University of Virginia Library

Search this document 


  

expand section 
expand section 
expand section 
expand section 
expand section 
expand section 
The London Thornton Manuscript: A Corrected Collation by Ralph Hanna III
expand section 
expand section 
expand section 
expand section 
expand section 
expand section 
expand section 
expand section 
expand section 
expand section 
expand section 
expand section 
expand section 

expand section 

122

Page 122

The London Thornton Manuscript: A Corrected Collation
by
Ralph Hanna III

In two very important notes, Sarah M. Horrall has advanced considerably our knowledge of the construction of Robert Thornton's two miscellanies, Lincoln Cathedral MS. 91 and British Library MS. Additional 31042.[1] Horrall carefully describes all the watermarks of the two codices; by matching watermarked and unwatermarked halves of Thornton's sheets, folded in folio, she can provide a collation of the British Library manuscript. This collation will be the basis for any further codicological understanding of the manuscript, for the pages are now mounted individually.

In most respects, Horrall's collation is beyond fault and an important contribution. But I do not believe that it is entirely correct. In suggesting improvements, I hope to shed some light on two mistaken assumptions which Horrall is not alone in making about the composition of English literary manuscripts.

Horrall divides Additional 31042 into thirteen codicological units, not all of which she is sure represent quires. Her collation runs: 1? (ff. 3-8, the fragmentary end of a quire) 224 (ff. 9-32) 322 (—22; ff. 33-53) 420 (ff. 54-73) 510? (—5, —8; ff. 74-81) 618? (—17; ff. 82-97, only 16 leaves) 74 (ff. 98-101) 810 (—2; ff. 102-10) 9? (—1, —2; ff. 111-19) 10? (ff. 120-24) 1122 + 1; (—20, —21, —22, +23; ff. 125-44) 1224 (ff. 145-68) 1318 (—14, —15, —16, —17, —18; ff. 169-81).[2] No one, I should think, will quibble with Horrall's reconstruction of the greater part of the manuscript: certainly, the results appear clearcut and sensible for ff. 3-73 and 125-81. More problematic is the fifty-odd folio gap separating these portions of Additional 31042, where Horrall's collation shows a variety of small clumps of folios and numerous authorial questionmarks.

At least three considerations, apart from Horrall's own illease, should make one hesitate at this point, I think. First, the six units into which Horrall divides ff. 74-124 are largely animated by the texts these folios contain, rather than by any codicological information. Second, with the exception of a fleeting glance in a footnote, Horrall ignores the emphatic and unequivocal


123

Page 123
codicological information she has so carefully unearthed.[3] And third, the resulting disposition of parts does not savor strongly of the behavior, as we know it, of the compiler-scribe Robert Thornton.

Horrall's six textual units contain only one quire (number 6) which at all resembles a usual Thornton product. This Yorkshire gentryman[4] is renowned for the sheer size of his textual units: for example, in the Lincoln MS., quires 15 and 16 contain 30 and 36 (perhaps originally 40!) leaves respectively. In fact, taking only units which are certainly quires into account, Thornton never copied in a format smaller than 16 folia (quires 4, 6, and 10 of the Lincoln ms.). Elsewhere in Additional 31042, Horrall's quires conform to Thornton's expected norms: quire size is not consistent but approximate, usually 22 or 24 folios.[5] In this context, Horrall's questionmarks are surely merited, for, without more extensive bibliographical data than she chooses to utilize, there is only a minimal likelihood that Thornton constructed the manuscript in the way she suggests.

Moreover, Horrall's reconstruction ignores, in ff. 74-124, such bibliographical evidence as the manuscript provides. Horrall routinely pairs leaves bearing watermarks and leaves without them; since the quires of the manuscript have been constructed by folding in folio, one watermarked and one unwatermarked leaf comprise a full sheet. But Horrall fails to note a very important corollary to this procedure: since watermarks are imposed symmetrically on sheets during manufacture, when sheets are folded to form a quire, they will show symmetrical patterns of watermarked and unwatermarked leaves within the quire. Thus, even without knowing quire boundaries, symmetrical patterning of marked and unmarked leaves generally provides evidence for at least a portion of a quire.[6] Hence, such features as a series of five successive watermarked leaves (ff. 104-108) and a series of five successive unwatermarked ones (ff. 111-115) surely suggest the presence of a single quire, rather than Horrall's two quires, in this part of the manuscript.

