University of Virginia Library

Search this document 


collapse section 
collapse section 
collapse section 
collapse section 
collapse section 
collapse section 


The printing-shop activities collectively called typesetting and presswork leave physical traces—in addition to the intended type—impressions of textual matter—in the finished products of those shops, and such traces are therefore available for recording by descriptive bibliographers. These details have not, however, been extensively reported in bibliographies, often because theyhave been regarded as lying outside the scope of such works. “Typesetting” (or “composition”)[1] obviously encompasses the work of one ormore compositors at their type-cases, setting the types into composingsticks and then shifting the composed type, a few lines at a time, to galleys we may also let the term cover such related decisions (not made by compositors) as whether to compose the pages in their numerical order or by formes. “Presswork” may loosely be defined as the routines that occur between the setting of type and the folding and gathering of printed sheets. It would thus include such events (when they occur) as the combining of composed type with skeleton-formes (containing recurrent running-titles), the determination of which forme of a sheet is to be printed first, the arranging of the furniture around the imposed type-pages in the forme, the fastening of sheets on tympan-points, the insertion of press figures, the making of stop-press alterations in the standing type, the progressive shifting and deterioration of individual types and ornaments (or progressive damage to plates and woodcuts), and the working up of quads and bearers so that they leave impressions on the sheets. Although occurrences of this kind produce visible evidence, much of it was not meant to be noticed or was even (as in the case


Page 2
of type damage or the impressions made by quads) not meant to exist; and descriptive bibliographers (who inevitably must make a selection of visual characteristics to be noted) have frequently felt that such details could reasonably be omitted. The fact that the analysis of compositorial spelling and of skeleton-formes, for example, has largely been undertaken by analytical bibliographers as a prerequisite to scholarly editing, and the fact that textual press-variants must obviously be examined by editors, have reinforced the notion that typesetting and presswork details are more relevant to analytical bibliography and editing than to descriptive bibliography.

Such a view, however, misconceives the relationships among all three; and a more thoughtful consideration of whether typesetting and presswork details belong in a descriptive bibliography forces one to examine the connections between analysis and description. The reason that “analytical bibliography” and “descriptive bibliography” are frequently spoken of as if they were separate fields has more to do with their endproducts than with their methods. Bibliographical analysis—the process of analyzing and drawing conclusions from physical evidence in books— often results in articles and books that concentrate on specific problems. A bibliographical description, on the other hand, is an account of an entire book; it may include detailed analyses of particular features, but in its totality it must encompass (on whatever level of detail has been selected) a whole book. The descriptive bibliographer, like the author of any other kind of broadly conceived work, may not realistically be able to pursue detailed analyses of every aspect of the subject to be treated; but any analyses that have already been performed constitute part of the evidence that must be dealt with. Descriptive bibliography, in other words, employs analytical bibliography as one of its tools—a point recognized by Fredson Bowers in the opening chapter of Principles of Bibliographical Description (1949), where he states that it is “the basic function of a descriptive bibliography to present all the evidence about a book which can be determined by analytical bibliography applied to a material object” (p. 34; cf. p. 31).[2]


Page 3

What is generally thought of as description is not, in any case, a fully separate process from analysis. Describing what one sees without any interpretation is not always possible, for we all bring to acts of observation a repertoire of experiences and associations that we often employ, almost instinctively, in accounting for what we “see”; but even when pure observation can apparently occur, it is often unrevealing until it is analyzed in the light of an accumulated body of relevant inductive evidence. A descriptive bibliography employing the strictest sense of description would not be as useful as descriptive bibliographies generally are; despite their name, descriptive bibliographies are regularly and properly more than descriptive. Some details in them may result from seemingly pure observation, such as the dimensions of a leaf, the direction of the chainlines relative to the leaf, or the conjugacy of certain leaves; but a statement of format is a conclusion based on analysis, and it refers less to what one can actually see than to a process of imposition, printing, and paper-folding that took place before the book became the object that is now viewable.

