University of Virginia Library

Common to all paper of the period is the presence of chain lines, with measurable spaces between them. Bibliographers have long noted

variation in the width of those spaces, but no one has pursued the implications of those differences. In his Introduction to Bibliography for Literary Students (1927), Ronald McKerrow merely observes that chains occur "from three-quarters of an inch to an inch apart" (p. 100); he seems to have in mind a variation between different kinds of paper. In the same vein, John Carter notes in ABC for Book-Collectors (1952) that chain spaces "vary in width between different makes" (p. 50); he too seems to suggest that those differences occur between paper varieties rather than within a single sheet. Fredson Bowers' Principles of Bibliographical Description (1949) advocates recording the measurement between chain lines (p. 446), but it does so specifically for nineteenth-and twentieth-century books, for which laid paper would usually be produced by machine and in which the features of the paper, including those distances, would tend to be more uniform.

In this as in so many other areas of paper study, Allan Stevenson was the first to suggest more specific applications of the evidence. One of the tests he recommends for distinguishing twin watermarks is a comparison of the chain spaces touched by each mark.[1] In the examples which follow his article he cites numerous cases where the spacing at the mark in one mould differs from that in another or where the chain spaces near the watermark differ significantly from the average openings in the rest of the sheet. In the text of the article, however, he is equivocal about the usefulness of recording chain line spaces. He confidently states, "The student soon learns the advantage of measuring the chain-spaces as part of the routine of examination," but he then notes, "In measuring large watermarks, internal chains may be ignored, for chain-spaces (after 1500) tend to average 20 mm." (pp. 68-69). Stevenson's attention, moreover, is directed to the chain lines around watermarks, not to those appearing elsewhere in the sheet. In another article, in which he shows how chain line indentations can help to determine conjugacy and cancellation, he again says of the lines, "Their regularity can be observed, their spaces measured."[2] But in the subsequent discussion he refers only in passing to "chain extensions," implying that agreement in the pattern of chain lines on two leaves can indicate whether the leaves are conjugate. Again he does not pursue his important insights by contemplating the usefulness of actual measurements.

Those who have heeded Stevenson's call to scrutinize chain spaces have followed him in devoting their attention to the spaces around watermarks and in considering the other spaces of a sheet to be the same as each other. Stevenson himself continues those emphases in later writings,

particularly in his examinations of the Missale speciale and of early English paper made by John Tate.[3] The latter study is particularly interesting from that point of view, for although he again speaks in terms of averages and focuses on chain space differences adjacent to the watermarks, he also records additional spaces—but without calling attention to the variations they reveal (p. 27). Subsequent summaries of paper evidence based on Stevenson's work reflect his emphases. G. Thomas Tanselle's authoritative survey of paper study urges a recording of chain spaces, though his sample descriptions all present evenly spaced chain lines and concentrate on those around watermarks.[4] Likewise, Philip Gaskell suggestively refers to "chain-lines differently spaced" but later specifies "the spacing . . . of the chain-lines . . . in the vicinity of the watermark."[5]

It is a measure of the significance of those earlier studies that they are able to provoke insights beyond their authors' original intents. Although the ultimate goal of determining paper varieties and moulds has sometimes been overshadowed by the intermediate goal of identifying watermarks, the methods developed for analyzing watermarks can be extended to serve the greater end by being applied to the vast quantity of paper which contains no such marks.

The measurement of the spaces between chain lines for an entire sheet of paper is probably the most fruitful extension of the earlier methods. Unlike watermarks, these lines are present in every leaf of paper manufactured before the introduction of wove paper in the later eighteenth century, and they can almost invariably be found in page margins, unobscured by type. Most important for identification, the spaces between the various chains in the paper I have examined are seldom the same. Equal chain spaces across a sheet are in fact rare enough to be a distinguishing quality of a variety of paper. In the paper of this period the openings usually vary by as much as 10% and sometimes even 20% within a given sheet, or by as much as 3 to 6 mm. Patterns become evident when the measurements are arranged in a series; these patterns serve to identify different paper moulds.

To measure the chain spaces I use a thin metric ruler. While theoretical considerations support the use of the metric system,[6] practical ones also do: this system provides units which are small enough to discriminate properly but remain easy to manipulate; and because chain lines are typically one millimeter thick, measurements to half a millimeter

can easily be made according to a binary system with chain lines sorted according to whether they fall more on a millimeter marking on the ruler or between such marks. Large formats serve the best, for they are the least interrupted by gutters, which obscure chains, and outer margins, from which some of the chain lines may have been trimmed. But sometimes when constructing a chain space model it is useful to take readings from several formats. When the same paper is used for an octavo and duodecimo, as, for instance, in the first 1728 Dunciads, the relative positions of the pages on the sheet will be different. As a result, one format reveals chains and spacings that are hidden or cut off in the other.

