The Wycliffite Bible and ' Central Midland Standard ' : Assessing the Manuscript Evidence

M. L. Samuels' article "Some applications of Middle English dialectology" (1963/1989) belongs to those rare pieces of scholarship which continue to be discussed and debated forty years after their original publication. Basing his arguments largely on orthographic evidence, Samuels outlined four types of incipient written standard in late ME (Types I-IV). Of these, Type IV ('The Chancery Standard') is the one most widely discussed by subsequent scholars with regard to its role in the evolution of Standard English (see e.g. Fisher 1996; cf. Benskin 1992, 2002). The interest shown in Type I ('The Central Midland Standard' or CMS) has been of a different, less diachronic kind, because its usage seems to have waned towards the end of the 15 th century. Why did CMS decline in spite of its apparent initial success and wide dissemination at the turn of the 15th century? The fate of its usage has sometimes been linked with that of the Lollards; since Wycliffite texts have traditionally been viewed as the core of the

M. L. Samuels' article "Some applications of Middle English dialectology" (1963/1989) belongs to those rare pieces of scholarship which continue to be discussed and debated forty years after their original publication.Basing his arguments largely on orthographic evidence, Samuels outlined four types of incipient written standard in late ME (Types I-IV).Of these, Type IV ('The Chancery Standard') is the one most widely discussed by subsequent scholars with regard to its role in the evolution of Standard English (see e.g.Fisher 1996;cf. Benskin 1992cf. Benskin , 2002)).The interest shown in Type I ('The Central Midland Standard' or CMS) has been of a different, less diachronic kind, because its usage seems to have waned towards the end of the 15 th century.
Why did CMS decline in spite of its apparent initial success and wide dissemination at the turn of the 15th century?The fate of its usage has sometimes been linked with that of the Lollards; since Wycliffite texts have traditionally been viewed as the core of the writing produced in Type I (Samuels 1963(Samuels /1989)), the stigmatisation of Wycliffism has been presumed to have cast a shadow over the prestige of CMS (e.g.Scragg 1974: 31, Burnley 1989).It is also possible to see the ultimate reason for the demise of CMS in its failure to make its way into administrative writing.Recent research on standardisation shows that this function is likely to be decisive in the ultimate establishment of a standard language (e.g.Milroy 1994, Rissanen 2000).The possible association of Type I with the University of Oxford (Black forthcoming), and its widespread use in medical and scientific writing (Taavitsainen 2000(Taavitsainen , 2001) ) make CMS appear primarily as belonging to the academic register, in a marked functional contrast to the administrative Type IV. 1   Before questions about the rise and fall of Type I can be meaningfully posed and answered, however, it is necessary to clarify in what sense, if any, can CMS really be regarded as a coherent, historically attestable entity.That Samuels' Types can hardly be viewed as standards proper is a position currently shared by most scholars.Although we have much evidence of late medieval awareness of orthographic variation in English, and comments about the communicative facility provided by Midland speech can be found in coeval sources, it is highly doubtful whether these attitudes as yet betokened an intentional prescriptivist ideology of standardisation (see Burnley 1989). 2 The methodological difficulty  The position taken by Samuels and subsequently supported e.g. by Sandved (1981) and Benskin (1992) views standardisation primarily as a process in which a variety of the written language spreads by imitation outside the immediate geographical location or social group where it originated.Typically such imitation is seen to be motivated by the high-prestige status of the variety, although communicative facilitation offered by the variety could also play a role in the process.Scholars emphasising the prescriptive nature of standardisation maintain that these criteria are not sufficient in themselves in defining a standard language; what is ultimately required is imposition of the variety from above and the presence of a language ideology which stigmatises variation and seeks to minimise it by explicit prescription (see e.g.Milroy 1994, Milroy 2000, Milroy & Milroy 1999, Stein 1994).For a good summary account of recent theoretical approaches to linguistic standardisation, see Taavitsainen & Melchers (1999).
of reconstructing a standard language almost entirely from orthographic evidence has also been noted by those researchers who prefer to see standardisation as a phenomenon extending to all levels of language (e.g.Wright 1996, Rissanen 2000).
