University of Virginia Library

II

In considering ways in which information about paper can be used, it is possible to distinguish at least three levels of application. Initially one can merely describe and identify the material at hand. From that, one can make inferences about the process of manufacturing the book, and beyond that consider the textual implications of those physical qualities. Each step considerably widens the range of possible implications and applications of the preceding level. In what follows I shall deal with the first two stages, beginning by showing how the procedures for identifying paper actually operate in several editions of The Dunciad.

The first 1728 edition was issued as both a duodecimo and an octavo (Foxon P764-765). The same general varieties of paper were used for both impressions: the chain lines are double, and the watermark, when present, is 'BF' and lies on the right-hand half-sheet. As a reflection of the fact that both impressions were imposed in half-sheets, about half of the gatherings are unmarked. The presence or absence of watermarks provides the first means of distinguishing the papers. The watermarked ones can be divided into two groups: some of the paper has chain shadows to the right and has tranchefiles, while other paper has the shadows to the left and no tranchefiles. The marked paper with tranchefiles can be further divided according to the position of the watermark: sometimes it lies in the chain space adjoining the tranchefile space, while at other times a complete chain space lies between the one occupied by the edge of the watermark and the nearest tranchefile space. The chain space measurements of those two varieties reveal that no further subdivision is possible. Other chain space measurements show that the marked paper without tranchefiles also divides into two varieties; overall, then, there are four varieties of paper with the 'BF' watermark.

Similar methods help to sort the unmarked half-sheets. Some of the paper lacks tranchefiles and has chain-line shadows away from the deckle edge. The paper that does have tranchefiles has the shadows toward that edge. Because there are no watermark positions to provide points of identification here, and because the variation in the wire line density is not great enough to be relied on, the chain spaces provide the next step of discrimination. The measurements of each class of paper fall into two groups, yielding the same number of varieties as the watermarked sections.

It is extremely likely that the eight varieties of half-sheets were originally mates of each other in some combination. But at this point it is not clear whether each whole sheet had one marked and one unmarked half or whether some sheets had two marks and some none. Although some combinations can be ruled out on the basis of the shadow direction (complementary half-sheets can not both have shadows going toward the deckle edge), not all possibilities are thereby tested. One way of positing matching pairs is to examine the distribution of each variety throughout numerous copies and make tentative links between half-sheets that appear in the same sections of the book. That process yields clear patterns here, but it can not be considered conclusive. What does yield decisive evidence is a feature of the imposition patterns of the two impressions of this Dunciad. A half-sheet in octavo is either the left or right half of the original sheet and ordinarily contains half of the chain spaces (which run vertically) on that sheet. A half-sheet in duodecimo,

on the other hand, is the top or bottom half of the original sheet and therefore contains part of every chain space on that sheet. One-third of a duodecimo full- or half-sheet is of course cut off, and there is little assurance that a particular cutoff will be reunited with its former companions. But that still leaves two-thirds of the original chain lines and spaces present. Because the paper varieties in both formats of The Dunciad are demonstrably the same, the chain space models for half-sheets of the octavo are effectively extended to two-thirds of a sheet when combined with the duodecimo readings. That means that chain space models from two ends of the same sheet will overlap for the center third and that the pairs of extended half-sheets can be reliably matched. The chain space models that result each have one end with tranchefiles and one without. Interestingly, those paper models match the characteristics of the paper Gaskell identifies as produced by end-to-end two-sheet moulds: "paper from pairs of such moulds will show combinations of four different examples of the watermark, not two, and it is also likely that only one of the shorter edges of each sheet will have the tranchefile and drip marks" (New Introduction, p. 64). If this 1728 paper, probably from Italy, is from double moulds, it is a fairly early example. Gaskell cites such paper from the 1690s, probably from Holland; the earliest documentation he finds for its manufacture in England and France, however, is from 1768 and 1788 respectively.

This first 1728 edition of The Dunciad proves to be a useful one, for it shows the extent to which various features of paper which are usually overlooked can help to differentiate varieties. While easily detectable characteristics define broad classes, I was able to note those differences most readily because I had before me the array of data collected from numerous copies of the works, usually in conjunction with chain space measurements. The final distinctions, however, had to rely on the time-consuming measurement of spaces. Only after the features of the varieties were ascertained could hypotheses be safely entertained about the nature of the moulds. As so often, watermarks played an equivocal role in the identification. Their position helped to distinguish the half-sheets in which they were present, but they were absent from about half of the gatherings. In addition, the portions of the initials that remained after trimming usually were too small to provide points of comparison on the marks themselves.