There are some circumstances under which the measurements of a particular chain space may vary for the same variety of paper: the sheets may have shrunk at different rates; a chain line may wander from adjacent ones as it runs from one end of the sheet to the other, and the different examples may have been measured at different parts of the sheet; the lines may have moved gradually during the life of the mould; or the measurements themselves may have been slightly inaccurate. The measurements ordinarily do not deviate by more than a millimeter per opening, however, and even when they do vary this much the pattern of wide and narrow chain spaces generally remains the same.

To help insure that a series is representative for a sheet, it is best to measure the leaves at the point that is most likely to provide an average reading, i.e. at the center of the sheet. That is also the usual location of watermarks, and measurements taken there when watermarks are present both will fit with previously published reports of chain spaces around watermarks and will yield measurements for future researchers who wish to cite only that information. For a folio, that means measuring across the center of the page; for a quarto, down the gutter; for an octavo, across the top of the leaf; and so on. The leaves one selects to measure will be based on standard imposition diagrams for the period; Gaskell's New Introduction to Bibliography provides handy illustrations of the usual formats. The measurements of leaves containing watermarks, however, should always be recorded. A researcher may also wish to measure other leaves or different places on the leaves to test the amount of variation between a pair of chain lines from the top to bottom of a sheet. Because the chains do not always lie square with the printing on the page, it is also useful to scan the unmeasured leaves of a gathering to see whether any reveal a chain—and thus the datum of an additional chain space measurement—that might have been trimmed off in other leaves. A record of the chain spaces of every leaf of a sheet would provide a means of detecting cancels, whose chain patterns would

almost certainly differ from the dominant one, but because of the time required a person may wish to test only previously suspected cancels this way.

Because the array of numbers produced by the measurements is potentially confusing, especially when consulted after the actual book is many days and many miles away, it is important to have a systematic way of recording the initial data. For formats with vertical chains, I place a ruler whose scale begins at the tip into the gutter of the book and note every point at which a chain crosses. Because of the tightness of most bindings the distance from the inside chain to the fold at the spine cannot be measured precisely anyway, so a slight bit of wear at the end of the ruler does not matter. On the other hand, it is useful to get as accurate a measure as possible, and hence the advantage of a thin ruler, which will slip farther into the gutter than a thick one will. The actual width of the leaf, which usually will have been recorded for the general description of paper but otherwise should be noted specially, and the amount of the leaf visible along the ruler together enable the bibliographer to infer the number of chain lines concealed between the innermost chains of conjugate leaves. That number eventually is important for constructing the chain space model of the entire sheet. Pushing the ruler snugly—but without undue force—into the gutter also helps to steady it during the measuring. Recording the positions of the chain lines seriatim across the rule seems preferable to measuring individually each chain space, for it is less confusing and therefore less prone to error. Such a series also provides a context for the individual measurements, one which can later help to explain errors which have crept into the record.

The procedure for formats with horizontal chains is essentially the same, though the measurement is then from the head to the foot of the leaf, or, usually, from the center toward the outside edge of the sheet. In this case the tip of the ruler lies flush with the edge of the page and the ruler itself usually can lie lengthwise in the gutter. Consistently measuring from the same side of the leaf reduces the confusion that can easily result from attempting too many processes simultaneously; I use the recto. That policy holds for watermarks too, even if that means reading and sketching them backwards. I make a record of the mould side of each leaf examined and, if necessary, make adjustments to the placement of the watermark when putting the information into final form. Meanwhile I have the information in a consistent format and can spot differences more readily.