Reconstructing sequences of printing-shop events involved in the production of specific books is an activity associated with the style of analytical bibliography that has developed over the course of the twentieth century as an essential ingredient in scholarly editing. This “New Bibliography,” as it was once called, differed from what preceded it in the degree of its emphasis on kinetic analysis—that is, analysis that deals not simply with identifying visual details (such as recognizing a particular type-design) but instead with inferring the actions of human beings that led to the specific array of physical evidence now before our eyes.[3] Thus patterns of variations in spelling, for example, can lead to inferences about which sections of text were set by one compositor rather than


Page 4
another; and patterns in the recurrence of specific settings of runningtitles can indicate how many skeleton-formes were in use.

That such details have largely been turned up in connection with textual investigations does not alter the fact that they are contributions to printing history. And it can scarcely be argued that any aspects of the printing history of a book are irrelevant to the bibliographical description of that book, for it has long been established that scholarly descriptive bibliographies on the highest level should aim to do more than simply provide a means for identifying editions, impressions, and issues. If they are to go beyond that function and offer histories of the production, finished appearance, and publication of particular editions, they not only must give fully rounded descriptions (recording details that are the same in all copies as well as those that vary);[4] they must also draw on all available evidence, both internal and external. Bibliographies regularly report information from copyright records and printers' and publishers' archives—information (reliable or not) that is often of a kind not inferrable from books themselves, such as their retail prices and publication dates or the number of copies printed. But in the process of supplementing internal evidence with external, bibliographers should not neglect the kind of internal evidence that is usually associated with the work of analytical bibliographers. The information it can disclose is just as much a part of the story as publication dates and edition sizes.

The criticisms that have been made of analytical bibliography in recent decades have no bearing on these points. One category of criticism simply reflects distaste for the perceived attitudes of the New Bibliographers and their focus on authorial intention;[5] it is easily dismissed


Page 5
since the validity of bibliographical analysis is a separate issue from the tone with which it is presented or the choice of editorial rationale within which the results of such analysis are employed. Another line of criticism, which has to be taken more seriously, does question the logical validity of bibliographical analysis as a procedure and the reliability of its results.[6] But pointing out instances of careless analyses and erroneous conclusions does not invalidate the approach itself, for the inductive search through the mass of physical evidence in books is the only way we have for attempting to extract some of the wealth of information embedded there. And although induction cannot lead to absolute certainty, it can, when handled sensitively and cautiously, contribute to a body of what may responsibly be called facts (which must always be tentative). In any event, the risks that must be confronted in bibliographical analysis are the same as those that face all reconstructors of the past. Descriptive bibliographers, like other historians, are constantly faced with questions of interpretation that require their informed judgment. Incorporating bibliographical analysis, or the results of it, into bibliographical descriptions does not inject into them anything epistemologically different from what is there in any case. This point, furthermore, applies just as much to the details drawn from external sources as to those derived from the books under examination, for archival documents are no easier to use, and no more foolproof in what they seem to be saying, than is the physical evidence found within books.

This line of reasoning leads to the conclusion that descriptive bibliographers have no valid theoretical reasons for excluding the kinds of details, such as those concerning typesetting and presswork, that have generally been considered the province of analytical bibliography. Whether in practice any given descriptive bibliographer decides to include them depends both on the nature of the material being treated and on the scale of the bibliography, as defined by the bibliographer. At


Page 6
the least it can be said that every descriptive bibliographer should consider the possibility of recording such details. Whether or not they are also analyzed, they can provide an initial body of evidence for other analysts. The descriptive bibliographer may sometimes use the results of previously published analyses, at other times engage in new analyses, and at still other times simply report relevant details without drawing conclusions—but in every case must employ well-considered evidentiary standards (which of course should be followed in every other aspect of the undertaking as well). The first prerequisite, however, is for bibliographers to abandon their all-too-common feeling (which exists largely through their failure to examine the matter) that such details are not properly part of bibliographical descriptions. Those details are perhaps the most obvious indication that the description is imbued with a knowledge of the book-production process, which should underlie everything the bibliographer places on record.

