OF LATE MEDIAEVAL
General Introduction, Index of Sources,
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
ABERDEEN UNIVERSITY PRESS
The Dot Maps in this volume serve two separate purposes according as the items to which they relate are or are not represented in the series of large-scale Item Maps in volume II. For the items included in that series, the dot maps are merely diagnostic, an aid to interpretation: they highlight the distributions of specified variants or classes of variant. Their function is thus comparable with that of a set of isoglosses drawn on the item maps themselves, though they have the advantage over isoglosses in that they do not in themselves imply artificially sharp linguistic divides; rather they reflect the way in which one distribution fades into another, and functional equivalents co-occur. In order to reduce the bulk of the Atlas, however, item maps have been printed for only sixty items; for the remainder, dot maps have been used as an economical substitute. Their obvious disadvantage is that they report only certain aspects of the distribution of variants for each item. Were they to be exhaustive, a dot map for each variant would have to be printed; and then maps of a generalised type, showing the distributions of classes of variant, would be desirable for interpretation.
It will doubtless appear that many of the maps were not worth printing, because the distributions reported are incohesive, or attestations are very sparse. The maps are, in the first instance, an answer to the question ‘What is the distribution of form x?’ That the distribution of form x turns out to be uninteresting or aesthetically unsatisfactory does not prevent the question’s being asked, nor need it follow that the question was not worth asking. A single dot map takes but little space; if it answers, efficiently and economically as it may, a question posed by only a handful of readers, then its inclusion is already justified. Yet, inevitably, the range presented here will be regretted in some respects by almost any user: ‘Why is there no map for the very form whose distribution is crucial to the particular research I have in hand?’ This applies, though with less force, to the diagnostic maps as well as to those dot maps which are the only cartographic representations for an item. Though dot maps are an economical mode of presentation, the overall constraints of space in an already extensive work have entailed a more restricted range of maps than the reader might have hoped to see.
The maps are presented in three separate series. The scale for the maps in each series is the same and is approximately 1:5350000 or 84 miles to 1 inch. The first series covers mainland Britain between Land’s End and the Firth of Forth; four maps are printed per page. The second series covers the northern part of the area surveyed and represents material collected for the NOR corpus only. The southern limit of this area, which does not coincide with the frame of the map, extends from the Welsh border through the Midlands, into north Suffolk. The area south of this line is a terra incognita in so far as the reported distributions are concerned. In this series, there are six maps to a page. The third series covers the southern area of survey, reporting material collected only for the SOU corpus. The northern limit of survey runs south and then westward from the Wash; for the reported distributions, continuities beyond this line are indeterminate. For the southern series of dot maps, the format is eight to a page.
The loci used in the dot maps are, of course, the same as those found in the Item Maps in volume II, but with the addition of locational points for the south-western Scots LPs. These fell outside the bounds of the northernmost sheet of the item maps.
Some LPs are a result of conflation of originally separate NOR and SOU LPs and so contain both items which were collected only for the NOR corpus and items which were collected only for the SOU corpus. The loci for these LPs appear on both the NOR and SOU series of maps, although, of course, for the NOR series only the NOR items are eligible for representation and in the SOU series only the SOU items are eligible. In the SOU series the dots for these LPs fall below the line marking the northern limit of survey, and in the NOR series they lie above the line marking the southern limit of survey.
Not all of the items treated in the SOU series are to be found in the LPs in volume III. The material presented as the Appendix of Southern Forms is reported in a coda to the main section of the County Dictionary in volume IV, pp. 311–25. Here, specified forms have been collected, but not the full range of variants that would have been elicited by systematic recording for the relevant items.