Once I have measured each sheet, I convert the series of readings into actual chain space distances and arrange them on a page according

to the way they would have lain in the original sheet. If the two leaves of a sheet in folio have been measured, for instance, the measurements from the gutter of the first leaf might be: 18 28 27 29 27.5 28 20 19 12, and in the second: 5 26.5 28 28 28 27 29 27 10. When the readings are combined to form the complete pattern, those of one leaf (here, the first) would be reversed so that the measurements at the gutter would be contiguous:

To align properly the corresponding chain spaces from different sheets, it is best to list the openings from the back fold outward, orienting the measurements around that point. Although leaves will not be trimmed at the inner margin, they may well be at the outer ones, with a potential effect on the number of chain spaces that are recorded. Once the measurements from different sheets have been arrayed, they can be grouped. For each group, a composite reading can be established and the model of an "ideal" sheet—and in many cases a nearly exact determination of the sheet length—can be formed for each paper variety which appears. Because of the potential variations mentioned earlier, slight differences may appear among the individual sheets which seem to be the same variety; the basic pattern of wider and narrower spaces usually does not vary. In establishing the model, however, I do not include examples which vary by 1.5 mm. or more for a particular chain space.

One of the challenges in grouping like sheets is to get all of the sheets positioned in the same direction. The mould side does not matter for paper without watermarks; merely turning the sheet top-over-bottom would change the sheet from the mould or felt side to the other one. But turning the sheet end-over-end reverses the pattern of spaces. Watermarks, tranchefiles, or any of the other distinguishing characteristics to be discussed later can serve as useful points of reference. Watermarks are a special boon at this stage. By determining how the watermark reads from the mould side and by positioning the mark so that the top and bottom of the mark itself are properly oriented, one can situate the watermark on the sheet and often at once discriminate paper varieties.

Although any procedure which works well is acceptable for initially recording data, it is advantageous that when formally presenting the findings a bibliographer avoid the idiosyncratic and work within a system that is already understood. For that reason, and because it works well, I have adapted Allan Stevenson's method of recording watermark positions to the presentation of chain space models for entire sheets.[7]

The essence of Stevenson's method is to separate chain spaces by

vertical lines and to enclose watermark widths within brackets. Thus, the notation 28 | 3 [24 | 22] 5.5 | 27 describes a watermark 46 mm. wide (millimeters being the standard unit of measure), 3 mm. from the chain on the left and 5.5 mm. from the one on the right; the adjacent chain spaces are 28 and 27 mm. To those elements I have added parentheses, ellipsis points, and a mark to indicate deckle edges. When a number in parentheses occurs amid a series of chain space measurements, it is an estimate of the width of an opening; such numbers are used when the opening occurs in the gutter or at the trimmed edge of a sheet and can not be measured directly. Parenthetical numbers at the end of a series represent the greatest width found for that opening when the sheet has been cut between chain lines. Ellipsis points at the end of a series mean that what is presented is only a part of a sheet—usually either a half-sheet or, as with readings from a duodecimo, one-third or two-thirds of a sheet. The percentage sign [%] indicates a deckle edge; I have chosen it because it has no other special use in bibliography. When no examples of a deckle occur but tranchefiles or other evidence imply the original edge of the sheet, that symbol occurs in parentheses. Because in most formats the watermark spreads over several leaves, a record of the width and position of both the top and bottom halves is often an aid to identifying paper in which only one part of the watermark is present. The chain space model can simply be made two-tiered at that point. Thus, one of the varieties of paper with the eagle watermark in the fine-paper copies of the 1729 quarto Dunciad Variorum can be described in this way:

 

(%) (9)

30

29.5

27.5

31

28

29

28.5

29

29

(29)

29.5

30

 

28.5

4.5 [24.5 7.5 [21.5

26] 4 20] 10

29

28

19.5

20

(13) (%)

Some special chain lines, the tranchefiles, have features not characteristic of other chains and therefore provide a second group of variables to distinguish unwatermarked papers. A tranchefile is usually described as an extra wire between the final chain wire and the frame of the mould, and hence also as the extra chain line such a wire forms at the end of the sheet. Tranchefiles look like other chain lines, except that they are spaced closer than usual to other chains and, because the tranchefile wire typically does not have a supporting wooden rib, they lack the customary bar shadows. While tranchefiles often occur at both ends of a sheet, sometimes a tranchefile appears only at one end, and sometimes at neither. Kenneth Povey and I. J. C. Foster provide examples of each of these varieties of "close chain-lines" in their discussion of paper with turned chain lines.[8] The number of ends of a sheet which

have a tranchefile, then, serves as one identifying feature of a sheet.