I am not aware of any descriptive bibliography that includes information on typesetting (as opposed to typography), and the only one, to my knowledge, that regularly includes a paragraph labeled “Presswork” in its entries is David L. Vander Meulen's A Descriptive Bibliography of Alexander Pope's DUNCIAD, 1728-1751 (1981).[7] This feature is just one of the innovative aspects of that pioneering dissertation, which deserves to be better known. His presswork paragraphs deal, as necessary, with one or more of four topics—first-forme impressions, point-holes, running-titles, and press figures. Two other landmark books that serve as major repositories of forms for reporting bibliographical evidence are Charlton Hinman's The Printing and Proof-Reading of the First Folio of Shakespeare (1963) and Peter W. M. Blayney's The Texts of KING LEAR and Their Origins (1982). Blayney's book is particularly useful for this purpose, with its appendixes on identifiable types, recurrent headlines, and press variants, and its trenchant commentary on the procedures involved in this kind of work should be required reading.

In the following pages, I shall draw substantially on the work of Vander Meulen and Blayney as relevant. What I propose is to make


Page 7
suggestions about the treatment, first, of two kinds of evidence (identifiable types and spelling variations) that are relevant to three aspects of the typesetting stage: whether the type was set by formes (rather than in page-number order), what type-case units were used, and how many compositors set the type. Then, under the heading of “Presswork,” I shall comment on typographical variations, impressions made by materials that were not meant to print, variations in furniture width, plating evidence, and the four categories of presswork evidence cited by Vander Meulen. The techniques thus far developed for analyzing typesetting are applicable mainly to books of the hand-press period; but several of the kinds of presswork evidence taken up here can be found in nineteenth- and twentieth-century books as well. (The analysis of paper, which has an extensive literature of its own, is not discussed here except when it occasionally enters into the analysis of presswork, since for the most part it is appropriately reported in the paragraph of a description dealing with paper.) Whether the result of examining these kinds of evidence should form a separate paragraph (or more than one) in a bibliographical description, or be incorporated into one of the now standard paragraphs, is something that each bibliographer must decide on the basis of what seems most convenient and reasonable in the given situation. In any case, the guiding principle is that forms for recording information ought to reflect understanding and promote further discovery.

I. Typesetting

The determination of whether a book was set by formes, as well as the identification of the type-case units from which certain pages were set (and in turn the shops where they were set), can often be made with certainty; but the history of attempts to identify pages set by different compositors shows how fraught with difficulty that effort is. These endeavors are extremely time-consuming and productive of unwieldy evidence. Whether or not descriptive bibliographers can reasonably be expected to include the full evidence in their descriptions (and it would not be unreasonable in every situation), they should at least understand the importance of offering some commentary on the matter.

The kinds of evidence primarily employed—damaged types and variable spellings—are in some ways similar to those used in certain categories of presswork analysis, but there is an important difference that should be kept in mind. The relevant type damage is damage that (at least in part) remains constant, so as to serve as an identifier of a particular piece of type; even undamaged types that remain undamaged, if


Page 8
they are of the wrong font or otherwise anomalous, can—in combination with each other or in conjunction with other identifiable types—help to identify a particular font associated with a given shop. And the spelling and layout variations relevant to compositor analysis are spelling and spacing inconsistencies in different parts of a book. In contrast, for presswork analysis the apposite type damage would be damage that first occurs or becomes progressively worse during the course of printing and thus is not identical in different copies, and the spelling variants would be those occurring at a single spot in different copies (they would be simply one category of stop-press alterations).

(a) Identifiable types

One way in which individual pieces of type can be identified is through the damage they sometimes suffer, either in the casting process or in subsequent use—as long as the damage is distinctive enough that there is no possibility of confusing it with that of another damaged type of the same letter. (In practice, the images of standard letterforms are primarily employed, and I shall concentrate on them here, but with a few added comments later on other typographic elements.) Although damaged types are not necessarily useful in identifying compositors (since the argument would involve showing that compositors always set from the same cases),[8] they can be decisive in determining whether the typesetting was done in page-number order (seriatim) or by formes (that is, according to the type-pages that would be on the press at the same time). Hinman, in a classic article of 1955, made clear how such a demonstration could be conducted; and by proving that certain gatherings of the Shakespeare First Folio were set by formes, he made a major contribution to printing history, for seriatim setting had previously been assumed for all books except page-for-page reprints. He thus inaugurated a considerable flurry of research, in which other scholars discovered evidence of setting by formes in other books of the period, and in other formats.[9]