In the selection of forms to be mapped, the traditional, neogrammarian categories of analysis are an obvious starting point, even though they are inadequate as a basis for detailed description (McIntosh 1956 and 1963). It can be assumed that a majority of users will want to know what distributions emerge for the traditional range of dialect criteria, regardless of what other matter the Atlas may contain. Many of these criteria are phonological in conception, but in spite of the present commitment to recording the directly-observable written forms of Middle English, as opposed to the phonological constructs that may be based upon them, it is in general possible to accommodate these criteria in a straightforward way. Thus, for IT, the distributions of (respectively) forms with and without initial -h are duly reported. It should not, however, be too readily assumed that the two classes of variants correspond to the spoken-language variants /hɪt/ and /ɪt/. No doubt for the most part they do so; but conservative spelling may conceal /ɪt/ within the h- areas; and the unhistorical use of h- in other words indicates that in Lincolnshire, if not elsewhere, hit and its congeners are not to be taken at face-value as evidence for /hɪt/. It would have been easy, here or elsewhere, to present the material in ostensibly phonological form; but although the rationale for grouping, say, all th-/þ-/y- variants for THEY which have ei or ey as their stem vowel, is implicitly phonological, we have preferred to let the spellings stand for themselves. In this case, as it happens, the phonological interpretation is less certain than the traditional accounts suppose. It has been generally assumed that the diphthong corresponding to written ei and ey was the same as that for ai and ay; in the idiom long familiar, these are ‘merely orthographic variants’. Yet it is remarkable that the geographical distributions of, on the one hand ei and ey, and on the other hand, ai and ay, far from corresponding, are largely complementary; and that is true of these digraphs generally, not just in the word ‘they’. It is possible, of course, that regional spelling practice is here independent of phonology: we do not suppose, for example, that when y is written instead of historically regular þ the spoken language has correspondingly substituted /j/ for /θ ~ ð/, even though the use of þ versus y-for-þ is very much a regional matter (see Benskin 1982a). We do not have to decide here whether there are two phonemes or one, and what phonetic form they (or it) took; that is a matter for interpretation which will require work beyond anything we have attempted. It may be observed, even so, that different conclusions may have to be drawn for different spelling systems, and that the problem itself would most likely not have emerged independently of the present treatment.
In general, the items have been treated segmentally, as sequences of consonants and vowel symbols. For each segment, the attested spellings are reported, either individually or as classes, on separate maps. So, for example, NOT can be regarded as a sequence CV(V)(C)(C)(V), where parentheses enclose optional elements. All variants have n for the first C; no map is needed. For V(V), the possibilities include a, au, av, aw, o, ou, ov, ow, etc. The second consonantal segment may be null, as in no, though that is rare; usually t is present, though forms like nogh are not uncommon; a cluster, as in naght, is typical. The final (V) is necessary to allow for spellings like note, rare though they are. Even on a segmental basis, it would obviously take many maps to account for all variants, though still far fewer than would be needed to report each form individually. Spellings are therefore grouped. For the stem vowel, the following are distinguished:
The rationale is in part phonological, of course: au, av and aw correspond presumably to the same spoken form. But the group can be defended in terms more appropriate to the written language, and to mediaeval spelling theory. u and v are merely different figurae of the same littera; w interchanges, in many systems, with u, though the variation may be subject to rule. The group is in these terms ‘the letter ‘a’ followed by the letter ‘u’ in any of its figurae’. It may of course be the case that the distributions of au, av and aw are complementary, that their selection is regionally determined. That would not invalidate the initial grouping, but it would call for more detailed presentation. Even so, the generalised treatment would not be redundant: it is all too easy for essentials to be submerged in detail.
Each dot map displays the distribution of that form or set of forms specified in its caption. Places from which the specified form is recorded are represented by black dots. There are three sizes of black dot, corresponding to the scale of relative frequencies used in the LPs and the County Dictionary. The largest dot-size represents a form that is either the only one there attested for the item, or a dominant co-variant. The medium dot represents a form occurring with a relative frequency in the range ca. 33% to 66% of that of the dominant form for the item; in the LPs such forms, and in the County Dictionary the LP numbers, are enclosed within single parentheses. The smallest dot represents a form relatively uncommon in the source text(s) for the place at which it appears, its relative frequency less than ca. 33% of that of the dominant form; in the LPs and the County Dictionary, the forms and LP numbers respectively are enclosed within double parentheses.