The pattern of the chain spaces around the tranchefiles also varies. In one common pattern, the outermost chain (or tranchefile) line is usually 5 to 10 mm. from the edge of the sheet. A space of about half the usual distance between chain lines then follows before the first chain line of the normal pattern in the sheet. In the Dunciad comprising volume V of the small-paper octavo Works of Pope in 1751, for instance, typical chain spaces at the end of a sheet are: % 5 | 13.5 | 27 | 27.5 . . . . In most other cases where tranchefiles are identifiable in Dunciads, there are instead two narrow spaces between chain lines. A common paper variety in the first 1728 edition thus measures: % 11 | 20 | 21 | 25.5 | 26 . . . . When there is a pair of narrow chain spaces, only the penultimate chain seems technically a tranchefile, since the outermost chain often has the customary bar shadow. Such a chain line is not part of the normal pattern of chains in the sheet, however, since the combined distance from it to the tranchefile and from the tranchefile to the next chain is significantly more than the usual distance between two chains. Regardless of the source of the chain line, the number of narrow chain spaces, as well as the actual width of those spaces, provides additional ways of differentiating paper varieties. Although several writers have made use of the presence of tranchefiles, Stevenson has again provided leadership by using the absence of tranchefiles, as well as the actual width of the spaces adjacent to tranchefiles, as bibliographical evidence. More recently, Annemie Gilbert and Sylvia Ransom have observed that the patterns of tranchefiles and their adjacent spaces do indeed differ in eighteenth-century papers.[9]

A third way of identifying unmarked paper again involves chain lines. Much of the paper in the Dunciads of the 1720s through the 1740s has double chains: about 6 mm. to one side of a chain is a second one, so that the chain lines in a leaf look like sets of railroad tracks. In one of the few acknowledgments of this feature, Edward Heawood points out that such "subsidiary chain-lines" as well as countermarks in the corner of the sheet (he emphasizes the initials CM over T, a combination which is frequently found in copies of The Dunciad) are characteristic of early eighteenth-century paper from Genoa. W. A. Churchill illustrates

the double-chain pattern in paper used from 1599 through 1744; he too cites Genoa as its source.[10] When double chains occur, the shadows caused by the wooden ribs beneath the mould may be centered on each of the chains, just as in single-chain paper, or they may fall between the chains, as they do in all examples from The Dunciad. When a shadow lies between two chains, one of the chains is more clearly visible than the other. Even when the second chain is nearly invisible, the shadow to one side of the primary chain can be seen.

Double chain lines can thus provide distinctions by whether they are present and by whether the rib or bar shadows center on the chains, but they can also do so according to whether uncentered shadows appear to the left or to the right of the primary chain. For those arbitrary directions to be meaningful, there needs to be some way of orienting sheets of paper consistently. For watermarked sheets, the standard sheet position for reading watermarks (from the mould side, with the top of the mark toward the top of the sheet) is appropriate. Unmarked sheets also have helpful features. Often they have tranchefiles at only one end; for ease of comparison, I consider such lines to be at the left edge of the sheet. Occasionally sheets may have a broken or bent chain line which can serve as a means of positioning them consistently. Other times the chain space pattern as the bibliographer has established it will serve arbitrarily to distinguish "left" and "right." Unmarked half-sheets appear especially often, for they come not only from unmarked sheets but also from the unmarked half of each watermarked sheet. In half-sheets there can be only one deckle edge parallel to the chain lines and, at most, a single set of tranchefiles. If either is present, the secondary chain can be described as being either toward or away from that edge (or, similarly, to the outside or inside of the sheet). In lieu of that specific evidence, one can usually determine from standard imposition diagrams the leaf edges which represent the outside edge of the sheet. In octavo half-sheets, for instance, the outer edge of leaves 1 and 2 will usually be at the end of the sheet.

When measuring chain spaces in double-chain paper I usually have not recorded the exact position of secondary chains. My chain space models for such paper, therefore, indicate the positions of only the primary chains. One reason for the omission is that often the secondary chains are not clear enough for the precise measurement which is necessary to make meaningful a report of the distances involved. Furthermore, in the examples that I have studied, those additional measurements have not yielded significant variations beyond the ones revealed

by the pattern of primary chains alone. That is not to deny, however, that in other instances such measurements could be important.

I have found the "direction" of the shadows to be consistent throughout a sheet, with one class of exceptions. At the end of a sheet where a tranchefile is present and where the chain lines are parallel to the deckle edge, the chain space pattern might be: . . . 27 | 28 | 27 | 20 | 20 | 8 %. The variation is this: when the secondary chains appear to the outside (here, the right) of the main chains, the penultimate chain or tranchefile usually has no secondary chain, and the outer chain or tranchefile has a secondary chain to the inside (here, the left).

