grid references are given with prefixed letters in the indices to those publications.

It should be emphasised that the grid references are here no more than a finding system for material entered on the maps; their precision is in no sense a claim to the accuracy of the linguistic placings.

Condensed presentations of forms: ‘square brackets’

On the maps, but not in the LPs or County Dictionary, square brackets are used to condense two or more related spellings into a single representation. The square brackets enclose an optional element, as for example in schold[e]: both schold and scholde are so intended. The optional element is usually a single letter, but may be a separate word in a phrasal group: so [þe-]whoch stands equally for þe-whoch and simplex whoch. Note that expansion of [þe-]whyle[-þat] produces four variants (þe-whyle-þat, þe-whyle, whyle, whyle-þat); multiple bracketing has been avoided for the most part, but where it is used, all combinations are intended.

Square brackets commonly enclose two or more elements separated by commas, as in qu[e,y]lk. The presence of a comma indicates that the segment contained within brackets is obligatory, but that it may be realised by any of the separate elements so enclosed. Thus qu[e,y]lk is to be read as quelk, quylk; but *qulk is inadmissible (one of the comma-separated elements must appear), and likewise *queylk (elements cannot be combined across commas). Note again that expansion of representations in which two sets of brackets are used requires multiplication: tog[a,e]dd[i,y]r contains 2 x 2 variables, whence togaddir, togaddyr, togeddir, togeddyr. (Three sets with two elements in each expand to 2 x 2 x 2 = 8 variants, not 2 + 2 + 2 = 6.) Representations with square brackets occasionally combine obligatory and optional elements: e.g. sch[o,u]ld[e] is expanded as schold, schuld, scholde, schulde. Note that the first bracket contains obligatory segments but the last segment is optional. Only where space is very cramped, however, has more than one set proved necessary.

Square brackets do not cut across the parentheses ( ) and (( )) used to mark relative frequency: all variants implicit in a square bracketed rendering are of equal status.

Page content

There are six pages of the Atlas for each item map in the series; the geographical division is constant, and a key map to the six sections appears on pp. 383–88.

The marginal overlap of adjacent sections calls for special comment. A section, as printed on an item map page, is bounded by a frame. The frame is an absolute limit for the print area of the map, and inevitably it cuts across text blocks. The blocks so intersected are removed altogether from the page in question: fragments of blocks are not printed. Accordingly, for a profile location in a marginal overlap between sections, the text block may appear on only one of the corresponding map pages, not on both. Either the sheet frame runs between the locational dot and the space assigned to its text block, or there is not space enough for the whole block as defined by its co-ordinates (the eligible print area) to fit within the frame: in neither case is the block or its locational dot printed on the page. However, if the sheet frame coincides with a dot and all of the eligible text space falls within the sheet frame, the text block is printed. Compare the text blocks for the LPs assigned to York on sheets 1 and 3: on sheet 3 four blocks are printed about the locational dot, but on sheet 1 only LPs 1001 and 1002 are printed, since the blocks for LPs 145 and 1352 fall outside the frame of that sheet.

The margins of a map page are therefore to be read in conjunction with the pages for the adjacent sections: what is blank space at the edge of one page may be a print area on the corresponding overlap. Note, however, that for one of the map pages in such a pair, printing of the text block is invariable.

The marginal contents of the map page are hence determined by the disposition of the text blocks, not of their locational dots, in relation to the frame. The bounds of the area that is comprehensively reported may be imagined as a second frame, step-sided and lying inside the printed one; east and west it is inset by the widths of the excluded text blocks, north and south by their heights. For the sake of legibility, it has seemed better not to print it, but for any of the six page formats, the line can be established by reference to the key map on pp. 383–88.

Null entries

If, for any map, a given LP contributes no information, then the number of that LP is recorded at the point on the map to which it belongs.

Point of origin for text blocks

For most of the maps, density of documentation is high, and at many places there is much information to be entered. Crowding is more or less inevitable, and it may sometimes be difficult, at first glance, to associate a text block with its proper point on the map. The maps will prove easier to read, however, if the following conventions of display are borne in mind.

(1) If the text block has a straight left edge, then the point to which it belongs lies just to the left of it. Otherwise, the block is directly above or below the point, and the point will be found within the left quarter of the block’s perimeter.

(2) If the text block has a straight right edge, then the point to which it belongs lies just to the right of it. Otherwise, the block is directly above or below the point, and the point will be found within the right quarter of the block’s perimeter.

(3) If the text block is centred, i.e. neither the left edge nor the right edge is straight, then the point to which it belongs is either above or below the middle of the block.

Multiple sources for towns, etc.