Page 9

The procedure, which can lead to demonstrable proof, is based on the fact that, in seriatim setting, no piece of type could be reused in any of the pages that would have to be set before a complete forme was available to be put on the press. For example, in a folio in sixes (like the First Folio)—in which each gathering consists of three quired sheets, producing six leaves or twelve pages—no piece of type could reappear within the first seven pages of a gathering, since that many would have to be set before a full forme (the inner forme of the inner sheet) would be ready for the press. If an identifiable piece of type does reappear earlier, then the setting could not have been seriatim. Instead, it would have had to involve “casting off copy” (estimating the amount of copy that would fit on certain pages) in order to make possible the setting of selected pages out of numerical order, specifically the pages needed for a single complete forme. The motivations for setting by formes could have in-


Page 10
cluded saving type (since seven double-column folio pages would have involved tying up a great deal of type) or completing a book more quickly (by allowing printing to start sooner).[10] But the price paid for such benefits was the likelihood that the casting-off would not be accurate and that compositors would have to make adjustments to compensate for those inaccuracies. Thus the presence of unusually crowded or spacious typesetting—on those pages that had to be made to end at a preassigned point in the text—offers supporting evidence that copy had been cast off for setting by formes; but such evidence cannot be conclusive by itself, whereas evidence from damaged types can indeed be conclusive.

Blayney has shown, however, that arriving at a conclusive demonstration for a quarto is more difficult than for a folio because quarto formes contain less text and thus potentially less evidence than folio formes. Yet he was able to achieve sufficient density of evidence for analyzing the first quarto of King Lear by tracing over two thousand appearances of nearly six hundred identifiable types, and he could conclude decisively that the book was set seriatim. Although he presents his complex evidence clearly, there is no way that it could be recorded simply: he uses a “distribution table” for each sheet, showing how many of the identifiable types on each page of it came from each of the immediately preceding six formes (pp. 89-150), supplemented by a 36-page appendix listing the specific types both in alphabetical order and in page order (pp. 504-539). Hinman had previously used a similar three-part presentation for the First Folio, his tables appearing throughout Volume 2 and his appendix requiring 66 pages (1: 425-490), plus another seventeen on reappearing center rules (1: 491-507).

When the evidence supporting conclusions about the manner of setting is of this magnitude and already published, a descriptive bibliographer cannot be expected to do more than report the conclusions and provide a reference to where the full evidence can be found. Even when the evidence is less extensive and published in an article, it would still be a wasteful duplication for a descriptive bibliographer to repeat all of it. But what one can legitimately expect a descriptive bibliographer to


Page 11
do is evaluate any previous work on the subject and provide a critical summary of the present state of research, perhaps including a selection of the evidence.[11] If no previous work has been done, or none that seems convincing, the bibliographer should be encouraged to conduct such an investigation—which, after all, is likely to lead to more satisfyingly definite conclusions than the attempt to identify compositors (as we shall see below) and thus perhaps more worth the great effort involved. That this suggestion will be seen as unrealistic is an indication of how unaccustomed bibliographers are to thinking of typesetting analysis as an integral part of a description. Bibliographers are of course free to set as they wish the precise limits of their research in any given instance; but the amount of work involved would not seem so unreasonable if there were a wider recognition that determining the order in which pages were set is not outside the scope of a description.

There is another important use to which the evidence of damaged types can be put: it can contribute to the profile of a specific font of type—the particular assemblage of types constituting the contents of one set of cases (upper and lower). Although the basic idea of identifying type-case units in this way was Hinman's, it was given remarkable development by Adrian Weiss in two landmark articles in 1990 and 1991.[12] Weiss showed how one can identify certain types, even though they are undamaged, by virtue of their being used in conjunction with each other. Wrong-font letters, for example, can enter a type-case unit by accident or they can be placed there intentionally to replenish a depleted or damaged supply of type, but if they remain and are routinely used as part of that particular font, they become identifying characteristics of it, along


Page 12
with the damaged types. With sufficient study, therefore, one can identify the various type-case units from which type was drawn for setting a given book and can associate each of them with specific pages.