Horrall ignores her evidence here and elsewhere because she believes she has a better way to recognize quire boundaries. She rejects information provided by watermarks in favor of what is known about the scribe's practice from the Lincoln MS.; there Thornton prepared his texts in fascicular units, small booklets usually composed of several quires.[7] Horrall assumes this


124

Page 124
discovery will also aid in determining the collation of the London MS. and attempts to pinpoint those folios which are likely to have stood at the end of fascicles (and thus of quires). Her effort is laudable, for the London ms. is very likely fascicular; but the attempt remains unconvincing because of Horrall's assumptions about how fascicle boundaries are to be identified. Inherent in Horrall's argument is the belief that all folio junctures where the verso concludes with a series of short texts and the recto begins with a new and substantial text are likely to be fascicle (and thus quire) boundaries. While such a situation is surely typical of fascicular manuscripts generally, it is not necessarily the case that the situation uniquely typifies the fascicle: some scribes simply prefer to begin new texts at the top of rectos and fill in intervening blank leaves with short materials.[8] In brief, Horrall has exchanged a powerful tool, symmetry of watermarks, for a dubious one, the disposition of texts; this error, when joined with the un Thornton-like collation she produces, renders her findings unconvincing.

One does better, then, to return to the incontrovertible codicological evidence. If one looks at a sequence of leaves and their watermarks, one can see very quickly that the marks form a symmetrical pattern. I give demonstrably lost folia in brackets[9] and place a letter, Horrall's identification of the watermark, above the number of those folia bearing watermarks:

       
[102a]  103  104  105  106  107  108  109  110 / 
117  116  115  114  113  112  111  [110b]  [110a] 
In the form I present the information, I hope to make clear that this sequence of leaves represents an 18-leaf quire, with its center at "/." If one

125

Page 125
reads the diagram vertically, each pair of folio numbers represents the two halves of an originally conjoint bifolium. In Robert Thornton's folio folding of large paper sheets, the watermark appears on only half the leaf, and thus ff. 103 + 116, 104 + 115, 105 + 114, 106 + 113, 107 + 112, and 108 + 111 form full sheets. By inference, ff. 110 + 110a were a full sheet, the lost leaf without watermark, and presumably ff. 102a + 117 and 109 + 110b were full sheets with watermarks on the lost halves of the bifolia.

The probability that ff. 103-117 represent a quire (an 18, lacking leaves 1, 10, and 11) is increased by the infrequency of this paper stock. It appears in Thornton's work only here, in fourteen or fifteen sheets of quires 7 and 15 of the Lincoln ms., and in the adjacent ff. 118 and 119 of this manuscript. One can demonstrate rather easily that these adjacent leaves cannot be part of the same quire as ff. 103-117. If one redraws the diagram above to include them, the results will not produce coherent full sheets:

       
[102a]  103  104  105  106  107  108  109  110  [110a] 
119  118  117  116  115  114  113  112  111  [110b] 
One putative sheet (ff. 105 + 116) bears two watermarks and two others (ff. 109 + 112, 110 + 111) have none. This represents a less plausible possibility and should be rejected.[10]

If one identifies ff. 103-117 as a quire, one immediately must deal with a small piece of text:

   
118  119  120  121  122  123  124 
Here there is an odd number of leaves, but the watermarks show a symmetry of sequence reminiscent of the much larger quire just surveyed:        
118  119  120 / 
124  123  122  121 
The folia appear to represent a six-leaf quire with an added singleton. This is, as I've argued before, far smaller than Thornton's normal quires and thus anomalous; it may be acceptable here as marking the end of a fascicle, an originally selfsubstantial series of quires. But more probably, the leaves represent only a piece of a quire with an indeterminate number of folia at the end cancelled. In that case, the smallest possible quire would be a 14, with

126

Page 126
f. 124 the seventh leaf. This possibility, especially given ultimate position in a fascicle, gains support from cancellation of leaves at the ends of fascicles in the Lincoln ms., quires 3 (leaves 11-18 cancelled) and 10 (leaf 16 cancelled).

