University of Virginia Library

Scribal Self-Corrections in the Thornton Morte Arthure
by
Mary Hamel

—Here endes Morte Arthure writen by Robert of Thornton[1]

The problem of editing Middle English texts is basically one of evidence: if a poem survives in a number of manuscripts, one must choose among alternative readings, basing the decisions made on judgments about which alternative is more likely to be original, which more probably in error. If there is only one manuscript of a text available, the editor may be forced to rely on his own common sense and knowledge of medieval language and style in order to identify errors. In either case the process requires some familiarity with the mechanics of medieval book production and some understanding of the habits of reading and writing characteristic of the age—for example, the fact that a scribe's verbal sense was more likely to be auditory than visual.[2] In short, the editor's task is primarily one of evaluation, by whatever means available or appropriate, of the body or bodies of evidence before him.

In recent decades there has been an increasing emphasis on the analysis of scribal tendencies as a means of evaluating such evidence—on the face of it a sensible approach, since an understanding of the kinds of errors to which scribes were commonly subject, or the kinds of changes they were likely to make, at least establishes a range of probabilities against which a questionable reading may be measured. It appears that generalizations are possible, not only about commonly recognized errors such as c/t confusion and

homoeoteleuton, but also about such things as the tendency of fifteenth-century scribes to regularize or increase alliteration in such a poem as Piers Plowman.[3] But generalizations about scribal tendencies, of course, may or may not be applicable to the tradition of any one text. An example is the alliterative Morte Arthure, which appears in the unique Thornton MS text (Lincoln Cathedral Library MS. 91, fols. 53-98v) to have a number of alliterative irregularities. Whatever one's conclusions about whether those irregularities are authorial or scribal, it appears that the poem's succession of scribes did not participate to any marked extent in the general tendency of their contemporaries to regularize or increase alliteration.[4] For descriptions of scribal tendencies are based on statistical inference from the scribal habits of individuals and are not necessarily applicable to every individual. On the other hand, intensive analysis of the characteristics of individual scribes is in most cases of limited usefulness, because of their normal anonymity; Scribe A's habits in transcribing Text X are relevant only to that text, if no other transcriptions can be identified as his.

But there are cases in which the habits of an individual scribe are significant, cases in which a known individual has copied a variety of texts, and in which therefore the conclusions derived from his transcription of one text may be useful in evaluating his transcriptions of others. Such a scribe is Robert Thornton, lord of East Newton in Ryedale, in the North Riding of Yorkshire, in the earlier fifteenth century.[5] Although apparently an amateur, he made copies of many important texts. The two extant volumes known to have come from his pen, the "Thornton MS" itself and BL Addit. MS 31042, contain many romances as well as other texts both devotional and secular; the scribe is a factor to be reckoned with by students of such texts as The Awntyrs off Arthure, Sir Degrevant, Sir Eglamour of Artois, Thomas of Ercildoune, The Parlement of the Thre Ages, Winner and Waster, The Siege of Jerusalem, Richard Rolle's treatises and poems, and a miscellany of other works.[6]

The importance of this scribe, and the recent publication of the Scolar Press facsimile of the Lincoln MS (see note 5 above), have led to increased interest in him in the last few years. The exhaustive researches of George R. Keiser, for example, establish Thornton as "a man of some moderate prestige and influence," who "had a fairly ordinary life as an active, though by no means a leading citizen of Ryedale and the North Riding"; more important, he was "an avid reader and a man of industry, enthusiasm, and ambition" whose access to texts available in his region must have been at least partly the result of active search and inveterate book-borrowing.[7] Other scholars, however, have been led by the careful organization of Thornton's two anthologies, the general "expertness" of his scribal operations, the surprising skill of some of his decorations, and the mere bulk of his material to question his amateur status.[8] Nevertheless, the birth-record of the scribe's grandson Robert "in Ridayll anno domini m cccc liij" on fol. 49v of the Lincoln MS makes the identity of the scribe Robert Thornton with the landholder and active country gentleman Robert Thornton more than merely plausible (Keiser, p. 159), and the scribe's amateur status must be accepted. It appears, in fact, that Thornton was an amateur in the best sense of the word, loving his avocation enough to train himself in the requisite skills and to employ them conscientiously, if not always consistently.

Indeed, it may be because Thornton was not a professional scribe that he stands out as an individual, and it is appropriate that at least one of his compilations has come to be known by his name. The history of the "Thornton MS" is closely connected with that of his family, for it remained at East Newton until the late seventeenth century.[9] The selection of works included shows that Thornton was copying texts for his and his family's private use: it is just such a compendium of entertainment, instruction, religious edification, and practical advice as one might expect to find in the domestic library of a moderately bookish fifteenth-century gentleman's household. And, from the marginal scribblings and evidences of early mending, it was clearly well used as long as Thornton's descendents were capable of reading his language.