Typically, the large dots appear in the heartland of a form’s range. Towards the periphery, co-occurrence with equivalent forms is expected: the ranges of two or more variants overlap, and free or conditioned variation is the norm. Medium dots are then characteristic. At the limits of its range a form may occur only as a rare variant, and it is then registered by the smallest dot. The scheme is idealised, but there are elements of it on many of the maps here. The maps thus reflect the patterns of fading and co-occurrence in a way not readily attainable by the use of isoglosses, of linear divides. The drawing of the line is of itself an interpretation, and inevitably arbitrary in some degree. The dot maps have their limitations, as will be considered further, but they are a valuable corrective in a study that has hitherto been obsessed with boundaries, and to the impairment of understanding.
At some places there are two or more source texts to be represented; that is, two or more LPs are assigned to the same point. In such a case, the size of dot printed is determined by the highest relative frequency attained by the form in any one of the contributory LPs. So, for example, if in one LP meny is the dominant form for ‘many’, in a second LP it is a minor variant, and from a third it is absent, and all LPs are assigned to the same place, then the largest size of dot will appear there on the meny map. If meny occurs in two or more LPs as a secondary variant–– status (meny)––and they likewise belong to the same place, then the dot-size is the middle one: the relative frequencies are not summed.
Most of the dot maps treat classes of variants rather than individual forms. Thus for ‘many’, the forms having e as the stem vowel are conflated for presentation on a single map; the several particular variants of this type––meny, menye, meni and so on––are not displayed one at a time on separate maps. A single LP may well contain two or more variants eligible for inclusion in the same conflation. So, for example, both manye and monye may contribute to a single record for the presence of the -ye ending in ‘many’. The size of the dot registered on the map is determined by the same principle as before: it corresponds to the highest relative frequency attained by any one of the variants in which the segment occurs. So a LP containing as its entry for ‘many’ mani (menye) ((monye)) would be represented by a middle-sized dot; one containing meni ((menye, manye, monye)) would contribute only the smallest size of dot. (There is obviously a case for summing the relative frequencies here, and recalculating the ratios to determine the appropriate dot-size; but our computing resources had to be directed to more fundamental matters.)
The black dots thus record the fact of a form’s occurrence, and, according to their size, the relative frequency with which the form occurs. A map consisting only of these symbols, however, can be surprisingly difficult to read. The source of the difficulty is the space between the dots: are they to be interpreted as interruptions to the distribution, areas within which some other form replaces the one portrayed, or are they merely gaps in the network of points to which the LPs have been assigned? The remedy adopted here is to print a grey dot at all points for which a LP is available. Although, as will be considered hereafter, there still remain indeterminacies in the pattern of the black dots, it is at least possible to establish the main outlines of the distributions, their main continuities and disruptions.
Some explanation of the preparation of the maps for printing may be of help to the reader. The grey dots, all medium size are superimposed (along with the coastlines) as a mask on all maps regardless. Where a grey dot coincides with a black dot both marking the location of the same LP, the largest black dot eclipses the associated grey dot. Medium black dots coincide with grey dots. Small grey dots show partial aurora of grey to their lower right.
For maps which report the distribution of just one or two forms, that is, just one or two manuscript spellings, the captions present no difficulty: the forms in question are listed, and any form not so listed is not included in the distribution. Commonly, however, maps report the distribution of classes whose individual members are too numerous to list within the space of the single line permitted by the present format. Accordingly, the following conventions are used.
(1) If only one or two forms are represented, they are cited in italics. Forms similar to, but not identical with them, are excluded. So, for example, a map stated as reporting the distribution of þey (italicised in the caption) would be valid only for that form, and would not imply the inclusion of the related forms they, thei, yey, yei and so on. The caption may, however, invoke a class of forms in so far as it designates a segment rather than a complete form as its subject. Thus, a map for ‘she’-variants advertised as ‘spellings with initial h-’ includes any form of ‘she’ which begins with the letter h, regardless of how the rest of the word is spelled: hi, hy, ho and their congeners all qualify.