If one turns to the head of this section of Additional 31042, another more or less symmetrical series of watermarks occurs (here Z is an unidentified mark unique to f. 90):

       
74  75  76  77  [77a]  78  79  [79a]  80  81  82  83  84/ 
[96a]  96  95  94  93  92  91  90  89  88  87  86  85 
Particularly important signs of symmetry are the F sheets (ff. 80-89, split by Horrall between two quires) and, surrounding them, the succession of four unwatermarked leaves (ff. 91-94) balanced by three watermarked leaves and a lost one (ff. 77-79). The diagram above identifies eleven full sheets, nine certain and two putative. Ff. 76 + 95, 77 + 94, 78 + 92, 79 + 91 represent full sheets of Horrall's stock E; ff. 80 + 89, 81 + 88, 82 + 87, 83 + 86, 84 + 85 are full sheets of Horrall's stock F. Folios 74 + 96a presumably formed a full sheet of stock E and ff. 79a + 90 of the unidentified stock. One must assume that ff. 77a + 93 were also such a full sheet, watermarked (with E?) on the lost f. 77a, and there remains the apparent anomaly of ff. 75 + 96, which bear no watermark.

This anomaly is, however, only apparent. For among the numerous paper stocks which Thornton utilized in his two manuscripts, there were at least a few sheets which lacked any watermark. Three examples, unnoted by Horrall, appear in the Lincoln ms. in clearly defined quires where they are still bifolia: ff. 51 + 52 (now singletons), 108 + 117, 296 + 299. If Thornton did have such sheets, and if he did use them occasionally, one can identify ff. 75 + 96 as the unique such example in Additional 31042. That identification allows one to group ff. 74-96 as representing a typically large Thornton quire—26 leaves, lacking 5, 8, and 26.

But the quire typifies Thornton's work in yet another way—in its mixture of papers and in the confirmation of that mixture. The center four bifolia represent a paper stock different from that of the majority of the leaves, which forms the outer sheets of the quire. This phenomenon occurs with some frequency in Thornton texts. Elsewhere in Additional 31042, quire 4 has three bifolia at its center of a different stock (Horrall's D) from the remainder of the quire (Horrall's C, with one odd sheet). In the Lincoln ms. the phenomenon is more frequent: for example, there are two odd bifolia at the center of quire 1, three at the center of quire 2, three at the center of quire 5, three (of two different stocks) at the center of quire 7.[11]

These fitful changes of paper stock suggest a mode of composing a codex


127

Page 127
which Thornton took as a given but which modern scholars generally consider an outre aberration. The example of Thornton, I think, forces one to consider whether the usual assumption of the singly-folded quire, in which a scribe selects a given format in advance and retains it, represents a universal model. The example of Thornton, which can be multiplied by several other English cases, suggests another model, in which a quire may represent something more than a single folding of sheets.[12]

That is, Thornton frequently appears to have supplemented his quires from their centers. Having begun copying with a certain stock of paper folded in folio, he at times increased his working space and his quire size by inserting separately folded bifolia into the center of his quire.[13] Such a procedure would be sensible for a compiler in Thornton's position. He obviously had, over a space of years, access on a fitful and perhaps limited basis (probably through loans) to a variety of manuscript sources. Moreover, the kinds of miscellanies Thornton could have borrowed would always have contained surprises: he could never be certain he would not, in the pursuit of text A, accidentally find a new literary text B which he would want to make his own by copying. But Thornton, if one judges his final product, also appears to have been reasonably methodical: although he had a remarkably catholic taste, the completed manuscripts fall into fairly precise arrangements by category, romances with other romances, for example. One method of facilitating this categorization would be to have several fascicular manuscripts in progress simultaneously. And, as a way of accommodating unexpected but desired texts, Thornton seems to have made room in fascicles already in progress by suppletion, by adding extra center bifolia, rather than by making new quires.[14]


128

Page 128

Recognition of this possibility also shows Thornton's quiring in a new light. Although he is generally thought of as rather sloppy, simply grabbing a (large to very large) handful of sheets, folding them, and filling them, the evidence of the paper stocks is a little different. If one looks upon many of the mixed-stock quires as in fact supplemented after an initial folding, Thornton's procedures appear much more normal. He typically began with eight or nine full sheets, a fairly consistent and normal quire size; only by chance additions and later foldings of further sheets did many quires achieve their current uneven and bulky appearance.[15]

If these arguments are convincing, there are only six further leaves with which to deal, ff. 97-102. These leaves will prove somewhat more problematic than any of the preceding:

   
97  98  99  100  101  102 
Ff. 98 + 101 and 99 + 100 can be taken as conjoint and from the same sheet. But ff. 97 and 102, each with watermark, cannot represent halves of the same sheet. They may provide an example, unique in Thornton's two manuscripts, of half-sheet imposition, perhaps the result of a cancel.[16] But there are other possibilities, at least one of which also explains the smallness of this quire.