One might expect, then, that the scribe, not intending public circulation, would exercise considerable freedom in adapting the texts he copied for personal use. Yet several editors of Thornton texts have attested that, dialect aside, the scribe exhibited a remarkable fidelity to his texts.[10] And even in

terms of dialect, he only "half-transform[ed] his original," adding "a Yorkshire veneer" to texts he copied from other dialectal regions.[11] Certainly, in preserving archaisms and rare words, including technical terms presumably outside his experience, Thornton seems to have been unusual among scribes. One might take as example the lively scene of preparation for the sea-battle at Southampton near the end of Morte Arthure (lines 3599-3711); here the poet's own familarity with and fondness for the details of shipping and ship-battles come through unimpaired. But this passage is interesting not only for its portrayal of topcastle and stem, sail and hatch, but even more for the terms inaccurately reproduced by the scribe—e.g., bilynge for "bilge" (3663), vpcynes for "upties" (3675).[12] Unlike many other scribes, Thornton seems not to have felt the need to rationalize or emend corruptions in his copytext; there are a number of places where he appears to have contented himself with reproducing words that looked as though they might mean something.[13]

In short, Robert Thornton's status as a country gentleman, devoutly religious, head of a substantial family, may have had some influence on his choice of texts to copy; certainly his education and experience must have influenced his treatment of the texts; but whatever his activities or background,

he appears to have exercised an extraordinary care in reproducing them, a care informed not only by his sense of their interest or usefulness but also apparently by a respect for them in themselves. Thornton's carefulness may have been due at least partly to his amateur status; lacking professional training as a scribe, he was free from the prescribed routines, techniques, and conventions of the scriptorium. What this means for the scribe's practice is that he was forced to rely on his models, his copy, as the only available guides to the representation of text. He was more liable than a professional scribe would have been to be influenced by the letter-forms, the spellings, the abbreviations, the punctuation, the line-divisions he saw before him, less prone to translate such incidentals of transcription into his own (or his shop's) systems and conventions.

This tendency has led one or two of his editors to see more than one scribe at work in the transcription of the two MSS. L. F. Casson, for example, the editor of Sir Degrevant, found a "considerable variation" in the handwriting of the Lincoln MS as a whole and cited the opinion of Eric G. Millar that the London MS was also probably in different hands. Millar, recognizing that "apparent variations in size or general appearance may be due only to a change of pen or a different quality in the ink," based his judgment primarily on the increased use of thorn and a "rather different" et-symbol after fol. 66v (Casson, p. ix). But with an amateur scribe like Thornton one must also take into account the fact that fol. 66v in the London MS marks the beginning of a new work, The Sege of Melayne; it is probably the scribe of Thornton's copy-text for this work who was responsible for its graphemic differences from the preceding material.[14] A similar circumstance may be found in the Lincoln MS; in the texts of the Prose Life of Alexander (fols. 1-49) and Sir Percyuell of Gales (fols. 161-176) the older form of thorn is used (that is, it has a straight descender without a flourish), whereas elsewhere in the MS thorn is normally indistinguishable from y; the difference once again is presumably due to the usage in Thornton's copy-texts for these works.

Thus, although the scribe's lack of professionalism might be seen as a reason for distrust of his texts, his very inconsistency of practice, and the dependence on his models that it implies, may offer the editor fuller access to previous stages of transmission than a professional scribe's systematic habits and conventions normally afford. This is not to say, of course, that Thornton was a scribe without habits and systems of his own; all his texts show evidence of a spelling system strongly influenced by his own dialect, for example. Nor can it be supposed that his texts are error-free; his lack of professional training did not prevent him from sharing in trained scribes' mistakes and misjudgments, not only reproducing those of his copy-texts but also contributing a number of his own.

It may be objected that it is a task of formidable difficulty to distinguish the errors of one scribe from those of his predecessors in the transmission of

a text. One aspect of Robert Thornton's care as a scribe has not been remarked by his editors, however: that is, his alertness in correcting his own errors.[15] In my own transcription of Morte Arthure, for example, I have found several hundred self-corrections. Many of these are quite minor, yet this tendency affords an unusual opportunity: to define a scribe's characteristic errors in terms of what he himself saw as error in his transcription. Although such an approach must leave out of account unrecognized errors, it does at least distinguish Thornton's errors from those of his predecessors.

Or does it? One question that must be considered in such an investigation is the extent to which such changes represent actual corrections—that is, reassertions of the copy-text's forms and spellings—and the extent to which they represent deliberate alterations influenced by the scribe's own dialect or by his own judgment about correctness, explicitness, intelligibility, emphasis, or elegance.[16] Definitive answers on questions of dialectal change must await the publication of McIntosh and Samuels' Survey of Middle English Dialects, although Professor McIntosh has provided a few clues in his 1962 article (see note 11 above). Until then one must depend on hints and guesses gleaned from more traditional and less precise surveys.[17] Other Thornton texts also provide some guidelines, particularly the "Thornton Medical Book," which appears to represent the scribe's own dialect more consistently than any other of his texts.[18] In spite of such uncertainties, it may be significant that, although some of these corrections do show apparent dialectal influence, none of them shows a deliberate change from the Midland <o> spelling of OE ā to the Northern <a>. The question of editorial (as opposed to dialectal) change may be more accessible to analysis.