The result is of course not the same as identifying compositors, for a compositor cannot be assumed to have used the same type-cases consistently and exclusively.[13] The kinds of evidence relevant to the two concerns are very different. Type-case units can be identified with greater confidence because there are fewer variables involved in recognizing groups of associated identifiable types than in deciding which of the idiosyncrasies in the deployment of types are attributable to compositors' habits (rather than accidents or the exigencies of the moment). But the two kinds of investigation should not be regarded as totally independent: even if compositorial stints cannot be equated with groups of pages set from different type-case units, one can at least hold the working hypothesis that a single proposed stint should not include pages set from different type-case units (though occasional mixing of type between cases must be allowed for, especially in instances of serious type-shortages). And if one can show that a given type-case unit belonged to a different printing shop from that named on the title page, one has even stronger grounds for this hypothesis.[14] The determination of the shops that possessed the type-case units, once those units have been identified, involves considerable effort, for obviously one must examine many books with different printers' names stated in them; but Weiss has shown how fruitful such examination can be. Indeed, one of his main contributions is to alert bibliographers to the possibility that anonymous shared printing occurred in the production of sixteenth- and seventeenth-century British books more often than had previously been thought. And the identification of the printing shops responsible for a given book is surely an important element in a bibliographical description.

Weiss's work takes descriptive bibliography explicitly into account. In his 1991 article, he explains at the outset how the determination of the printer or printers of a book (going beyond a mere acceptance of a printer's name on a title page) is fundamental both for bibliographical description and for scholarly editing, since both activities rest on as great an understanding as can be achieved, for each relevant book, of “the


Page 13
circumstances of its production” (p. 183). He properly sees the function of descriptive bibliography as recording not only the “physical characteristics” of a book and “corresponding inferences about its printing history” but also “details which may contribute to the general understanding of early printing” (p. 184)—or, one may add, the printing of any period. The implication of this view is that even if one cannot positively identify the printer (or sharing printers) responsible for a given book, any “divisions of labor” that can be recognized, along with other typographical characteristics (such as those that are used in a typography paragraph in a bibliographical description), should be recorded as “initial clues to the identity of a font,” which other investigators will regard as “of utmost importance in the printer-search process” (p. 185). With Weiss's valuable practical hints for pursuing such work as an exemplary guide, bibliographers of all periods should be encouraged to extrapolate from his carefully stated principles and to regard their descriptions as repositories of typographical information.[15]

Such information can of course come from all printed images, including those of ornaments (headpieces, tailpieces, factotums, and ornamental capitals), rules, and punctuation marks, as well as of basic letterforms. But the latter are the most helpful for typesetting analysis because they constitute, in most pieces of printing, a much larger body of usable evidence. Ornaments and rules can certainly be distinctively damaged, and an identifiable recurrence can in some instances be as revealing as a recurrent standard letterform, but naturally there are not as many opportunities for ornaments and rules to be reused. And the commonest punctuation marks, though they occur more frequently than ornaments, do not suffer readily distinguishable damage as often as letterforms: commas and periods especially, given their shapes, offer more limited scope for identifiable damage, and one cannot always say with confidence that two seemingly identical images of a damaged comma or period were produced by the same type. Identifiable ornaments, when they do turn up, can sometimes provide additional evidence—along with that from font analysis—for recognizing work from particular shops; but the ease with which these items could be shared underscores the need for great caution in making such identifications. (See Blayney's acknowledgment of this point in the record of identifiable ornaments in Appendix III of his Texts of KING LEAR, pp. 432-503, and Weiss's comments on ornament damage and borrowing in his 1991 article, pp. 191-203.) McKerrow


Page 14
also warned (in An Introduction to Bibliography, 1927) of the dangers of assuming that ornaments between text units (“outside the text-page”) were necessarily put in place when the text was set (p. 33); they may not, therefore, furnish any compositorial evidence at all (but rather may reflect the imposition stage). In the end, although one should remember that identifiable recurrences of any typographic materials are potentially useful, one may realistically expect that the letterform images will play a larger role than other images in the analysis of typesetting.

One example of a concise pattern for combining the details thus far discussed may be in order:

TYPESETTING. Set by formes throughout (see [citation] from three type-case units, two of them (I, II) in the shop of —— (as named on the title page) and the other (III) in the shop of —— (see [citation]). The formes set from each type-case unit are as follows: I: A(i), B(i), C(i), D(o); II: A(o), B(o), C(o), D(i); III: E-G.