The temptation to incorporate anomalous chain space patterns also arises here. In at least two copies of the octavo, gathering B has a countermark 'G' in the lower corner. Both of these copies are at libraries which might be outside the group that a bibliographer would be most likely to check—the National Library of Wales, and Edinburgh University.

A difference in shadow direction indicates that the two examples are different. What is implicit, therefore, is that these samples have corresponding half-sheets, perhaps unmarked ones that are in the 10% or so of gatherings that I have been unable to assign with certainty. A bibliographer who would force those remaining examples into the models to which they were only approximately similar would therefore risk obscuring evidence of variant papers.

Another interesting array of paper varieties occurs in the first octavo edition of The Dunciad Variorum in 1729 (Foxon P776-780). The collation for the most common forms of the book is: 8°: A4 (—A2, A1 + χ2) B-Aa4 Bb4 (± Bb3) Cc-Gg4 Hh1. A1 is a blank and Hh1 an errata leaf; an engraved plate of either an owl or an ass may appear as the frontispiece or may be inserted at the beginning of Book I or II in the poem. The greater part of the work, gatherings B-Bb, is printed on the same two varieties of 'S'-marked paper with double chain lines that comprise the quarto edition of the poem, which was circulating privately about a month before the octavo was first advertised. The problem of matching the correct marked half-sheets with unmarked ones again arises but here is solved by recourse to the quarto, where full sheets reveal the entire chain space sequence.

The watermark 'TM' with the corner countermark 'CM' over 'T'[16] occurs in the half-sheet gatherings A and Cc-Gg. These can be subdivided into two groups according to the frequency of wire lines and the position of the watermarks. Four varieties have fine wires (33/3), and they never have the countermark and main mark in the same half-sheet. The directions of bar shadows indicate that these four varieties constitute two pairs of half-sheets, or two sheets. With only one exception in the seventeen copies checked, only this fine-wired paper is used in gathering Cc. A few scattered examples also occur in three of the other gatherings in this non-'S' section.

The other sub-group, with noticeably coarser wire lines and with the watermark and countermark on the same half-sheet, occurs in two varieties with wire lines of 24/3 and two with wires of 26/3. Both of the 26/3 papers have the mark 'TM' upside down with respect to the countermark. Again the shadow directions clearly indicate which of the unmarked gatherings are the likely complements of the marked ones. Only papers with wires of 24/3 occur in Dd, and they are by far the predominant varieties in Ff and A. Ee and Gg, on the other hand, consist mainly of varieties with wires of 26/3.

The cancel Bb3 and the errata leaf, which were printed together and are occasionally found conjugate, sometimes have a 'TM' watermark. On the basis of the chain-space models established for varieties of paper with that watermark, unmarked examples of these leaves can also be identified as coming from those sheets—almost invariably ones with the coarse chain lines. The same is true for the cancel leaves χ2, the new title page and the owl frontispiece which replaced the original title page. Sometimes that owl plate appears separately later in the volume; then too it is on 'TM' paper. The ass plate, on the other hand, is never on that paper. Sometimes portions of the watermark 'BF' appear on that plate, and sometimes a fleur-de-lis, suggesting that the paper may be the kind represented by Heawood's design 1715 (dated 1729), with one of the marks in each original half-sheet. The initials suggest a possible connection with the 'BF' paper of the 1728 editions, but a comparison of the chain spaces with the 1728 models reveals that the paper is not the same. Again, then, the features of chain and wire lines provide a means of identifying paper varieties. In a moment I shall consider some possible implications of that information.

Other special leaves in 1729 editions can also be identified by recourse to chain space models. Most copies of the second octavo edition of The Dunciad (Foxon P781) have a half-sheet of errata. It is on paper different from the rest of the book but, with the hint provided by an occasional watermark and the confirmation from other paper characteristics, it can be shown to be on the two 'TM'-marked sheets with wires of 24/3 that appear in the first octavo. 'TM' paper shows up in another place as well: the addenda leaf at the end of the 1729 quarto is printed on marked and unmarked sections of these sheets. Meanwhile, back in the second octavo edition the three cancel leaves and single-leaf errata list can be proven to be from the same paper as the rest of the book.