The purpose of this study, then, is by the detailed analysis of Thornton's self-corrected errors in Morte Arthure to derive conclusions about his characteristic errors, tendencies to err, and possible editorial tendencies, as well as about his habits and systems of transcription. Conclusions about the scribe's practices are supported by a statistical analysis of frequency of error at various points of the alliterative long line, which also identifies patterns of weakness within the line. The conclusions reached, one hopes, will be of use not only to students of other Thornton texts but also to critics of Alliterative Revival texts more generally; for Thornton's patterns of error are probably not unique to him.

An understanding of Thornton's tendencies to err depends in part on familiarity with his characteristic handwriting, which, like other hands of the period, was likely to give rise to certain kinds of error. In Derek Brewer's words (p. vii) Thornton's is "a fairly typical mid-fifteenth-century cursive hand" which "varies considerably." The New Paleographical Society notes that it "varies a great deal, but is at best a carelessly formed cursive type of script."[19] Confusion of minims in m, n, i, and u was inevitable in any hand of the period and in Thornton's case may also include w-errors; although it was quite different in form, he tended to use w interchangeably with u, especially in the ou digraph. The grapheme <y> is also ambiguous, standing for both y and þ, but it is nevertheless seldom confusing since it is ordinarily clearly distinguishable by context and position. With the meaning þ it occurs with few exceptions only initially or (when medial) before a suspension (e.g., þow 'thou,' oþer 'other');[20] with the meaning y (i.e., /i/) it is normally used only medially or finally (again with a few unambiguous exceptions, e.g., ynowe 'enough'). The scribe of Thornton's copy appears to have been less careful in distinguishing y and þ by position, however, and a few cases of medial þ without a following suspension do crop up. Thornton never uses <y> for /j/, preferring <3> for that sound in all positions (thus <yow> is always "thou," while "you" is spelled <30w>). That this sound was spelled earlier in the tradition of this text with a thorn-like y, however, is shown by the uncorrected form 30fe for þofe in lines 1938 and 2854.[21]

Other closely related letter-forms are f/long s, c/t, and o/e. So alike in form are c and t, in fact, that it is often difficult to decipher which letter the scribe himself intended.[22] Beyond these easily confused forms, there are other pairs that are similar enough to cause occasional confusion, especially if carelessly made: y/r, h/b, initial v/d, final e/sigma s, and gg/w or dd/w. One final peculiarity of letter-form may be noted: like other northern scribes, Thornton often used 3 to indicate a final sibilant (here transcribed as z); otherwise, he normally chose the sigma-form of s finally, long s initially or medially.

The scribe used standard suspension-symbols for -er, -es, -ur, -us, and nasals; the nasal suspension occurs (at times superfluously) in a variety of forms, straight, curved, or looped, sometimes dotted and sometimes not. One must note also Thornton's often-remarked habit of adding apparently superfluous flourishes to final nasals and of crossing doubled ll and (less frequently) th. Although early editors interpreted these marks as suspensions

for a following -e, the more recent consensus is that they are usually meaningless.[23] Final -e after d or g was, however, indicated by a superscript dagger.

The Corrections

Thornton used three methods to correct his transcription of the poem: first, he might strike out a word or part of a word and (if the error was not simple dittography) immediately thereafter write the preferred form. Second, he might superimpose a preferred letter over a rejected letter in spelling corrections (this habit has led to some problems of legibility) or, in a few cases, convert one letter to another (e.g., r to y). These two methods were obviously part of the immediate process of transcription; the third method, in which corrections were inserted above or at the end of the completed line, may indicate either an immediate review of each completed line or a later proofreading effort. When this method was used for omitted words, the point of omission was marked by short double virgules or the more familiar angled caret-mark slightly below the line, the mark repeated under the word to be inserted (if at the end of the line). Caret marks (in either form) were also used for word- or spelling-changes by the insertion method. Two of these end-of-line corrections, the insertion of ayer in line 455 and the correction of lordez to londes in line 878, were performed in a hand different from that of the text—one more carefully formed, with letters more uniform in size, and with book rather than cursive e.

The text contains no corrections of large-scale error such as omitted or repeated whole lines or garbled passages; presumably if the scribe did recognize and decide to correct such errors he recopied the page. One may infer from the size of his two volumes that he had access to a good supply of paper.

In the following summary, Thornton's self-corrections are divided into two major categories, spelling-changes and word-changes. Within these broad categories items are further subdivided into corrected omissions (in which the scribe has added a letter or word), corrected intrusions (in which he has deleted a letter or word), and alterations (in which he has exchanged one letter or word for another). Within these subcategories, further subdivisions emphasize particular tendencies. A third major category then summarizes corrections involving possible suspension errors. Since this system of classification relies on greater certainty about the actual nature or cause of the error corrected than is sometimes available, a number of items are included in more than one category or subcategory (indeed, a few items show multiple correction); the final "Statistical Summary," however, counts each item only once.