A fourth identifying feature not dependent on watermarks is the density of the wire lines. Like chain lines, these marks are a useful guide because they are present in all laid paper. There has been little published consideration of the recording of information about wire lines, though Gaskell and Stevenson reveal two main approaches. Gaskell recommends trying to record the spaces between the impressions, just as with chain lines: the bibliographer should "measure the distance across a particular number of marks and . . . divide the result by that number" (New Introduction, p. 76). Stevenson, on the other hand, quotes a writer at the turn of the century who reports the number of wire lines per inch.[11] In an important article on American paper making in the late eighteenth and early nineteenth centuries, John Bidwell shows that the ledgers of the mould-making firm N. & D. Sellers do in fact describe moulds in terms of the number of wires per inch.[12]

Because of the historical precedent and because of the small distance between the lines, the record seems better made in terms of density than of spacing. Eliminating the process of division both saves a step in the recording of the data and also yields more meaningful results. The difference in spacing which is created by an additional wire per inch in the common densities, for instance, is only about a tenth of a millimeter —a difference impossible to detect without special equipment. A bibliographer who wishes to compare paper at hand against the published measurement therefore either has to reconvert the published one to a reading for a larger distance or else carry out a mathematical operation upon the new measurement. It is much easier to report the number of wires per unit. I have found no problem that results from expressing the measurements for the chains and wires in slightly different terms; there is no inherent connection between the two anyway. The person

who does need to know the distance from one wire to another has available the numbers to make that calculation. On the other hand, including the density fraction does not mean that the spacing distance could not be added in parentheses.

The measurement should be made on the metric scale rather than in inches. Again standardization is an argument for the choice of metric, but again there is also a special practical consideration: because wire lines average about a millimeter apart, the researcher can make a quick initial estimation of the lines by seeing whether they occur more or less frequently than the marking for each millimeter on a ruler. I have found it useful to report wire lines by recording the number of lines per three centimeters (about 1.2 inches). That measure is long enough to record subtleties of spacing that a shorter one would have missed (e.g. 28/3, where wires in a single centimeter would probably be read as 9) while short enough to preserve the eyesight of the person counting. The number of lines per sample varies slightly in a sheet, but seldom by more than two. In fact, a difference of two wires per three centimeters is sometimes sufficient to distinguish paper samples. In analyzing paper in the first octavo edition of the 1729 Dunciad Variorum I was surprised to find that the paper varieties which I had identified by other means corresponded in all but one case with the division of wire lines between 24/3 and 26/3. That one exception from a book now thousands of miles away may have been the result of an inaccurate measurement and serves as a reminder of the need for vigilance even when one is confronted by the imminent closing hour of a library.

While the difference in wire line frequency is thus of use to modern students, there is evidence that the eighteenth century also used it as a basis for discriminating kinds of paper. The description "fine" was normally used as an indication of quality, with the other main categories being "second" and "ordinary."[13] But in Charles Ackers' printing ledger from this period, the term several times stands in opposition to "coarse," suggesting that the words refer more specifically to the texture and perhaps the wire lines of the paper.[14] The Oxford English Dictionary specifies that when "coarse" and "fine" are contrasted in the eighteenth century, "fine" indicates delicacy: "consisting of minute particles or slender threads and filaments." In the only Dunciad in which the size of an ordinary paper and "fine" paper issue are the same but the wire lines different (the quarto Dunciad Variorum of 1729), the regular

copies have wires of 24/3 while those of the special copies are 29/3. Bidwell's account of the Sellers moulds at the end of the century confirms this pattern of wire density: the wires in moulds for writing-quality paper had the thinnest wire and the greatest number of wires per inch, while moulds for printing and wrapping qualities were increasingly coarse (p. 312).

Other qualities of unmarked paper also have the potential to characterize specific varieties. Some are general qualities and provide help mainly by their presence or absence. Turned chain lines, for instance, merely distinguish the paper from the more common kind, unless uncut copies are available in order also to reveal from edges whether the paper was made as a double sheet or as side-by-side sheets. The basic distinction between turned and ordinary chains can be useful, though; in a 1742 edition of The Dunciad which appeared as volume III, part ii of Pope's Works (Griffith 579),[15] it instantly reveals which of the leaves are cancels. Various other qualities are not amenable to easy quantification or categorization. Variations in color in the leaves may be too slight to be distinguished without more elaborate measurement than a color chart provides, but they still may offer hints. In a slightly different manifestation of color, the sprinkling on the edges of at least some copies of the 1751 small-paper Dunciad has discolored at different rates according to the variety of paper on which it lies. The texture of paper is also difficult to describe. What Stevenson speculates might well be "knots from woolen underwear" in some paper does not serve as an adequate description of the paper's texture but does provide him with a clue for linking papers possessing different watermarks ("Tudor Roses," 27).