A further difficulty with Samuels' Types as standards proper has to do with the degree of their internal consistency in terms of the widely recognised criterion for a standard formulated by Haugen (1966) as "minimal variation in form" (see Smith 1992, Black forthcoming). 3The presence of internal variation in CMS was recognised early on by Samuels himself when he identified a "more northerly sub-type" in addition to the main variety (1969/1989: 141).That variation is likely to be even subtler is suggested by Sandved's (1981) cogent argument about the boundary between CMS and more local dialects as clinal rather than absolute.
Recent empirical studies by Taavitsainen and Smith contribute further to the recognition of variation within CMS.Taavitsainen (2000) observes that the variety of CMS used in a large number of ME medical manuscripts differs to some extent from that found in Wycliffite writings.Smith's (1996: 70-71) comparison of single manuscripts of six texts traditionally associated with Lollardy reveals considerable variation within the body of Wycliffite writings itself.Given that Wycliffite texts played a pivotal role in the original definition of CMS by Samuels, it turns out, rather surprisingly, that none of the texts Smith analyses contains the expected forms of all those diagnostic items he investigates.The degree of variation witnessed prompts Smith to view CMS as a standardised/focused language/usage rather than as a standard one in the strict sense.This concept, Smith (1996: 66, 67) explains, stands for "a centripetal norm" -"a sort of mean towards which scribes tend" -rather than a fixed and prescribed entity followed by scribes (see also Smith 1992Smith , 2000;;Black 1999, forthcoming).
Haugen's perhaps slightly misleading term for this purely language-internal criterion is codification.
See Smith (1992) for the application of this and Haugen's three other criteria to Samuels' Types.

The present study: aims, methods and materials
Smith's and Taavitsainen's findings imply that it is difficult to sustain a 'grand unifying theory' about CMS as a monolithic usagea system conceptualised as a single entity by late medieval English scribes working at different times in different locations and on texts of different genres and registers.Although there does not seem to be any immediate reason to discard the concept when used as an umbrella term for supra-local late ME orthographies consisting of central Midland features, it can be argued that its more concrete historical contextualisation and textual pinpointing would necessitate a bottom-up approach, in which the first goal of research is the description of CMS as it appears in manuscripts of a single ME text.This method would ideally enable the researcher to relate orthographic concerns of individual scribes with constraints of textual tradition and of the immediate copying situation. 4  The present paper reports the results of a pilot study in which the method is applied to a sample of manuscripts of the Wycliffite Bible (WB).This choice of primary material is motivated by the special status of WB among proposed CMS texts.Judging from the more than 230 extant manuscripts which contain the text in whole or in part, the production figures of WB must have surpassed by far those of other proposed CMS texts; this is also likely to reflect its degree of popularity and extent of dissemination in late medieval England. 5 Moreover, Samuels ' (1963/1989) reference to Skeat and Dibelius as scholars who had previously discussed Type I implies that CMS was in the first place largely extrapolated from WB usage (see Skeat 1895-96, Dibelius 1901;cf. Fristedt 1953: 39-42).With respect to the conceptual history of CMS it therefore seems legitimate to regard WB as its archetypal representative.
The very large number of surviving copies makes the linguistic and codicological uniformity associated with manuscripts of WB appear particularly intriguing.This association has been made in particular for the so-called Later Version (LV), represented by some 85% of the surviving manuscripts (as estimated by von Nokken 1997).The idea itself goes back at least to the mid-nineteenthcentury edition of WB by Forshall and Madden, who in their introduction asserted that "[t]he copies of this later version present so great an uniformity, that their peculiarities scarcely admit of an observation" (Forshall & Madden 1850: xxxi).Samuels' (1963Samuels' ( /1989: 79, n. 5: 79, n. 5) comment that "[p]ractically all copies of the later version" are written in CMS reflects the same position from a more stricdy linguistic perspective.Codicological aspects of this uniformity are most clearly evident in Doyle's (1983) and Hudson's (1989) identification of similarities between the mise-en-page of LV New Testament manuscripts.Doyle (1983: 169) notes the survival of "innumerable copies" of the text "produced ... to common patterns, in similar text-hands and with decoration of professional standards"; in a similar vein, Hudson (1989: 131) observes that "[t]he manuscripts vary in type and quality a good deal less than might be expected given the history of the text".Whether these similarities are explained by the association of the manuscripts with patterns of professional book-production in the metropolitan area, as Doyle (1983Doyle ( , 1990Doyle ( , 1997) ) surmises, or whether some provincial Lollard scriptoria might have been responsible for what looks like the 'mass production' of copies, are questions which require a considerable amount of further scrutiny and comparison of manuscript data to be resolved.