All examples below are cited from the MS without emendation; Thornton's

somewhat erratic capitalization is preserved, except that all line-initial letters are presented as capitals; long i is presented as j except when capitalized as I; y and þ are distinguished; all suspensions are expanded in italics. Although all items are cited from the MS, they are identified for ease of reference by the editions' line numbers rather than by folio. All available editions include notation of folio pages for those who wish to refer to the facsimile. It should be noted, however, that some of these corrections were made in very light ink; for this and other reasons, some corrections may not be visible in the facsimile.

For the examples also, an attempt is made to indicate the method of correction used by the scribe. When a single (right-hand) bracket is used, the element to the left of the bracket was crossed out and the element after inserted, either immediately, above the line, or at the end of the line (the last indicated by an asterisk *). When double brackets include two letters joined by an arrow, the second letter was superimposed over the first (e.g., [r→y]). Double brackets including single elements, in the case of corrected omissions contain inserted material, and in the case of corrected intrusions contain deleted material.

I. Spelling changes

• A. Spelling errors by omission of a letter or letters (77 cases).
• 1. Consonant omissions or omissions from consonant clusters (37 cases). Examples; el] Erledoms 42; child[r→y]ren 3208; filnesql filsnez 881; ch] schotte 2467; vescowte] vescownte 1984; schok[e→k]e 4114.
• Of these errors, thirteen involve omitted liquids (10 r, 3 l); four, the /š/ phoneme; four, omitted nasals; and seven, failure to double consonants after short or lax vowels.
• 2. Omitted vowels (25 cases). Examples: kenly] kenely 1243; berynge] beryenge 2377; br] burneste 3846; malez] maylez 1857; chanse] chawnse 2368.
• Of these, four involve metathesis of liquids; six, the ai/ei digraph; nine, the au/ou/eu digraphs.
• 3. Omitted syllables (9 cases). Examples: wyrp] wyrscheppez 1059; peste] potestate 2327; wre] wandrethe 2370.
• 4. Preterites and participles (6 cases). Example: regnd] regnede 2034. Like the example, most involve the omission of an inflection-vowel.
• B. Spelling errors involving intrusive letters (44 cases).
• 1. Dittography or contamination (19 cases). Examples: swether[w]yke 47 (correction in light ink); Wa[y→w]ayne 1302; hawrawde] harawde 3029; Etten] Eten 'eaten' 2716.
• Of these, five involve doubled consonants after long or tense vowels.
• 2. Other intrusive letters (11 cases). Examples: swyche] syche* 76; charye] chare* 1886; togers] Toges 3189; browndys] brondis 3695.
• 3. False starts (14 cases). Examples: aye] ayers 1259; ti(er?)] title 2363; sha] hade 2472; s] sette 4043.
• C. Spelling errors involving the exchange of letters (108 cases).
• 1. Hesitations between possible variant spellings (76 cases).
• a) e/y hesitations in unstressed syllables (18 cases). Examples: ned[e→y]s 470; capteefes] captyfis 2340; brawnd[e→y]sche 3359; h[e→y]s 384, 3572.
• Four of these involve unstressed function words like the last example above. Of the total, all but three change e to i/y.

---

128




• b) e/y hesitations in stressed syllables (19 cases). Examples: l[e→y]gmane 420; re] ryuer 920; f[e→y]ste 2233; r[e→y]alle 3373; th[e→y]then 4345.
• Again, the scribe's alterations consistently prefer i/y to e. Some of these changes may reflect etymological variants ("liege," "royal"); a substantial proportion of the rest involve open-syllable phonological change, but four cases are closed-syllable.
• c) Special cases: "high" (6 cases), "through" (2 cases), "eyes" (2 cases), "rescue" (3 cases). Examples: h[e→y]e 620; th[o→r]ughe 1857; eyne] eghne 2962; bowez] bowghez 921; resch] rescows 4137.
• d) Others: a/e/ai respellings (6 cases), e.g., heþe] haythen 2974, 3arnes] 3ermys 3911, aventall] aventaile 2572; o/u/w respellings (7 cases), e.g., auke] awke* 13, co] cunvayede 482, trouthe] trowthe 2585; consonantal respellings (11 cases), e.g., seluen] selfen 1210, wynnez] wynes 1028, j] gesserawnte 2892, lik] lygham 3281, schragkys] schragges 3473.
• 2. Misreadings and other errors (32 cases).
• a) Misread letter-forms: e/o, e.g., be] bot 4070; e/sigma-s, e.g., Romayn[e→s] 2234; f/long s, e.g., f] syghte (with sigma-s) 1949; h/b, e.g., he] be 3190; n/u, e.g., 30w] 3one 2726; y/r, e.g., Eyrthe] erthe 527; y/þ, e.g., no yng] nothynge 2442.
• b) Contamination: e.g., Raylide with reche[d→s] 3263 (the sigma-s does not obscure the d, leading to editions' misreading).
• c) Transliteration or metathesis: e.g., ffor] fro 1698 (plus two other line-initial cases); ax] aske 2350; many] man 'main' 4071 (y erased).
• d) Verb forms: e.g., likyde] lykes 140, sai[d→s] 868.
• e) Others: r/l, i.e., her] helpe 346, war] walkes 946; u/y (one example), th[u→y] 4323 (perhaps confusion of þu and þi); anomaly, schikf] skilfull 1561.