One must face the possibility that the quire is substantially incomplete. Horrall assumes, properly as an initial effort, that only one leaf, the twenty-sixth and last of the preceding quire, has been lost after f. 96. Her argument appears to have considerable merit, for she demonstrates that the materials missing at this point could have filled a single leaf. F. 96 contains Lydgate's "Kings of England" (IMEV 3632) but breaks off in midstanza and lacks lines 61-105, while f. 97 begins at line 18 of Lydgate's "Dietary" (IMEV 824). These missing pieces of text, says Horrall, provide "exactly the right number of lines to fill one manuscript folio."


129

Page 129

But three objections can be lodged against this argument. First, in spite of Horrall's assertion, the missing sixty-two lines are not "exactly the right number" to fill a missing leaf. Thornton, who varies a good deal in the number of lines per page, is averaging in this portion of the codex about seventy-six lines to the folio, and the missing material represents substantially less than a folio. There is thus no assurance that f. 96a contained both the end of "Kings" and the opening of "Dietary"; these texts almost certainly occupied separate folios, with some intervening and unrecoverable text.

Moreover, Horrall does not take into account other possible losses within the folios under discussion. One cannot definitively state that no folio is missing after f. 97. Although f. 97v clearly is loaded with filler, given that the four lines of poetry at the page foot (IMEV 3778) are unique, one cannot assert absolutely that they form a complete poem. Further, the subsequent quire 7 opened with a leaf now lost, and the text at the end of quire 6 is incomplete—and again unique (IMEV 990). Thus one cannot be absolutely certain whether a number of folios, rather than only a single one, may be missing after f. 102.

Following this reasoning, the diagram for these folios given above needs to be redrawn. But such a redepiction has to acknowledge substantial indeterminacy:

   
(. . . [96n]?)  97  ([97a]?)  98  99  100  101  102  (. . . [102n]?) 
If Lydgate's "Kings" ended on f. 96a and that folio did not bear the opening lines of the "Dietary," one cannot be certain how many folios or how many texts have been lost in this section of the codex. If one assumes that no folio has been lost after f. 97 and only one after f. 102, the minimum quire under examination would be an 8 (lacking ff. 1 and 2) with ff. 97 + 100 and 98 + 99 surviving full bifolia. If one assumes that a single leaf is gone after both ff. 97 and 102, the minimum quire would be a 12 (lacking ff. 1-5) with no surviving full bifolia. And if more than one leaf has disappeared after f. 102, the number of possible original quire sizes increases. One can represent the minimum quires in the simplest cases, with two folios lost after f. 102 (one at the head of the next quire):                
(a)  [96b]  [96c]  [96d]  97  98 / 
[102b]  102  101  100  99 
(b)  [96b]  [96c]  [96d]  [96e]  [96f]  [96g]  97 / 
[102b]  102  101  100  99  98  [97a] 

130

Page 130
Obviously the quire may have been far larger than a 10 or a 14, but the essential point remains the same: very likely, too few folios here survive to make any definitive judgement about the original status of the quire.[17]

Having identified ff. 97-102 as a fragmentary quire, I propose a new collation for ff. 74-124: 526 (—5, —8, —26; ff. 74-96) 6? (ff. 97-102) 718 (—1, —10, —11; ff. 103-117) 86 + 1 (+7; ff. 118-124) or 814? (—8, —9, —10, —11, —12, —13, —14; ff. 118-124). And, since this interpretation reduces the number of quires from Horrall's thirteen to eleven, the conclusion of the manuscript can now be designated 922 + 1 (—20, —21, —22, +23; ff. 125-144) 1024 (ff. 145-168) 1118 (—14, —15, —16, —17, —18; ff. 169-181).

Folios 74-124 then comprise four quires. Two of these reflect Thornton's normal tendencies. Two are unusually short; one is a fragment, whereas the other shows Thornton's tendency to copy texts in ongoing fascicular units. Lincoln Cathedral 91 has a fascicular format—six parts, each with reasonably homogeneous textual contents—and the London MS. seems similarly constructed:

Part I = quires 1-2 two long excerpts from Cursor Mundi.