For the present paper, the spelling of eight diagnostic CMS items was analysed in manuscripts of the Later Version housed in the Bodleian Library (BOD), the British Library (BL), Cambridge University Library (CUL), the John Rylands Library (JRL), and Lambeth Palace.To facilitate explicit textual comparisons between the copies and a straightforward conversion of the results into percentages, ten fixed loci for each item in the gospel of Matthew were monitored.Matthew was chosen because it provided the largest number of surviving manuscripts.If the output of more than one scribe was clearly discernible in the monitored loci in one manuscript, the scribe responsible for a majority of the forms was focused upon.If the number and division of hands was judged to be ambiguous, no analysis was attempted.As a result, 68 LV manuscripts (hands) were included in the study (see Appendix I).
The items analysed derive from Samuels' early work on CMS.In the original formulation of the case (Samuels 1963(Samuels /1989)), seven diagnostic forms for Type I were given: sich for 'SUCH', mych for 'MUCH', ony for 'ANY\ silf (or 'SELF', stide for 'STEAD', ^ouun for 'GIVEN', and Ä> for 'SAW\ Of these, 'MUCH', 'ANY', 'SELF', 'GIVEN' and 'SAV were chosen for the present study; the disqualification of'SUCH' and 'STEAD' was due to their very low number of occurrences in Matthew.In Samuels (1969Samuels ( /1989)), the doubling of vowel symbols -of i in particularemerges as a further characteristic of Type I, with spellings such as wijf for 'WIFE'.AS representatives of this category, three frequendy occurring items were chosen: 'FIRE' (expected CMS form fijr), 'LIFE' (lijfi, and 'LIKE' (lijc).(See Appendix II for the loci analysed).
Before moving on to discuss the results, it is necessary to consider for a moment the preferable level at which orthographic variation should be measured in the present case.This consideration is motivated by a methodological decision made by Samuels, whereby the concept of CMS became further abstracted and its status as a written usage was rendered more ambiguous.Samuels' (1963Samuels' ( /1989Samuels' ( , 1969Samuels' ( /1989) characterisation of CMS shows, somewhat oddly, that even though Type I is viewed as a written standard, the diagnostic forms are represented at a level which essentially corresponds to variation in the spoken system.Thus mych, for example, reflects at least the scribal forms "mych", "myche", "mich", "miche"; si% stands for at least "si3", "si3e", "syj", "sy3e"; and ßouun represents not just "3ouun", "30UÜ", "3ouen", and "3oue", but also the corresponding spellings with an initial "y" or "g".While such conflated forms make sense from a traditional dialectological perspective preoccupied with explaining variation in the spoken system, it is doubtful whether in trying to explain scribal behaviour and the conceptualisation of a model of spelling to be imitated, a qualitative distinction can be drawn between say scribal preference or mych , moch or much on the one hand and mych , mien or "miche" on the other.What seems to be crucial here is whether the scribe -when tending towards the mean in his mind's eye (cf.Smith 1996) -seeks to adhere only to those elements of orthography which correspond to phonological variation, or whether the mean also incorporates into itself elements that pertain solely to the graphological domain (cf.Mcintosh 1974Mcintosh /1989).It appears likely that observing the orthography of scribes at the graphemic level (see Robinson & Solopova 1993) will increase the degree of variation witnessed and thereby further attenuate the original case for CMS as a historically attested single model.At the same time, however, this methodological decision can be expected to bring to light new patterns of variation at a level which is closer to concrete scribal output, thereby helping us reconstruct less abstract and historically more tenable orthographic models followed by scribes.