For some places, mostly towns, more than one LP has been entered. Provided that the adjacent parts of the map are not densely printed, it is possible to display up to four such LPs separately; in the case of four, as (for example) at Canterbury, each of the four text blocks is disposed in a quadrant of the circle having the town as its centre, and the inner edges of the blocks are all straight. (If one of the LPs contributes nothing to a map, the LP number is entered in the usual way.) In some places, however, the requisite number of LPs cannot be displayed separately, and for cartographic purposes only, they have been conflated as a single LP.

The principle of conflation is that two or more LPs are run together, as a single LP, having a separate number of its own; any form attested in a contributory LP appears also in the conflation; the relative frequency registered for any form is the highest relative frequency that it attains in any of the contributory profiles. The facts pertaining to a conflation are set out in the introductory matter at the heads of the relevant LPs as printed in volume III of the Atlas, and in the corresponding entries in the Index of Sources (volume I, both County and Repository Lists). There are full crossreferences by LP number from the cartographic conflations to their constituents, and from the constituents to their conflation and to each other. The conflations are not separately printed in the volume of linguistic profiles, however, nor do they enter into the County Dictionary (volume IV).

Sub-profiles present a special case of conflation, which is treated more fully in section 8 of the introduction to volume III. Briefly, a sub-profile is a supplement to a full LP: in most cases, the full LP depends on the first hand of the text or manuscript, and the sub-profile records the peculiar contributions of a subsequent hand whose language is substantially the same as the first. There may be several such subsequent hands, and hence several sub-profiles appended to the same LP. Where sub-profiles are used, it is in the nature of the case that similar local origins are assumed for all. Accordingly, some text blocks on the maps are interrupted by one or more instances of semi-colon plus numeral, e.g. whoche, whych 1; wech 2; wich 3. The first semi-colon here terminates the contribution of the full LP; ‘2;’ ends the contribution of the second, and ‘3’ (without following punctuation) that of the third. An entry woch, wyche 3; wuch 5 indicates that only the third and fifth subprofiles contribute information for this item, and that the main LP is void.

For some places, of which Hereford is the most notable example, sources which could have been assigned to the same point have in fact been slightly dispersed. Here, there is non-linguistic evidence (duly noted at the heads of LPs) to the effect that several manuscripts were all written in the same town, and they are admittedly in broadly similar language; yet local immigration into the towns was a fact of mediaeval life, and when detailed analysis points to out-oftown affinities for the language of ostensibly urban writings, it has on occasion seemed better to indicate the probable local character of the variation.


Map of Britain in the Dark Ages (scale: sixteen miles to one inch). Southampton: Director General of the Ordnance Survey, 2nd edn. 1966, corr. reprint 1974.

Map of Britain before the Norman Conquest (scale 1 : 625000, about ten miles to one inch). Southampton: Director General of the Ordnance Survey, 1973.

Map of Monastic Britain (scale 1 : 625000, about ten miles to one inch). Southampton: Director General of the Ordnance Survey, 1978.

Map of Roman Britain (scale 1 : 625000, about ten miles to one inch). Southampton: Director General of the Ordnance Survey, 1978.

Bennett, M.J. (1979) ‘Sir Gawain and the Green Knight and the literary achievement of the north-west Midlands: the historical background’, Journal of Medieval History 5, pp. 63–87.

Bennett, M.J. (1983) Community, Class, Careerism: Cheshire and Lancashire Society in the age of SIR GAWAIN AND THE GREEN KNIGHT. (Cambridge Studies in Medieval Life and Thought) Cambridge University Press.

Benskin, M. (1977) ‘Local archives and Middle English dialects’, Journal of the Society of Archivists 5, pp. 500–14.

Benskin, M. (1981a) ‘The Middle English Dialect Atlas’. In Benskin and Samuels (1981), pp. xxvii–xli.

Benskin, M. (1981b) ‘A linguistic atlas for late mediaeval English’, Medieval English Studies Newsletter [Tokyo] 4, pp. 5–12.

Benskin, M. (1982a) ‘The letters <þ> and <y> in later Middle English, and some related matters’, Journal of the Society of Archivists 7, pp. 13–30.

Benskin, M. (1982b) ‘Marian verses from a Hedon manuscript: some new materials for the Middle English dialectology of the East Riding’, Revista Canaria de Estudios Ingleses 5, pp. 27–58.

Benskin, M., Cowham, R.H, and Doyle, A. (1985) ‘A computer-aided system for printing variable text blocks on crowded maps’, Association for Linguistic and Literary Computing Journal 5, pp. 1–24.

Benskin, M. and Samuels M.L., ed. (1981) So meny people longages and tonges: philological essays in Scots and mediaeval English presented to Angus McIntosh. Edinburgh: the Editors.