Part IIa = quires 3-5 romances: The Siege of Jerusalem, The Siege of Milan, Roland and Otuel, with a few briefer items, especially at the end, ff. 94-96.

Part IIb = quires 6-8 religious verse: The Quatrefoil of Love, Lydgate's "Virtues of the Mass," a versified Three Kings, with briefer items.

Part III = quires 9-10 Richard the Lionhearted (romance) and a verse narrative of the infancy of Christ (IMEV 250).

Part IV = quire 11 two dialogue poems: The Parliament of the Three Ages, Winner and Waster.[18]

I assign the designations "IIa" and "IIb" to the problematic section of the manuscript on the chance that the lost folios may have included a fascicle boundary.

I hope that this correction, in company with Stern and Horrall's pioneer work, will stimulate further the study of Thornton's miscellanies. As Horrall has already suggested, common paper stocks may imply contemporaneous copying and thus common sources, and someone may thus be able to identify some of Thornton's exemplars. This step, in turn, may make it possible to attain greater specificity about the construction of the manuscripts—which fascicles were in progress at what time (a relative chronology for the minimum of two decades over which Thornton's copying appears to have extended). And such studies may end with major advances in general literary history, for they will clarify the movement of texts in the north of England, c. 1430-1450.

Notes

 
[1]

"The London Thornton Manuscript: A New Collation," Manuscripta, 23 (1979), 99-103; "The Watermarks of the Thornton Manuscripts," N&Q, 225 (1980), 385-386. These notes grow out of a response to the important reevaluation of Additional 31042, Karen Stern, "The London 'Thornton' Miscellany," Scriptorium, 30 (1976), 26-37, 201-218.

[2]

The foliation of the manuscript includes four flies (ff. 1-2, 182-83), in fact pastedowns from an earlier binding. Except for catchwords on ff. 8v, 32v, and 73v (which occur in nonproblematic positions), the manuscript lacks catchwords and signatures.

[3]

Manuscripta, 23 (1979), 103, n.13.

[4]

For the best summary of biographical information, see George R. Keiser, "Lincoln Cathedral Library MS. 91: Life and Milieu of the Scribe," Studies in Bibliography, 32 (1979), 158-180. For collation of the Lincoln manuscript, see D. S. Brewer and A. E. B. Owen, The Thornton Manuscript (Lincoln Cathedral MS. 91) (1975), pp. xii-xvi; and Owen's correction (the lost bifolium ff. 147a+159a ignored in the Scolar description), TCBS, 6, iv (1975), 221.

[5]

The final quire, an 18, may have been deliberately small.

[6]

The classic demonstration, although dealing with quarto folding and complex in its attention to the direction of the folds, is Stephen Spector's "Symmetry in Watermark Sequences," SB, 31 (1978), 162-177.

[7]

See Brewer and Owen, Thornton Manuscript, pp. viii-ix, for a description of the separate fascicles of Lincoln 91.

[8]

This failure to distinguish necessary and sufficient conditions for the identification of a fascicle or booklet also bedevils the most significant recent work on the subject, Pamela Robinson's "The 'Booklet,' A Self-Contained Unit in Composite Manuscripts," Codicologica, 3 (1980), 46-69; Brewer makes similar statements, Thornton Manuscript, p. viii. I examine the ramifications of the problem more fully in a forthcoming article on Westminster School MS. 3. For an example analogous to Thornton's technique here one might compare British Library MS. Harley 2250. This codex, like Additional 31042, now exists as mounted single sheets and has not been collated, but one should note the tendency to begin new texts unconnected with preceding materials at the head of the next available page (not necessarily a recto, and thus surely not a quire boundary) and the concomitant presence of filler items; see ff. 49v, 72 (column 2), 83v, 84v (column 2), and 87rv (the end of the ms.). The last three instances likely occur within a single quire. See the description provided by Clifford Peterson, Saint Erkenwald (1977), pp. 3-6.

[9]

These losses have been recognized on three grounds—survival of stubs in the manuscript (as at ff. 110a and 110b), incompleteness of texts when compared with other copies (as at ff. 143a, 143b, and 143c, although there could be only two missing leaves and the quire a 22) and by inference in the case of unique texts (f. 77a in The Siege of Milan is posited on the basis of a break in the stanza form and some narrative discontinuity). For a definitive description of contents, see Stern, Scriptorium, 30 (1976), 214-18; for identification of lost leaves, see Stern, Scriptorium, 30 (1976), 28-30, with corrections by Horrall, Manuscripta, 23 (1979), 101-102.