II. Whole-word changes (111 cases).

• A. Omitted words.
• 1. In stress position (25 cases). Examples: That Caerlyon was [callid]* : with curius walles 61; The [redyes them stricken] romaynes redyes þan 1427.
• 2. In unstress position (28 cases). Examples: to [þat] alde wyf 986b; I witter þe [þe]* Emperour 1239a; Gawayne the gude : [ha stricken] he has 3706.
• B. Intrusions (22 cases). Examples: j not [watte] watte it ment 977; sir Ewayne [sir] fytz vriene (in light ink) 2066; lady[ne] : ne 3081 (the dele marks read as long -ss- by the editions). Almost all cases in this category are simple dittography errors.
• C. Word exchanges (36 cases). Examples: hyghe] lowde 124; þe] 30w 225 (in light ink); knyghtez] kynges 523; honoure] one 704; treson of lordez] lōnmacrdes 878; Enamellede] ennelled 1294; spourres] speres 2542; full] so 2571; neuen] mene 2869.
• Among these, we might note three cases of "knights/kings" confusion (523, 3687, 4169), all at line-end. Including these, more than a dozen cases apparently involve confusion of alliterative formulas, especially at line-end.

III. Suspensions.

• Of the examples collected in parts I and II, 20 may have involved confusion of nasal suspensions; for example the correction luppe] lumppe] 1814 (I.A.1) may reflect the exemplar's lūppe; the correction chanse] chawnse 2368 (I.A.2), the exemplar's chāūse. Two additional corrections also reflect hesitation over nasal spellings: hemmes] hēmmacres 3253 and reng] rēgmacrne 4005.

  ---

  129

  
  
  
  Six of the corrections discussed above may have involved the -er suspension, and there are in addition two cases in which a superfluous -er suspension was cancelled (e.g., couererde] couerde 3633). Six further cases may have involved the -ur suspension, one the -us, and two general contractions (e.g., peste] potestate 2327, I.A.3). An interesting case is the word-exchange honoure] one 704 (II.B); since the upper limb of the 7-shaped -ur suspension was horizontal, it could be confused with the nasal suspension, and one may infer that the exemplar had onmacre, first interpreted by the scribe as onure.

Statistical Summary

This survey has dealt with 339 cases of corrected error: 229 spelling changes (including suspensions) and 110 word changes (additions, deletions, and alterations); these counts avoid duplication and may differ from those above. These corrections are distributed fairly evenly throughout the text; the only striking imbalance is in word-additions and -deletions, which occur twice as often in the first half of the text (1-2199) as in the second (2200-4346).

Following is a tabulation of corrected errors by position in the alliterative long line; it shows some interesting patterns. In this analysis, A is the first stave, B the second, C the third, and D the fourth, while unstressed words are marked x. In the two or three cases of triple first half-lines (i.e., a a a : a x), the words are identified as A or B according to the principal stresses in the half-line. The x-positions, of course, may include more than one word; although six cases are identified of word-omissions before the first stave, in no case was an initial word omitted.

Statistical Summary

                       

Correction Types	x	A	x	B	x	:	x	C	x	D	x		Totals
I. Spelling Changes
A. Additions	1	15	0	22	0	:	0	24	0	13	1		76
B. Deletions	3	8	3	10	2	:	0	7	1	10	0		44
C. Exchanges	7	27	7	30	0	:	5	12	0	17	0		105
D. Suspensions	0	1	0	1	0	:	0	1	0	1	0		4
II. Word Changes
A. Additions	6	9	14	12	0	:	4	4	4	0	0		53
B. Deletions	3	3	1	2	2	:	1	4	3	3	0		22
C. Exchanges	(3)	1	5	2	0	:	3	6	2	16	0		35
	---
Totals	20	64	30	79	4	:	13	58	10	60	1		339

In line IA, spelling changes by addition, one may note that all but two of these corrections occur in staves, and of these two exceptions the correction in final position bears secondary stress. Even such function words as wyll (2664) and also (3485) are in stress-position. The middle staves, B and C, appear slightly more liable to this kind of spelling error (omission of letters) than A or D.

Line IC, spelling changes by alteration, shows that such corrections occur more than twice as often in the first half-line as in the second, an

imbalance particularly noticeable in dialectal hesitations. It may be that Thornton was simply more likely to make such changes in the first half-line, rather than that the first half-line was more liable to error.

Line ID includes only those suspension-errors in section III that are not included in other sections.

Line IIA, additions of omitted words, shows that omissions are over three times as likely to occur in the first half-line as in the second. For unstressed words, the position between A and B is particularly vulnerable; for stressed words, the B-stave.

The tabulation for deletion of intrusive words in line IIB is somewhat misleading, for there are five cases of cross-caesura contamination—three in which an A or B stave is repeated at C, two in which an unstressed word occurring immediately after the caesura is prematurely inserted (that is, the caesura is misplaced).