Finally, some features such as paper thickness may be quantifiable but ambiguous in their significance. Gaskell notes the possible variation of thickness within a given sheet (New Introduction, pp. 76-77); my own checking in The Dunciad has turned up variations of as much as 100% within a single leaf. But although it is unwise to assign paper to a particular group solely on the basis of its thickness, it does seem worth-while to record the thickness, both to have available the range of measurements for particular paper and to provide the information for determining an average thickness. I have found that when the readings from different groups of what is demonstrably the same paper are averaged, the result is the same, within a tolerance of ±.001 inch or .025 mm. Although none of these paper features provide as much help in identifying paper as the four discussed earlier, they often do provide clues that the paper researcher may find valuable.

The method providing the most precise identification of paper varieties, then, is that which establishes a record of the skeletal structure of a mould, the pattern of its chain lines and spaces. The procedure requires no specialized equipment, merely a ruler. In examining paper it is useful to apply as many tests as are convenient, of course, for the identity is thereby established not only more conclusively but also in many cases more easily. The various grosser differences can often help to define broader classes of paper and in the process also provide initial groupings for the more precise chain space patterns. In all attempts at identification the bibliographer is also aided by the knowledge that varieties were normally produced in multiples of two or, in the case of double moulds, four.

The primary difficulty with the procedures I have been suggesting, especially the measurement of chain spaces, is not a new one to the bibliographer: they take time. The efforts can ultimately save time, of course, in facilitating the identification of paper samples. In some ways radiographs or other reproductions would be preferable for such work; several narrow strips across the direction of the chain lines could provide the necessary information and serve as paradigms against which to check other samples. Practical difficulties with such reproductions are their cost and availability. Radiographic prints from the major libraries that can provide them appear to average about $10 each. Because the libraries often have the legitimate wish to retain the original negatives for their own files, the bibliographer who wishes to have them in order to compare patterns by overlaying them often needs to pay an additional $5 per reproduction to have duplicate ones made. As a result, even a record of each paper variety in a single book such as the first octavo Dunciad of 1729, whose sixteen different half-sheets would require at least thirty-two exposures for complete coverage of a single copy, would be prohibitively expensive. Photographic reproductions are potentially cheaper, but the libraries that have the necessary books are also the ones most likely to require the work to be done by their photographic departments—at nearly the same cost as the radiographs. A researcher who wants a reproduction of each watermark design does not fare much better. A watermark appearing in an octavo is often spread over four leaves and requires four exposures for each variety, with still more if countermarks are present. Apart from other considerations, then, a bibliographer may wish to use chain space models instead of reproductions for financial reasons.

There are also theoretical grounds for presenting chain space models instead of or in addition to reproductions. An obvious function of descriptive bibliography is to describe—that is, to give an account of the

material at hand in words or other typographical symbols. By presenting the chain space model, the bibliographer conveys information such as measurements not immediately evident from the actual leaf or its reproduction and also provides a standard form of the data that can be more easily re-used and transmitted. Although reproductions are particularly useful supplements to verbal reports of complex designs like watermarks, they provide little additional insight into the features of chain spaces.

Another problem the bibliographer must be mindful of is the temptation to force aberrant patterns of paper characteristics into the procrustean beds of established designs. Given the number of times paper is handled between its removal from the vat and its placement on the tympan, it is easy for anomalous sheets to slip into a grouping. For instance, one half-sheet gathering in one of the seventeen copies of the first octavo edition of the 1729 Dunciad that I have examined has the watermark of a horse inside a large circle. Some copies of James Bramston's The Art of Politicks produced by the same printer and same publisher the same year also have that watermark, and the lot of paper for that book may be the source of the stray sheet. What is important here is that the other half-sheet of that paper is probably unmarked and thus would prove puzzling if it were recorded in a Dunciad. Another danger in trying to fit unusual paper samples into existing models is that sometimes paper varieties may be represented by only one or two examples in a book. For instance, the sheets used for gatherings A and B of the 1743 edition of The Dunciad, in Four Books (Foxon P796) are similar to paper used throughout the volume but do not match exactly. The bibliographer who resists the temptation to group those examples with the established ones is rewarded to find those anomalies occurring as one of the main varieties in a companion edition of Pope's Essay on Man published in February 1744.