CMS in manuscript context: preliminary results
After the orthographic data had been collected, a computation was carried out to establish the majority (graphemic) spelling form of each diagnostic item in each manuscript.A summary of this information for all manuscripts indicates that the most common majority forms for each item comprise myche (simple majority form in 55 of the 68 manuscripts/hands), jfor (54), silf (52), ltjf(48), ony (47), lijk (43), siß (32), and ßouü (27).As illustrated in Figures 1-8 in Appendix III, these forms clearly emerge as the leading ones in the manuscripts.Only for the last item 'GIVEN' do the forms jouen (18) and ßoue (13) appear as substantial rivals.That ßouü was, however, for many scribes the expected written form of 'GIVEN' and not just an abbreviated variant of jouun used at constrained positions which required the saving of space (typically at line-endings) is suggested by its presence as the sole form of 'GIVEN' in no less than 18 manuscripts (see Figure 8).With jouen and ßoue, the corresponding figure is significantly lower -3 and 0 respectively -which may indicate that scribes more readily tended to perceive them as alloforms of a single variant.
If this list of the commonest majority forms is compared with Samuels' diagnostic forms, it will be seen that with the exception of fier they can all be situated under the broadly conceived CMS umbrella.The striking dominance of fier (54) over the expected fijr (6) indicates that individual scribes clearly did not give an identical treatment to all common words for which a potential spelling with a doubled vowel existed, but that their preferences varied from one item to the next.This situation introduces a strong caveat against generalising statements concerning the diagnostic feature in question.
The emergence of a single conspicuous majority form for each diagnostic item investigated might at first glance be interpreted as evidence for a relatively systematic and wide-spread sub-variety of CMS in manuscripts of LV.Such an appearance is somewhat deceptive, however, because the results just discussed do not truly reflect the output of individual scribes in terms of their actual repertories of majority forms.This can be verified by a database query designed to find those manuscripts (hands) in which all eight forms in question (fier, ßouü, lijf, lijk, myche, ony, sij and silfi occur as the majority variant. The surprising result of the query is that among the 68 manuscripts there is only one which fulfils this criterion -a copy of the gospels in BOD Seiden Supra 49.Under closer critical scrutiny even this single finding proves to be somewhat suspect in terms of its relevance and comparability, since the manuscript is one of those two codices which were included in the analysis although the output of two scribes was identified in Matthew.Of these hands, that with the larger proportion of monitored loci (up to 14:30) was analysed.However, as it contains only 42 of the usual 80 (i.e.10x8) occurrences (with only 2 instances of 'ANV, 3 of 'LIFE', and 4 of 'SELF'), it can be suspected that the significandy lower number of instances has created a false impression of orthographic systematicity.To sum up the discussion so far, it seems thus safe to conclude that although the profile of majority forms identified above {fier, silf, lijf, ony, lijk, siß, ßouü) does in broad terms fulfil the criteria for an abstract Type I mean for the present sample of LV manuscripts (and quite likely also for the manuscripts of LV on the whole), it cannot be regarded as a concrete sub-variety of CMS actually attested in any individual scribal profile investigated.
Rather than focusing on majority forms only, the adherence of individual scribes to a standardised orthographic usage can more meaningfully be measured by investigating the degree of variation a scribe exhibits in reproducing the diagnostic items.A database query along these lines shows that among the 68 LV manuscripts examined there is one strikingly uniform instance in which no spelling variation at all can be witnessed among the ten loci of each item.This is a copy of the gospels in CUL Add.6684, a pocket-size book (ca.120x85 mm) written in small textura semiquadrata with a single-column layout.As is often the case for WB, nothing is known about the medieval provenance of the manuscript.The profile of diagnostic items -fyre, ßouü, liyf, liyk, miche, ony, saw and self -indicates that despite the high degree of systematicity in evidence, the orthography clearly cannot be situated even within the broadly defined Type I. Somewhat ironically, the orthographically most highly focused hand in the allegedly archetypal CMS text thus represents a non-CMS usage.