Line IIC, corrections of mistaken words, shows that such errors are three times as likely to occur in the second half-line as in the first; in particular, the number of corrections in the fourth stave (D) is striking. Nevertheless, for unstressed words the position between A and B again shows the highest number of corrections. The three line-initial for/fro metathesis errors, already tabulated in line IC, are included in parentheses here but are not counted in the totals.

Of the long line as a whole, the first half has a slight preponderance of corrected error (58%), with its second part particularly vulnerable. And a stressed word is over three times as likely to show an error of some kind as an unstressed word. Yet the highest incidence of whole-word errors occurs in the vulnerable unstressed position between the first and second stave.

Conclusions

The Text of Morte Arthure: Direct Evidence

The examination of the scribe's corrections provides certain kinds of direct evidence unnoticed by previous editors of the poem. The most obvious kind of direct evidence is the corrections that have simply been overlooked, because they were made in light ink, because a caret-mark was misplaced, or because the scribe's habit of correction by superimposition has misled the transcriber. Lighter-ink corrections previously missed include these: the second w cancelled in Swetherwyke (47; MS reading Swetheryke); þe crossed out, 30w inserted after it above the line (225)—thus there is no need to emend to 30w þe, for the MS reading is simply 30w; the second sir-contraction in sir Ewayne sir fytz vriene (2066). A special case is the scribe's correction of ladyne (3081), a misdivision that also violates the caesura; the two diagonal slashes with which he crossed out ne look rather like doubled long ss, and a portion of the e remains visible, leading to the editions' reading of ladysse (the MS reading is lady). Then, editors have felt the need to defend their

transposition of his in (1797),[24] not recognizing that the misplacement of his is no more than a misplaced caret-mark in the correction of an omitted word. Examples of misleading corrections by superimposition are ryuere for reuere (424; here the y does not quite overlap the e, leading to the editions' reading reyuere); hyghe for heghe (2651; not heyghe); reches for reched (3263; see above, I.C.2.b); leders for ledars (3832).

A second kind of evidence comprises clues to the exemplar's form obscured by incomplete corrections. An example is the scribe's failure[25] to complete his task in correcting treson of lordez (878; see II.C above). The error was one of transposition, not entirely corrected by the change of lordez to londes; the original phrase occurs in line 991, tresour of landez (with Thornton's Northumbrian a replacing the Midland o of the line 878 correction). As another example, the scribe's omission of romaynes in line 1427 (II.A.1) was corrected by the deletion of the premature redyes them, a phrase retranscribed after the insertion as redyes þan; the emendation of þan to them thus needs no defense. In line 4071, the partial correction of the transliteration many by the erasure of y justifies the emendation of the resultant man to mayn 'main.' Finally, the correction of togers to Toges in line 3189 justifies a similar correction of the togers of line 178.

A third kind of evidence from corrected errors is the confirmation of questioned readings. Filsnez (881) was queried by some early commentators before the OED identified this unusual form's ON root; that the scribe questioned it too and confirmed it from his exemplar is demonstrated by his correction of the error filne.[26] A similar case is the scribe's line-end correction of charye to chare (1886); although the word is difficult (the MED's treatment, s.v., is conjectural), emendation to carye or care would be precipitous.[27] On the other hand, the form valewnce (2047) has not previously been questioned; editors have interpreted the word as "Valence," a place name (cf. valence, line 41). But the scribe rejected his first spelling valence in this line, presumably respelling the word from his copy-text; the original reading here may be valewe or even valiaunce (the exemplar's form was possibly valeunce). Finally, in view of the scribe's correction of forssy to forsesy in line 3300, should one emend the result to forsy (as Krishna does), when the correction seems to insist on this odd spelling?

Indirect Evidence

Before we come to Thornton's own scribal characteristics, it is worth looking at the evidence the corrections give us for the characteristics of his copy-text for this poem.[28] We can infer from the errors and their corrections that the hand of the previous scribe was a similar cursive script, at least by the evidence of o/e confusion (these graphemes are less likely to be mistaken in book hand). Besides the universal scribal problems of minims (m, n, u, i), c/t, and f/long s, the exemplar also had similar pairs h/b, r/y, e/o, e/sigma s, and identical y/þ. Moreover, the copy-text seems to have used similar suspensions and to have made similar suspension-errors—in particular, to have been similarly careless about nasal suspensions. On the other hand, Thornton's preference for w-spellings of the au/ou digraphs seems not to have been shared with his predecessor, a fact which may increase the likelihood of u/n error. The previous scribe also appears to have been more willing to use þ medially without a following suspension; the spellings heþen (e) (3687, 3704) and seþen (1977) are probably his. Dialectal characteristics suggest that other corrected forms taken from the exemplar include beryenge (2377; I.A.2 above), wenge corrected from wyenge (4002), and—a special case—both the error swyche and its correction syche (76; I.B.2 above).[29] Finally, although the conclusion is not based on Thornton's self-corrections, there is some evidence that the previous scribe's hand was uneven, as in the uncorrected error Cyruswitrye for cyrqwitrye (2616), in which the q had apparently crept far enough above the line to be misread as the -us suspension.