The next example, the octavo edition of The Dunciad which appeared as volume V of Pope's small-paper Works in 1751, differs from the previous ones by having turned chain lines, single chains, and many sheets with no watermark at all. Some of the features useful in identifying paper thus are not present (though their absence is itself a distinguishing factor), and the effectiveness of the chain space measurements has been reduced because there are fewer spaces—and hence, fewer potential variations—per sheet.

The presence of watermarks provides the first distinguishing feature of the paper. Because most of the gatherings in the volume are full sheets, it can be determined with reasonable certainty whether they are marked. Gatherings a-b and A-G never have a mark, while the remaining ones—πA, c-d, and H-R—usually have the initials 'RW'. One way to begin

sorting the marked sheets is to classify them according to the portion of the mark that shows. The method is by no means foolproof, for the sheets might have been placed in slightly different positions on the tympan, thereby causing the printed impression—whose outline, and not the edges of the sheet itself, were used by the binder as folding guides—to vary in respect to the position of the watermark. But in the present case the watermark sections do help. Some sheets have the top of the mark showing, and others the bottom. A comparison of the chain line spacings reveals three subgroups of each kind. Double-checking the space patterns of these six groups against each other confirms that they are separate varieties.

Six varieties of paper is an unusual number for sheets which, as the turned chain lines indicate, were almost certainly produced on double moulds. These sheets would be expected to occur in multiples of four: two sheets per mould, with a pair of moulds. The occasional unmarked sheets which appear in the sections of the book which are usually watermarked provide an explanation. When the chain spaces of these sheets are analyzed, they yield two patterns, bringing the total number of varieties to eight. The patterns are different from those in the marked sheets, meaning that they are not merely states of the marked ones in which the designs had come loose from the moulds. Nonetheless, these two varieties may have had marks of some sort at one time. It might also be that one pair of moulds had marks on the two sheets on one mould but no marks on the other mould, or, more likely, one marked and one unmarked sheet on each mould.

The section of the volume where only unmarked sheets appear poses greater difficulties, for there are no clear distinguishing features to provide initial subgroupings. But most of them have tranchefiles at only one end of the sheet; these provide a way of orienting them for easier comparison. When the chain spaces are checked against each other, four distinct patterns emerge. It is possible that some of these varieties are the ones occurring throughout the watermarked sections, but another check of the chain spaces reveals that all are distinct. Finally, it is also necessary to consider the volume's four cancel leaves and the errata leaf (which appears in two settings), none of which is marked. Their chain spaces match none of the twelve identified varieties in the book. That lack of correspondence fits the information in the paper ledger of the book's printer William Bowyer; the paper he ordered was sufficient only for the 20½ original sheets of volume V. Cancels and errata lists are not planned before the printing of a work begins, and they were apparently machined on other paper.

Thus far the second part of this essay has been concerned with ways

in which the techniques discussed in the first part can be used to discriminate varieties and determine characteristics of sheets of paper. But already we have moved beyond that initial stage by considering how that information can be used in identifying special single sheets such as errata and cancel leaves and in helping to detect double moulds. In the remaining part we shall look at additional applications of the descriptions. Those uses are not exhaustive, but they do suggest the kinds of insights to which the material can lead.

Paper without watermarks can be used for many of the same bibliographical purposes as that with marks, but because more of a sheet is now relevant, the potential frequency of use increases. The newly noted characteristics can provide new ways of performing old tests—such as that for conjugacy, for instance. When one leaf of a suspected pair has a type impression on only one side and the test for first forme impressions does not work or when the leaves have been pressed so smooth that neither that test nor the one for the mould side of the paper can apply, a check of the chain-space pattern, shadow direction, and wire line frequency may solve the problem. In the first 1729 octavo Dunciad, for instance, one can determine on those grounds that the owl frontispiece and cancel title page are conjugate but that the ass plate and title never are.

Paper identifications can also suggest relationships between different works, even if the precise meaning of those relationships is not immediately clear. As noted earlier, two gatherings of preliminaries in The Dunciad, in Four Books in 1743 are of the same paper as several in the Essay on Man published a few months later. One possible inference clearly is that those sections may have been machined at the same time. A more complex relationship occurs in the 1729 editions, all printed by John Wright and published by Lawton Gilliver.[17] The first Dunciad of that year was a quarto edition, issued on both ordinary and fine paper. The ordinary paper has varieties of an 'S' watermark; as we have seen, this paper is also used for most of the first octavo edition of that year. The quarto's addenda leaf, when present, is on 'TM'-marked paper, the same as in the rest of the first octavo. That paper appears still again in the half-sheet errata section in the second octavo edition.