(2) If a range of forms is reported, but the manuscript forms cannot all be listed, then the type is identified as a representation in inverted commas. So, for example, the formula ‘‘they’ type’ covers any spelling that can be regarded as a figura of that particular potestas, whether ‘th’ is realised as th itself, or þ or y or other regular correspondent of ‘th’ in the manuscript in question, and whether ‘i’ is realised as i or y. There is of course no suggestion that the ‘th’ or the ‘y’ of the representation for the type is canonical or an underlying ‘real’ version: these are merely particular tokens which have been used, arbitrarily, to stand for the type. Usually, the type is followed by two or three examples cited in parentheses; these, as manuscript forms, are of course italicised.
It has sometimes been necessary to use the symbols ‘C’ and ‘V’ for ‘consonant’ and ‘vowel’ respectively. Thus for HUNDRED, the formula ‘-dVrC’ means any form ending with d, followed by any vowel, followed by r, followed by any consonant. Covered here, for example, are honderd, hundert, hundirth. Excluded are forms like hundereth which have a vowel between r and the final consonant.
(1) In terms of the present corpus, the greatest deficiency implicit in the use of dot maps is the exclusion of so much of the northern material. Because it cannot be localised with any confidence, it does not appear on the maps, and the distributions of northern Middle English forms are correspondingly sparsely reported. By contrast with the more southerly variants, which belong to areas in which the network of localised material is fairly dense, the northern distributions are skewed.
It may appear, for example, that ‘swilk’ is predominantly a north Midland form in that the bulk of its attestations are from those parts. It is no less characteristically northern, as scrutiny of the County Dictionary will reveal; but because there are relatively few points for which information is available in the north, and because many of these depend on documents in which ‘such’ does not appear, the northern distribution is seriously understated. The northern distributions must always, therefore, be read in conjunction with the County Dictionary (volume IV).
(2) The variable density of the network of informants leads naturally to consideration of the extent to which a given LP––in present terms, a given dot on the map––may be taken as representative for the usage of the surrounding area. The space around the dot may be interpreted in two ways. Firstly, when it surrounds an ‘informant’ for which the only evidence of local origins is linguistic, then selfevidently it marks a degree of indeterminacy for the local origin of that ‘informant’. Secondly, whether the ‘informant’ is securely localised or not, that space marks an indeterminacy in the distribution of the forms entered at the point to which the LP is assigned: they may have been current over the whole of that space, or have given way to equivalent forms attested from other LPs on its periphery. For want of a denser network of informants, the degree to which the LP is representative for the whole of its surrounding area cannot be known. (This reservation applies equally to the distributions reported by maps for living languages, save that here the indeterminacy can be reduced by re-survey.)
In the more sparsely populated areas of England, it is not to be supposed that the space between LPs is ‘occupied’ by the set of forms attested in the enclosing LPs, or even by a sub-set of those forms: there may simply have been no writers or speakers living there. In northern England especially, there are large areas of upland from which no evidence at all is to be expected. Settlement confined largely to the dales, and the concomitant absence of a dialectal continuum over the watersheds, pose considerable problems for a reconstruction of the mediaeval distributions, and the present account is no more than a very cautious beginning.
(3) General statements about the distributions of variant spellings for segments of an item may be combined to generate unattested forms. Consider, for example, a pair of dot maps for WHICH, one of which reported the distribution of forms ending in -lk (‘whilk’ and congeners), the other, forms with e as the stem vowel (‘whech’ and congeners). Now suppose that the two distributions overlap: there is some area in which forms ending in -lk occur beside forms having the stem vowel e. It is perhaps natural to suppose that the segments combine, and that forms like whelk are current in that area. The inference, however, would not be justified on that evidence alone. It could well be that the -lk forms had only i or y as the stem vowel, and that the associated e forms depended on variants of the ‘whech’ type: ‘whilk’ and ‘whech’ co-occurred in the usage of individual writers. Thus far, even the existence of ‘whelk’ is uncertain: it can be generated from the maps, but whether it is attested is another matter.
The maps must therefore be used in conjunction with the County Dictionary. It will there be found that the ‘whelk’ type is indeed attested, though it is less common than the distributional overlap of the -lk and -e- variants implies. (Its usual form is qwelk, rather than whelk, and it is characteristic in Cumbrian documents.)