Thus, when speaking of the scribal tendencies in Thornton's copy of Morte Arthure, one should keep in mind that many of them were shared with his immediate predecessor (as well as late medieval scribes more generally). Nevertheless, one may legitimately infer from the foregoing analysis certain habitual procedures on the part of Robert Thornton.

The broadest conclusion one may draw is that Thornton transcribed, not word by word, but line by line. This conclusion is based primarily on the patterns of error shown in the "statistical summary," especially the greater tendency for the omission of words in the first half-line and the greater tendency for the wrong word to be copied in the second half-line, especially the last stave. What these patterns appear to show is a scribe hurrying through the first half-line in his writing in an attempt to complete the line while his

memory of the whole was fresh—but not always succeeding. The tendency to mistake the final stave would be aggravated, of course, by the poet's frequent dependence on almost interchangeable formulaic tags in the second half-line; as mentioned earlier (II.C above), a number of these errors can be attributed to Thornton's confusion of tags.[30] An additional factor is the normal absence of alliteration in the final stave.

While transcribing the line, Thornton frequently performed his corrections in the process of writing, by superimposing a preferred letter or by crossing out and adding the correct (or an omitted) form immediately; many of these corrections may have been made without rechecking to copy. Twenty-five or so of the corrections, however, were performed after the transcription of the line was completed, for they were inserted at the end of the line with the deleted error or a caret-mark or both indicating the appropriate place. Among these, of course, are the two corrections by the second hand mentioned earlier.

We can see the scribe, in summary, reading a line in his exemplar and committing it to short-term memory, transcribing it with immediate corrections, occasionally checking the exemplar in the process, and then briefly rechecking it before going on to the next line. That brief look back, of course, while it helped him to correct error, may also have helped to introduce new error in the form of contamination from neighboring lines. One cannot claim that the scribe's attention to correctness in transcription was always at a high pitch, or that he always rechecked the exemplar before going on, in any case. There are too many errors remaining in the text, and Thornton undoubtedly committed his share of these.

The Scribe's Tendencies to Error

We may first take up spelling elements that appear to have caused particular problems for the scribe. The problem of nasals (22 errors) should occasion no surprise, since the ambiguity of minim-graphemes and the availability of alternative means of representation (i.e., suspensions) must inevitably have caused confusion. Four of these corrections, however, may show some uncertainty about the distinction between the alveolar and velar nasals (nayknmacrgmacr] nakyn 2350, rynge] ryngne 3214, lenghne] lenghe 3265, reng] regne 4005).[31] Problems with liquids (24 errors, including omission, metathesis, and r/l exchange) were perhaps more closely related to speech patterns, in view of the importance of auditory memory in scribal habits.[32] Among other consonants, although only four corrections involve ch/sch spellings (bewch] bewschers 1047, basche] basschede 2121, ch] schotte 2467,

Iriche] Irische 4123), in the text as a whole the scribe frequently hesitated over such forms (cf. charpe 3600 vs. scharpe 3842, ravichse 3539); it is difficult to know whether the ch spellings were the scribe's or his exemplar's.[33] Other consonantal respellings appear to be related to adjacent vowels: in fact, Thornton seems to have had some difficulty spelling for vowel length, with five corrections of single consonants after short vowels, five corrections of doubled consonants after long vowels, and two corrected omissions of lengthening-vowels (kenly] kenely 1243, knes] kneys 4274).[34]

Turning to vowel problems, we find nine cases of hesitation in the spelling of the ai/ei diphthong, usually resolved in favor of the exemplar against the late ME tendency to level the diphthong in unstressed syllables (Jordan, §247) and perhaps the tendency in parts of Scotland and Yorkshire to level it in stressed syllables (§§132, 233 Rem. 2). The situation with o/u/ou hesitations is a little more complicated. Six corrections simply show the scribe's preference for w-spellings where u might be ambiguous; but six other ou/ow spellings correct initial errors in u or o (four in tonic syllables, two in atonic). The one or two a/au hesitations may also reflect leveling of a diphthong, of which there is some evidence elsewhere in the text, though au forms predominate (see Jordan, §§240, 287).

Most striking among the vowel-changes are the large number involving e and i/y. In the fourteen cases of e/y change in unstressed syllables, the scribe's preference for i/y spellings (in four verb inflections and four noun plurals, among other forms) is predictable for his own northern dialect (Jordan, §135); but in the text as a whole he by no means converted all such unstressed e's, inflectional or otherwise, to y's or i's.[35] Indeed, the twice-repeated correction of hes to hys (384, 3572) rejects a peculiarly northern form (see MED, s.v. "his" pron. [1]) that appears nowhere else in the poem.