Meanwhile, the varieties of eagle-marked paper of the fine-paper quarto recur throughout the second octavo. A curious problem arises here: why would this special—and presumably more expensive—paper be used for a reprint edition in a small format? The answer may lie partly in the significance Pope attached to the edition; he wrote Swift

that he wanted it to be used as the copy for future printings.[18] But the answer may also lie in other facts revealed by paper. An octavo piracy ('Printed for A. Dob'; Foxon P774-775) published by a group of London booksellers within days of the authorized first octavo was on the same varieties of good paper. The publishers had also taken special effort to reproduce the engravings of the title page and the headpiece at the beginning of Book I. In Gilliver's lawsuit which followed, the defendants noted that, among other problems, the copies of the quarto edition tardily deposited with the Stationers' Company were not on the better paper but "upon much worse."[19] The pirates' main strategy in court was to show that copyright for the book had not been clearly established and that therefore their edition was in fact not a piracy. Pointing out the inferior quality of the quarto deposit copies served the logical point of challenging the copyright claims for that book but also served the ethical point of showing these pirates' good intent, for their edition appeared not as a cheaply constructed attempt to capitalize on a popular work but rather as an expensively produced edition of a work not formerly available in a small format. It may be, then, that to produce an octavo edition on paper of comparable quality Pope specified that the next edition under his control, the second octavo, be printed on the same eagle-marked sheets. Whatever the ultimate explanation, it is again paper evidence that reveals connections among these four books.

Precise identification of paper varieties can also on occasion help to establish the dates that the paper was used. Bowyer's paper ledger reveals that the paper for volumes II and V of Pope's small-paper Works of 1751 was ordered together.[20] That spring and summer 92 reams were received on March 21, 52 reams on April 9, 40 reams on June 27, 24 reams on July 2, and 12 reams on July 12. All of the 14½ original sheets of volume II and 9 of the 20½ sheets of volume V are on unwatermarked paper, though the varieties in the two books are not the same. As indicated earlier, the remaining 11½ sheets of volume V (the last gatherings and some of the preliminaries) have the mark 'RW' or one of two special varieties of unmarked paper. The ratios of the number of similar sheets in the volumes to the total number of sheets are similar to the ratios of reams in individual shipments to the total number of reams ordered for these volumes:

         

14.5 sheets	(unmarked, vol. II)	[1:2.4]	92 reams	[1:2.4]
9	(unmarked, vol. V)	[1:3.9]	52	[1:4.2]
11.5	(marked, and unique	[1:3.0]	40}
	unmarked, vol. V)		24} =76	[1:2.9]
			12}

If those correspondences are in fact significant, and if, as is likely, the paper was used in approximately the order purchased, then within this group volume II was printed first, followed by the unmarked sections of volume V and then by the marked ones—i.e. the last gatherings of the poetic text, the appendix and indexes, and most of the preliminaries. From the dates the paper was purchased, the actual printing dates of the sections could also be estimated. A final determination of the correspondence would require a test of the typographical evidence as well as possible adjustments for the number of sheets in these particular reams, the paper which the ledger indicates was returned to the wholesaler, the wasted sheets, leftover paper from other projects which found its way into this one, and the number of copies actually printed (which, as Bowyer's ledgers reveal, often differed from the number officially ordered.) The estimation of the printing dates based on paper evidence is by no means conclusive, then, but the investigation itself is possible only because the similarities and differences among the paper varieties have been precisely determined, primarily by the measurement of their chain spaces.