No other profile of majority forms in the material corresponds to that in CUL Add.6684.Between other manuscripts with a highly focused orthography, however, some interesting correspondences can be found.Let us take a look at two examples.
Perhaps the most conspicuous case is that of the New Testament manuscripts Lambeth Palace 1150-51 and Sion College Arc L 40.2/E2, which share an identical profile of both majority and minority forms.The degree of orthographic variation in both manuscripts is distinctively low: fier, 30UÜ, myche, liyf, liyk, ony and silf all occur as the sole forms; for the eighth item 'SAW", say represents 80% of the occurrences, with siß as a 20% minority form.That the two instances of siß are found at identical loci in both manuscripts (Mt.14:30 -Vulgate videns, Early Version seende; Mt. 22:11 -Vulgate vidit, Early Version saß) may bespeak a close textual relationship between the copies (see Benskin 1977: 510 for a similar instance in two mss of the Secreta Secretorum).Discrepancies between the manuscripts in their intralinear glosses and their chapter-internal subdivision of text suggest, however, that it may be premature to view one as a direct copy of the other.A proper collation of the copies is clearly called for.
Another explanation of the identical profiles shared by the two manuscripts is obviously that they do not result from close scribal attention and imitation, but that both profiles represent the output of one and the same scribe.Both manuscripts are written in small well-formed textura semiquadrata/rotunda of a kind broadly characteristic of a large number of LV manuscripts.While a detailed comparison of the graphetic minutiae of the hands on the lines advocated by Mcintosh (1974Mcintosh ( /1989) ) and Doyle (1994) must await, it is worth observing at this point that in both manuscripts the scribe has an unusual preference for writing above the top line of ruling.
Although the two manuscripts vary in their mise-en-page to an extent which does not immediately suggest their production in a single scriptorium or through a shared network of artisans' workshops (cf.Christianson 1999), it is interesting that among the 68 manuscripts examined there is also a third codex which contains an almost identical profile of majority and minority forms -a copy of the New Testament in BOD Laud misc.361.In fact, the only difference between the profiles is that in the Laud copy there is one interfering instance of lik at Mt. 13:44, written in a compressed manner which indicates that the scribe was probably constrained by the line-ending position to produce the shorter lik instead of his regular liyk.That the manuscript seems to bear a close relationship to the Lambeth and Sion College copies textually too is further suggested by the occurrence of its two instances of sij in exactly the same positions as in them (i.e.Mt. 14:30,Mt. 22:11). 6 Among the WB codices investigated for the present paper, there are also two other New Testament manuscripts (BL Harley 4890, BOD Laud misc.388) in which si% is found only at Mt. 14:30 and 22:11 while say occurs in all other monitored loci.Although their majority/minority profiles do not otherwise conform to those of Lambeth 1150-51, Sion Arc L. 40.2/E2 and Laud misc.361, it is worth noting that in all five manuscripts there is in general very litde graphemic variation in the spelling of the eight diagnostic items.