This is intended only as an illustrative specimen, not as a pattern for others to force their evidence into. Situations vary so greatly that each bibliographer will have to decide how best to place this kind of detail into the context of a particular descriptive bibliography. But the underlying principle to be kept in mind is that a description is an appropriate central site for typographical (as for other book-production) evidence; usably presented details—or citations of previous gatherings of them in printed or electronic form—are valuable elements in the cumulative body of observation on which printing historians can draw.

(b) Spelling and layout variations

Tracing inconsistencies in spelling through a book is one of the oldest techniques of bibliographical analysis: the theory is that, if a word is repeatedly spelled one way in one part of a book and another way in another part, these spellings may reflect the habits of different typesetters who set different sections of the book. If other varying spellings (and punctuation practices) fall into clusters that coincide with the same sections, the conclusion that one has identified the characteristics of two different compositors is of course strengthened. And features of layout over which typesetters had control (as they often did in Elizabethan and Jacobean times over the style of speech headings, the placement of stage directions, and the spacing of scene headings in plays) can be used in the same way to help work up profiles of individual compositors' habits.[16]


Page 15
In this process, one must naturally be aware of situations in which a compositor's usual habits might be less likely to operate (as in instances of type shortages or in justified lines, where uncharacteristic spellings might be used to make the lines fit),[17] and one must recognize the possibility that a page set by one compositor was corrected by another.

This approach to the determination of how many compositors set a given book (and which pages each of them set) is clearly a promising one. But in practice it has frequently led to results that are easily contested. One recognizes, of course, that as an inductive procedure it can produce only hypotheses, not certainties; but the range of variables involved, the multiplicity of evidence to be examined, and the numerous alternative possibilities it presents for arrangement into patterns have caused hypotheses about compositor identification to be perhaps the most easily overturned hypotheses in all bibliographical analysis. Nevertheless, the examination of spelling and layout evidence must continue; and if it is to be performed at the most rigorous level, it requires a consideration of all spellings, spacings, and design choices, not a selective search based on some prematurely formed notion of what appears to be significant.

Descriptive bibliographers, in dealing with this matter as with all others, must use their critical judgment to decide how strongly to support the results of particular compositor analyses, whether performed by others or by themselves. When previous analyses, and possibly conflicting ones, exist, descriptive bibliographers should study the evidence and present informed critical summaries of the situation. Whether or not they should perform additional independent investigations of their own—or perform the initial ones in instances where no previous analysis has been published—is a matter that each bibliographer will have to determine. This kind of analysis is extremely time-consuming, and some


Page 16
bibliographers will decide that as a practical matter they must limit themselves to assessing any previous analyses rather than making new tabulations on their own. But what they should not decide is that they can ignore the subject altogether. Given the possibilities for compositor analysis that exist, primarily but not exclusively in books of the hand-press period,[18] and the large body of work illustrating cautious and incautious ways of approaching such evidence,[19] descriptive bibliographers cannot simply avoid the topic as if information about compositorial stints and the effect they had on the printed pages were outside the scope of a bibliographical description.

The report in a description should, at a minimum, contain four elements (not necessarily arranged in this order): an indication of what work has been done to ascertain how many compositors set the book and which pages each one set; an assessment of the reliability of this work; a summary of the conclusions thus far reached; and a brief listing of the evidence employed. Such a report could consist of a single sentence (perhaps following the coverage of identifiable types in a “Typesetting” paragraph), as in this hypothetical example:

J—— D—— in 1948 ([citation]) argued convincingly that this edition was set by two compositors: one (characterized primarily by the spellings do, go, and here and the speech heading Hen) set A1r-B3v, D2v-G4r, L1r; the other (characterized primarily by doe, goe, and heere and the heading Henry) set B4r-D2r, G4v-K4v.

Obviously many variations on this pattern are possible, as circumstances and stylistic considerations dictate; but this sample shows that the bib-


Page 17
liographer's assessment of previous work can sometimes be conveyed by an adverb (“convincingly” in this instance) and that the evidence recorded can be limited to the strongest pieces (as indicated here by “primarily”). Any reader who wishes to have more information about where the cited spellings occur and how many of each there are, as well as what other evidence was considered, can refer to the original study (which will have more details, even if it, too, is a selective presentation).[20] In some situations one would have to report disagreements between previous investigators and try to adjudicate the debate, perhaps by an entirely new investigation that turns up additional relevant evidence. In the latter case, the bibliographer may wish to publish the detailed argument as an independent article, simply referring to its conclusions—as one refers to articles by others—in the bibliographical description.