Still another use of such detailed paper evidence is to determine the original relationships or imposition patterns of specially printed leaves that are not currently conjugate. This procedure depends on an earlier establishment of chain space models and of the other relevant features of the paper. The 1729 quarto Dunciad provides the first example. K1.2 is a pair of cancel leaves, and gathering P is a half-sheet. In the ordinary-paper issue, both of the predominant varieties of the 'S' watermark occur in each section. In the seven copies for which I have measurements, one of the half-sheets is always marked and one is not, and the chain spaces reveal that they are always from the same variety of paper as each other. The tantalizing inference is that these half-sheets were originally part of the same sheet. That conclusion is confirmed by the Edinburgh University copy, in which the half-sheets are bound together and are conjugate. With that copy available, the evidence provided by the paper is largely superfluous, although besides corroborating the other finding it presents evidence of the binder's practice in apparently inserting both halves of a particular sheet into the same book. The example

is useful in showing the potential of such paper evidence. The facts here lead to the correct conclusion, although they are not themselves conclusive. But by suggesting a relationship, they encourage the researcher to seek further evidence. Here, for instance, the bibliographer who would go on to examine running titles would find that those on K1.2 and P2 are the same ones as elsewhere in the book and that when those from the two half-sheets are combined they form the same set that appears in other gatherings.

A second case in which an imposition pattern may be inferred comes from the second octavo Dunciad of 1729. With only one exception, which will be considered in a moment, the cancel leaves D3, E2, and P3 and the single-leaf errata 2 B1 are all from the same two varieties of eagle paper as the rest of the book. Tranchefiles occur only in examples of D3 and 2 B1, indicating that these leaves were printed at the end of the sheet of paper. On the other hand, the most easily identifiable portion of watermark, the portion of an eagle design which occurs toward the center of the sheet, occurs only in E2 and P3. These leaves, then, were printed as the inside ones of the octavo imposition. The absence of tranchefiles in any examples of these leaves in the three dozen copies checked support that positioning.

That the leaves come from adjacent parts of the sheet suggests that some of them may have been pairs. This is certainly true for 2 B1 and the new P3, for they are bound together and conjugate in a copy at Oxford. The paper evidence also supports Foxon's surmise that "the cancels for D3 and E2 were probably printed together" (P781). But Foxon does not think that these two quarter-sheets in turn were conjugate: "Deposit copies apparently have one pair of leaves or the other, which suggests they were not part of the same half sheet and were therefore printed at different times, though before publication." The word apparently is very likely inserted because of the difficulty in ascertaining copyright deposit copies. But in at least one of those apparent deposit copies, that at Aberdeen University, both pairs of leaves are present. Among the other deposit copies various combinations of old and new leaves appear, a pattern reflected in the non-deposit copies. It is by no means clear, then, that the two sets of leaves were printed separately. The details of the leaves indicate that they were all printed on the same two varieties of paper, and those details are at least commensurate with those leaves being part of the same half-sheet. 2 B1 and D3 would be the outer leaves on the sheet, and P3 and E2 the corresponding inner ones. That hypothesis gains further strength from paper evidence in the only identified exceptions to the pattern of double-chain, eagle-watermark paper for the cancels. In a personal copy, formerly that of James M.

Osborn, 2 B1 and the new D3 and E2 are from the same setting of type as the other examples but are all on single-chain paper. (P3 was can-celled but then the original leaf reinserted.) That all of these anomalous leaves are on the same paper suggests that at least in this case—and very likely in the other instances—the cancels were printed together.[21]

In the pursuit of paper evidence, just as in other bibliographical study, it is important that the bibliographer search as widely as possible. As shown by the single occurrence of single-chain paper in the cancels of the second octavo Dunciad Variorum, that search first of all needs to encompass as many copies as possible. But ultimately that research should also extend to other works produced by the same publisher or printer, and even to works by other publishers and printers produced about the same time. We have seen hints of the benefits of such study in the identification of the anomalous horse watermark in one Dunciad and in the explanation of a couple of unusual sheets of a 1743 Dunciad by reference to a slightly later edition of An Essay on Man. In practice, such widespread investigations remain merely ideal goals. But precisely because such goals are difficult for a single worker to achieve, they serve to remind researchers of the importance of recording the results of smaller projects so that those findings may cumulatively form a body of reference material.

Throughout this essay we have considered reasons for recording particular characteristics of paper that are usually overlooked. They can often extend the range of insights available from more familiar features such as watermarks, but even more importantly they can provide evidence which those other qualities cannot furnish. These methods are valuable even if they merely duplicate other findings, for they give additional credibility to those interpretations. Likewise, they are important even if they are not conclusive, for they often stimulate new approaches which do yield such results. Finally, if they do nothing else, these ways of analyzing paper serve as another reminder for bibliographers to be cautious when claiming that two items are printed on "the same paper."[22]