Berndt, Rolf (1982) A History of the English Language. Leipzig: VEB Verlag Enzyklopädie Leipzig.

Bloomfield, M.W., and Newmark, L. (1964). A Linguistic Introduction to the History of English. New York: Alfred A. Knopf.

Davenport, M., Hansen, E., and Nielsen, H.F., ed. (1983) Current Topics in English Historical Linguistics. Proceedings of the Second International Conference in English Historical Linguistics, Odense University, April 1981. Odense University Studies in English 4. Odense University Press.

Davis, Norman (1952) ‘A Paston hand’, Review of English Studies N.S. III, no. 11, pp. 209–21.

Dees, A., van Reenen, P.Th., and de Vries, J.A. (1980) Atlas des formes et des constructions des chartes françaises du 13e siècle. Beihefte zur Zeitschrift für Romanische Philologie, 178. Tübingen: Niemeyer.

Deutscher Sprachatlas, auf grund des von Georg Wenker begründten Sprachatlas des Deutschen Reichs...unter Leitung von Ferdinand Wrede. 23 sheets in 2 vols. Marburg: N.G. Elwert’sche Verlagsbuchhandlung, 1927–56.

Francis, W. Nelson (1968) ‘Modal daren’t and dursn’t in dialectal English’. In Studies in honour of Harold Orton, ed. Stanley Ellis (Leeds Studies in English N.S. II), pp. 145–63.

Francis, W. Nelson (1983) Dialectology: an introduction. London and New York: Longman.

Gradon, Pamela (1979) Dan Michel’s AYENBITE OF INWYT. Volume II. Early English Text Society, Original Series 278. Oxford: for the Society, Oxford University Press.

Gray, D. and Stanley, E.G., ed. (1983) Middle English Studies presented to Norman Davis in honour of his seventieth birthday. Oxford: Clarendon Press.

Hennig, W. (1979) Phylogenetic Systematics. Urbana, etc.: University of Illinois Press. (2nd edn.; 1st edn. 1966).

Hull, D.L. (1967) ‘Certainty and circularity in evolutionary taxonomy’, Evolution 21, pp. 174–89.

Jones, Charles (1972) An Introduction to Middle English. New York, etc.: Holt, Rinehart and Winston.

Laing, M. (1978) Studies in the Dialect Material of Mediaeval Lincolnshire. Diss. Ph.D., University of Edinburgh; 2 vols., unpublished.

McIntosh, A. (1963) ‘A new approach to Middle English dialectology’, English Studies 44, pp. 1–11.

McIntosh, A. (1981) Review of Atlas des formes et des constructions des chartes françaises du 13e siècle (Dees et al. 1980), Medium Ævum 50, pp. 136–42.

McIntosh, A. (1983) ‘Present indicative plural forms in the later Middle English of the North Midlands’. In Gray and Stanley 1983, pp. 235–44.

McIntosh, A., and Samuels, M.L. (1968) ‘Prolegomena to a study of mediaeval Anglo-Irish’, Medium Ævum 37, pp. 1–11.

Milroy, J. (1983) ‘On the socio-linguistic history of /h/-dropping in English’. In Davenport, Hansen and Nielsen 1983, pp. 37–53.

Milroy, J and Milroy J.L. (1985) ‘Linguistic change, social network and speaker innovation’, Journal of Linguistics 21, pp. 339–84.

Orton, Harold (1962). Survey of English Dialects (A). Introduction. Leeds: E.J. Arnold, for the University of Leeds.

Parkes, M.B. (1983) ‘On the presumed date of the manuscript of the Orrmulum: Oxford, Bodleian Library, MS Junius 1’. In Stanley and Gray 1983, pp. 115–27.

Revard, C. (1979) ‘Richard Hurd and MS Harley 2253’, Notes and Queries 224, pp. 199–202.

Samuels, M.L. (1963) ‘Some applications of Middle English dialectology’, English Studies 44, pp. 81–94.

Samuels, M.L. (1981) ‘Spelling and dialect in the late and post-Middle English periods’. In Benskin and Samuels 1981, pp. 43–54.

Samuels, M.L. (1984) ‘The dialect of the scribe of the Harley lyrics’, Poetica 19, pp. 39–47.

Stanley, E.G. and Gray, D., ed. (1983) Five Hundred Years of Words and Sound. A Festschrift for Eric Dobson. Cambridge: D.S. Brewer.

Strang, Barbara M.H. (1970) A History of the English Language. London: Methuen.

Sundby, Bertil (1970) ‘Present-day trends in Middle English dialectology’, Revue des Langues vivantes+Tijdschrift voor levende Talen 36, pp. 566–80.