[10]

Other efforts to place all these sheets into the same quire, or into two adjacent, but separate, quires are possible, but all will show similar inconsistencies. For example, splitting into two ten-leaf quires will give three full sheets with no watermark (ff. 103 + 110, 112 + 117, 114 + 115) and two sheets with two watermarks (ff. 105 + 108, 106 + 107), as opposed to only five normal full sheets.

[11]

And more complicated examples in quires 10 and 15, where three stocks of paper are involved.

[12]

Other examples include National Library of Wales, Peniarth 392 (the Hengwrt Canterbury Tales), quire 22, where the supplemented quire reflects difficulties in the scribe's supply of copy (see A. I. Doyle and M. B. Parkes, The Canterbury Tales [1979], pp. xxxi-xxxii); two examples where scribes increase their writing space from without, by quiring an existing gathering within other bifolia, Huntington HM 144, quire 9 (see John M. Manly and Edith Rickert, The Text of the Canterbury Tales, 8 vols. [1940], I, 289-292) and Trinity College Cambridge R.14.45, part 6, quire 1 (see G. S. Ivy, in Francis Wormald and C. E. Wright eds., The English Library Before 1700 [1958], p. 42). Similar, but after the fact, are the several insertions of originally independent quires into existing ones in the Findern Miscellany, a text prepared by about thirty scribes; see Richard Beadle and A. E. B. Owen, The Findern Manuscript, Cambridge University Library MS. Ff.i.6 (1977), pp. viii-xi.

[13]

This view, of course, assumes that Thornton had a rational paper supply in which he obtained sheets of the same manufacture in largish lots, rather than constantly working with odds and ends. But certainly not all examples reflect later suppletion: quire 5 of Lincoln Cathedral 91 looks as if Thornton was at the end of his stock B and simply filled out the (eighteen-leaf) quire with stock L, which he had in profusion. (A total of 44 sheets in Lincoln 91 represent this stock.) If that was the case, quire 9, later in the same fascicle, and quire 3 of Additional 31042 were already partially copied before quire 5 was constructed. On Thornton's exemplars, no work has succeeded Angus McIntosh's ground-breaking "The Textual Transmission of the Alliterative Morte Arthure," in Norman Davis and C. L. Wrenn eds., English and Medieval Studies Presented to J. R. R. Tolkien (1962), pp. 231-240.

[14]

A fair amount of evidence supports such conjectures. The leaves of the manuscripts show very uneven wear, but particularly heavy wear on folios one might hypothesize to have been the outsides of individual fascicles. (In fact, text-bearing first folios of quires have fared particularly badly in Thornton manuscripts: at least six, two with their conjugates, are missing in the sixteen quires of Lincoln 91.) The fascicles may have had several years of use as loose booklets before binding; they required some tailoring to be fitted together into codices. Thus the verso of Lincoln 91, f. 52, originally leaf 10 of quire 3, was given a catchword and the eight subsequent folios cancelled, probably at the time of binding. Partial correspondence of paper stocks (the two manuscripts share five stocks) seems to indicate that the texts were in progress simultaneously and their contents only divided later.

[15]

Although the enormous quire 16 of Lincoln 91 was created as a self-contained book of medical recipes, the thirty-leaf quire 15 could have originally been an 18, with extra leaves inserted in its center.

[16]

Halfsheets appear fairly regularly in paper manuscripts, and it is surprising that there are no clear examples in Thornton's work. As an example, Huntington HM 114, with paper sheets folded in quarto within a vellum sheet (which becomes the center and outer bifolia of each quire) has about 10% halfsheets. Given the unbound state of his fascicles, Thornton may have preferred replacing a full bifolium to using halfsheets.

[17]

My colleague, William O. Harris, has offered me helpful advice on the whole paper, but especially on this section: his arguments convince me that trying to view ff. 97-102 as substantially complete only creates difficulties, the postulation of such features as cancels or inserted singletons.

[18]

Most of the lost sheets at the end are probably cancels, since Winner appears nearly complete, with only a few lines lost at the end.