Although the e/y changes in stressed syllables also show a consistent preference for i/y spellings, the dialectal evidence is harder to interpret. On the one hand, nine of the rejected forms (e.g., reuer, menystre, etc.) show the Northern lowering and lengthening of i to /e:/ in an open syllable (Jordan, §§36, 226), reversed by the scribe in his corrections; such spellings do occur elsewhere in the text, however. Four of the rejected forms, on the other hand, show e spellings in closed syllables (e.g., fefty, messe-do), corrected

by the scribe to y; if the rejected forms reflect a later ME lowering of i to e in closed syllables south of Humber (Jordan, §§36, 271), they presumably were found in the copy. I have found little evidence of such lowering in other Thornton texts, and in Morte Arthure there remain only scheftez, en (2456, 3627), and possibly one or two others.[36]

Treatment of other dialectal forms seems to be similarly inconsistent. Though the two corrections for the word "eyes" (eghne 2962, eyghen 3985) seem to reflect Thornton's own preference—the -gh- form predominates not only in this but in other Thornton texts[37] —several of the corrections for "high" reinstate the exemplar's form hye (e.g., 620, 1646). A good example of the scribe's dialectal indecisiveness is the corrected form childyren (3208, corrected from an initial childr-), a kind of compromise between the northern childir, -er and the more southerly children (the northern form dominates in this text three to one).

To sum up, although a number of spelling changes show a predictable preference for the spellings of the scribe's own dialect, not all of them do so, by any means. Indeed, the changes he makes are surprisingly limited; there is little evidence in the corrections, for example, of o > a change (in forms either from OE ā or OE a plus a lengthening group), yet both o and a forms are found uncorrected in the text.[38] The largest group of dialectal vowel-changes, in fact, is that involving e > i/y correction, and as we have seen the dialectal direction of these changes is mixed. It seems probable that the large number of e/i/y changes among these corrections reflects the scribe's own uncertainty in view of the phonological changes occurring in this and other dialects at this period; some spelling decisions elsewhere in the text, for example, appear to have been influenced by the raising of long tense e (e.g., bieldez 'dwell' 1242, chiftayne 2732) and the lowering and

diphthongizing of long tense i (e.g., weysse 'wise' 2514, theyn 'thine' 3403; see Jordan, §§277, 279). These and certain other phonological changes must have complicated the scribe's choices considerably. The evidence of other dialectal spelling changes is also inconsistent, however; yet Thornton's basically conservative approach is shown by such corrections as that of thys to thes rather than thir "these" (52), and that of swyche to syche (76), apparently reproducing his copy's own correction (see note 29 above).

One must also take some note of spelling changes that may be attributed to common scribal errors, often described elsewhere: dittography, contamination, transliteration, confusion of similar letter-forms, and particularly suspension-error. Thornton also shared with other scribes, amateur and professional, the tendency towards overleaping of phrases, words, and parts of words, and towards carelessness with familiar literary formulas. But there is little evidence in these self-corrections of deliberate changes in the directions Kane (Piers Plowman, pp. 125-128) pointed out of the more explicit, "correct," intelligible, emphatic, or elegant statement. It is possible, of course, that at least some of Thornton's word-alterations reflect his judgment of appropriateness rather than his review of the copy-text; for example, the substitution of hym for he in line 2227 may be an attempt to clarify a somewhat confusing encounter between the Emperor Lucius and sir Lionel in battle—but even with the substitution it is still not clear who struck whom in this line. On the other hand, the evidence of such corrections as jrrtayne (575), valewnce (2047), chare (1886), and filsnez (881), words which were obviously unfamiliar to the scribe, argues the priority of fidelity over intelligibility in his eyes.

Finally, we may identify certain patterns of error. Certainly in this work, probably in other alliterative texts from Thornton, perhaps in alliterative texts from other hands, one is more likely to find:

• 1. Omitted letters in staves than in unstressed words;
• 2. Omitted words (stressed and unstressed) in the first half-line than in the second;
• 3. Respellings in the first than in the second half-line, and in staves than in unstressed words;
• 4. Wrong staves in the second half-line than in the first, and especially in the D position.

And as a broad generalization, one may say that, at least in Thornton's hands, the weakest part of the alliterative long line—the part most subject to error—is the latter part of the first half-line, the B-stave and the unstressed word or words that precede it.

Applications

The real usefulness of this kind of analysis will perhaps become apparent only when its conclusions are applied in the evaluation of uncorrected error in the text under study.[39] One hardly needs, after all, to know that the scribe

had corrected a certain number of transliteration errors to emend fonode (205) to foonde, swrede (47) to swerde. But, on the other hand, familiarity with the scribe's problems with the ai/ei diphthong—the errors he corrected tend to show the omission of one or the other vowel-element—may cast some light on a form like wye 'weigh' (740), which occurs as the C-stave in its line, a position in which Thornton corrected some twenty-four letter-omissions. Thus, although the OED records this spelling as a Scots and Northern variant of the fifteenth-sixteenth centuries (citing only this instance, however), the emendation weye is fully justified by the patterns of error shown in Thornton's self-corrections. The greatest benefit available from the study of the scribe's self-corrections, in short, is simply familiarity with the scribe's habits of mind and pen. Familiarity of this kind allows the editor to focus more clearly on the causes and kinds of error in the text he must deal with and to deal with them on a more rational basis.

Notes