II. Presswork

The eight kinds of presswork evidence commented on here are not equally available in books of all periods. For example, variants resulting from stop-press alterations (rather than cancel leaves) are more likely in books before 1700,[21] whereas press figures are largely confined to the eighteenth century and a few decades on each side of it, and plating is, for all practical purposes, a process of the nineteenth and twentieth centuries. One of the few kinds of evidence regularly found in books of all times is that resulting from typographical accidents; and evidence of first- and second-side impressions and of variations in furniture width can occur in books of any period. The comments below on those matters are therefore applicable to machine-printed books as well as earlier ones. There has not yet been much analytical work on books of the nineteenth and twentieth centuries, and the printing technology of those centuries makes the discovery of useful evidence particularly challenging. Some research, however, has been done on methods for detecting unmarked impressions, recognizing the use of plates, and identifying imposition


Page 18
layouts of machine-printed books,[22] and further work will undoubtedly produce more approaches to books of the past two centuries. The earliest printed books—those from before the early or middle 1470s—also form a class not fully amenable to some of the analytical routines worked out for hand-printed books of the following three centuries: since those early books were customarily printed one page at a time rather than from imposed formes, any of the techniques discussed below that assume the use of multiple-page formes do not apply to these books without adjustment.

It thus seems safe to say that no single bibliographical description will need to report on all the matters discussed below. An adequate account of any one of them, however, is likely to occupy a distinctive enough niche in a description that setting it up as a separate paragraph will perhaps in most instances seem the best course. For the sake of logic, when more than one are dealt with, such paragraphs could be linked as subdivisions in a comprehensive section headed “Presswork.” The following pages take up (in this order) evidence of (a) headlines; (b) point-holes; (c) first- and second-forme impressions; (d) typographical variations; (e) impressions from materials not meant to print; (f) press figures; (g) variations in furniture width; and (h) plating.

(a) Headlines

Many books contain headlines—made up of such elements as running-titles, rules, chapter numbers, and the like—at the head of the text-block on each page; and when the bulk of their content remains the same from page to page, printers sensibly held them in standing type for reuse. A set of such headlines for a single forme (two for folio, four for quarto, and so on) and any other material to be repeated on each page (like


Page 19
box-rules), all ready for new type-pages to be combined with them, is called by bibliographers a “skeleton-forme.” (Printing historians use the term to refer to all the reusable elements in a forme, including the “furniture” surrounding the type-pages; but since there is no way of proving the reuse of material that leaves no impression, analytical bibliographers limit their discussion of skeleton-formes to the elements that print.) The pattern of skeleton-forme use is a basic fact about the printing of a book. If, for instance, the same set of headlines recurs in a number of consecutive formes (and if no other set is repeated), the book was printed with one skeleton; but if two different sets recur (as when one set is used for inner formes and another for outer), two-skeleton printing was employed. It is not difficult, in many pre-1700 books and some later ones, to recognize a particular setting of a running-title when it reappears, for each one may have identifying characteristics such as broken types, distinctive spacings (not only within the title but also between its ends and the left and right type-page margins), or type-styles and spellings that do not reappear in other running titles.[23] Any recurrences of headlines (even single ones, rather than sets, in no discernible pattern) and any reuse of portions of them (as when the letterpress, or part of it, reappears independently of the surrounding rules) constitute facts of the printing history of particular books and are worth noting as evidence of the imposition process. But often headlines do recur in sets corresponding to formes, and bibliographers can then state with confidence how many skeleton- formes were used and where they reappeared.