In the second example, a shared profile between manuscripts seems more clearly than in the first one to betoken their common origin.The codices in question are the Bodleian Library manuscripts Bodley 183 and Fairfax 11 -both professionally executed copies of the New Testament, the former with some Old Testament material appended to it.De Hamel (2001: 178) argues that "the two books were certainly written by the same scribe and were apparently decorated by the same illuminator" and thinks it likely that they were "disseminated through the London book trade" (see de Hamel 2001: 179 for a coloured image of the mise-en-page of both mss).Scott (1996: 29) associates the initial and border illumination in Fairfax 11 with the influence of the Southern English "Carmelite large-initial style" of the turn of the 15 th century -a style archetypally represented by the exquisite Carmelite missal which now survives in BL Add.29704-05, 44892 and in other fragments.De Hamel's paleographically motivated argument of a single scribe is strongly supported by the identical majority profiles of the manuscripts (fier, joue, myche, lijf, lijk, ony, say, silfi and the close agreement between the types and frequencies of their minority forms (Bodley: jouen 30%, sai 30%; Fairfax: jouen 40%, sat 20%, liyf10%).Further evidence in support of a single scribe/scriptorium is provided by similarities in the style of catchwords in the manuscripts, their shared complex two-column ruling pattern, and -last but not leastby the strikingly matching measurements between the various proportions of their ruling.While the types of initials used for the lower levels of the ordinatio in these two manuscripts (i.e. for prologues and chapters) do not exacdy correspond -perhaps for reasons having to do with the resources of the customers for whom the codices were madeit is more telling with regard to a common origin of the codices that the line height reserved by the scribe for the initials of these elements of textual hierarchy is identical in both copies.De Hamel (2001: 178) tentatively also associates a third LV New Testament manuscript with the same "commercial production line".In the absence of a systematic graphetic study of the hand of Bodley 665, its similarity with Bodley 183 and Fairfax 11 remains to be properly assessed.The adoption of another type of ruling pattern and the different placement of catchwords, together with a distinctively different orthographic profile with 100% majority forms such as eny and fyke, indicates, however, that the identification of the scribe is by no means self-evident.7

Conclusion
In the pilot study reported in this paper, features of Type I orthography were examined in a large sample of manuscripts of the Later Version of the Wycliffite Bible.The investigation showed that when scribal data for all 68 hands surveyed were combined, a particular spelling variant clearly emerged as the dominant one for each diagnostic CMS item.No single hand, however, contained these dominant forms of all eight items.While the dominant profile consisting of fier, jouü, lijf, lijk, myche, ony, sij and silfcan therefore be regarded as an abstract CMS mean for LV scribes, in accordance with Smith's concept of a focused usage, the alleged status of LV as the prototypical 'invariable' Type I text has to be questioned when variation is measured at graphemic level.
When the degree of variation in the output of individual scribes was gauged, a few instances of highly standardised orthography were found.None of these, however, was stricdy placeable under Samuels' original diagnostic profile for Type I, and the most focused of the usages in CUL Add.6684 fell clearly outside even the more broadly conceived notion of CMS as an abstract superordinate category.
To exemplify the framework within which further investigation of late ME standardised orthographies appears most fruitful, two instances of a shared orthographic profile between LV manuscripts were briefly reviewed.Here analysis of spelling forms was combined with codicological and text-critical concerns to assess the implications of orthographic uniformity for the circumstances in which the codices were produced.The case of Bodley 183 and Fairfax 11 in particular encourages the idea that relationships between orthographic and codicological features are worth pursuing further, and that by such correlation it may eventually be possible to associate particular sub-varieties of CMS with the output of individual professional scribes working in specific historical locales for book-production (cf.Mooney 2000).Needless to say, for manuscripts of widely copied texts such as WB, a thorough exploration of the issue would require a formidable investment in photographic/digital resources to enable meticulous comparison of orthographic and codicological minutiae in manuscripts housed in different repositories.
In the long run, it would be logical to extend these concerns to other texts found in the manuscript contexts of the text first scrutinised, thereby working our way gradually towards larger textual networks which are historically motivated through shared patterns of transmission.I would like to argue that only by means of such a multistage inductive procedure will it ultimately be possible to make sense of scribal patterns of variation identified under the broad CMS umbrella.In the Figures, "g+" stands for 5, "u$" for Ü; and "e$" for é.

1
Black's (forthcoming) well-argued association of the origins of CMS with Oxford introduces a meaningful contextual link between the seemingly disparate Wycliffite and medical/scientific textual components of Type I.I am grateful to Merja Black for kindly allowing me to read her paper in manuscript. 2

Figure 2 :
Figure 2: Distribution of majority forms for 'FIRE'

Figure 1 :
Figure 1: Distribution of majority forms for 'MUCH'

Figure 3 :
Figure 3: Distribution of majority forms for 'SELF'

Figure 4 :
Figure 4: Distribution of majority forms for 'LIFE'