(4) It should not be assumed that the grey dots on a map automatically indicate the exclusion of the variant form(s) in question from the places in which those dots appear. In general, the grey dots do indeed imply the limits of a form’s geographical domain; but in particular cases, a grey dot may reflect only the absence of any information at all for the item at that point. Suppose, for example, that in the heart of the swilk-domain as registered by a dot map, there appeared a grey dot. If the dot were there merely because the LP for that point contained no evidence of any sort for the word ‘such’, then the grey dot could not reasonably be interpreted as a discontinuity in the distribution of swilk. Rather, we should suppose that, in default of evidence to the contrary, the form of ‘such’ current in that place was likewise swilk. If, on the other hand, ‘such’ were indeed attested in the LP, but never as swilk or some close congener, then the grey dot would indicate a linguistic fact of potential significance.
Accordingly, the use of some symbol to mark nonattestation or exclusion was considered. The reasons for rejecting it were twofold.
Firstly, on maps of the present small scale, it is all too easy for the main facts of a distribution to be lost in a mutliplicity of symbols. At a larger scale, more complex cartography might have been attempted, but the cost would have been enlargement of an already bulky work, or reduction of content. Since the material on which the dot maps are based is presented in full, both as LPs and a County Dictionary, the small-scale maps were preferred.
Secondly, the use of a symbol for non-attestation or exclusion would imply a degree of reliability to which these maps cannot in fact attain. The problem is that if a nonattestation symbol were used, it would of itself confer the status of an exclusion symbol on whatever sign was used to represent those other locations in which the variant form(s) at issue did not occur; and exclusion is not the straightforward notion that it may appear to be. The occurrence of some one form for an item, and only of that form, need not entail the exclusion of co-variants from the scribal dialect in question. Consider, for example, some place represented by a text in which the relevant item occurs only once or twice: in a sample (were such available) that contained many more instances of the item, the appearance of co-variants might well be expected. Such expectation is in part statistically informed; but it depends also on the degree of variation for the item found in adjacent areas, and on the individual scribe’s own degree of consistency or discipline. A scribe whose usage is highly consistent for those items which occur very commonly in his text––each word, only one or perhaps two spellings––may perhaps be trusted to reveal his habitual usage even for those items which occur only rarely in his text. From just two instances of chyrche, we might be reasonably confident that such a scribe always, or at least regularly, wrote ‘church’ in that way, and that chirche and chyrch and still more cherch and kirk were all absent from his active usage. Such individual consistency, however, may well mask the degree to which variant forms for the same item co-occurred in the writings of other, less disciplined scribes from his own neighbourhood. As an informant for dialect mapping, such a scribe may mislead: he is atypical, and implies a greater local uniformity than may in fact have obtained.
An attempt to register the geographical exclusion of some given variant, as opposed to its fortuitous non-attestation, would therefore self-evidently require numerical treatment of the material collected, and establishment of statistical confidence levels for the sample presented by each linguistic profile. For the present corpus, such analysis is operationally daunting, and far exceeds any resources ever controlled by the compilers of this Atlas. In addition, however, some assessment of the skew implicit in the samples presented by each informant would have to enter the calculations. Such assessment would have to attend not only to the textual character of each sample, but also to the behavioural characteristics of its scribe; and these would have to be judged at least in part by the character of the other samples available for the surrounding areas.
The coastlines used in the series of dot maps are based on an original map compiled on GIMMS, kindly provided by Ann Carruthers of the Department of Geography, University of Edinburgh.
The positioning of the dots on the following maps is based on the National Grid taken originally from the Ordnance Survey 1:625000 Outline Map of Great Britain (1975), with the permission of the Controller of Her Majesty’s Stationary Office.
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.
McIntosh, A. (1956) ‘The analysis of written Middle English’, Transactions of the Philological Society, pp. 26–55.
McIntosh, A. (1963) ‘A new approach to Middle English dialectology’, English Studies 44, pp. 1–11.