The recurrence of headlines can sometimes be employed by bibliographers to establish the method of half-sheet imposition used to produce certain half-sheet gatherings: when, for instance, the same headline appears on two pages of a half-sheet gathering, the type-pages could not have been imposed for “work-and-turn” (where all the type-pages are on the press at the same time) but instead must have been imposed with other material to make up two full formes for ordinary printing.[24]


Page 20
Whether other details of printing-shop procedure can confidently be inferred from skeleton-forme analysis is less clear. Such analysis was a prominent feature of the post-World War II analytical activity, having been embarked on by Bowers and Hinman even before the war, at the beginnings of their bibliographical careers. A great deal of effort has been expended on attempting to extrapolate, from the pattern of skeleton use, the way in which proofreading was undertaken and the number of copies that were printed; but such analyses are now recognized as flawed because they postulate printing efficiency without taking into account other jobs simultaneously in progress in a shop and also because they make unjustifiable assumptions about the regularity and predictability of compositors' and pressmen's work-rates.[25] More cautious analysis may in the future be able to extract more reliable information from skeleton-forme patterns; but in the meantime the discovery of the recurrences of individual headlines (or portions of them) and whole skeletons is a significant contribution to understanding the production history of any book. And since these facts can often be established with certainty and relatively little effort (at least in books of the earlier hand-press period) and can be reported concisely, they ought to be regularly included in bibliographical descriptions. Indeed, Bowers's experience with


Page 21
running-titles caused him to say in the Principles that “with some books the precise information about the running-titles can constitute bibliographically the most important part of the description” (p. 188).[26]

There are obviously two stages involved in the fullest kind of reporting of headline evidence: first, describing the idiosyncrasies that make particular headlines recognizable; second, listing the recurrences of these headlines (or parts of them), once they are identified. In many instances the first of these can reasonably be omitted, on the grounds that the identification of individual headlines is often unquestionable (even if the differences are not always easy to detect without magnification or a collating machine) and that what is of primary interest for presswork analysis is the pattern of headline recurrence, not the peculiarities of the headlines themselves. On the other hand, those peculiarities are physical features that one might expect a bibliographical description to note, and there is always the possibility that, if the basis for discrimination among headlines is made known, other persons may find certain identifications unconvincing. This kind of evidence is more cumbersome to report (but not necessarily more space-consuming) than are the conclusions about recurrences, and each bibliographer must weigh the potential usefulness of the full report in a given situation against the labor and space involved. But regardless of whether the evidence for identifying headlines is stated, bibliographers should make the effort to identify them and should report any recurrences that are located.

A convention of sorts—using roman numerals to stand for identifiable headlines—grew up among the headline-analysts of the decades after World War II; but Blayney, in his 1982 book, calls this “a custom which should be abandoned as soon as possible” because such numerals are not only awkward but also undesirable (given their structure) for dealing with arrangements of units (p. 540), and he proceeds to suggest and demonstrate a clearly preferable system. His compact and usable style of report employs lower-case letters to stand for individual headlines, and


Page 22
they are listed for each forme in a standard order. In the example below for a quarto (adapted from p. 541), the order moves from 1r (for outer formes) or 2r (for inner formes) counterclockwise around the printed forme (clockwise around the forme of type):[27]

E(o)  b,   d,  f,  h: 
E(i)  The L. Cokes charge,  
F(o)  e2*   a,   c,  e,  g: 
F(i)  c2   giuen at Norwich Aſſiſes.  
G(o)  c3*  
G(i)  e2  
H(o)  c4  
H(i)  e2  

This plan makes it easy to see whatever patterns there are, whether the recurrences are of whole skeletons or only of individual (if sometimes variant) headlines. The order of the letters could be explained in words, as above (and as Blayney does), or by providing the following guide at the beginning of the table:

$(o)  1r   2v   3r   4v  
$(i)  2r   1v   4r   3v  

One could also place the outer and inner formes of each sheet on the same horizontal line, as follows:

(o)  1r   2v   3r   4v   (i)  2r   1v   4r   3v  
e2*   c2  
c3*   e2  
c4   e2  

It is true that the price paid for this compression is that the pattern is not quite as immediately discernible; in instances of this kind, where skeletons remain relatively intact but sometimes shift from inner to outer formes (or vice versa), one might wish to retain Blayney's layout. But in cases where the patterns are not as neat, less is gained by the more spacious plan. In an article two years later, Blayney altered the arrangement somewhat to report the following situation:[28]


Page 23

B(o)  D(o)  G(o) 
B(i)  D(i)  G(i) 
C(o)  F(o)  H(o) 
C(i)  F(i)  H(i) 

One could argue that the following presentation might be at least as useful, and perhaps more so:

(o)  1r   2v   3r   4v   (i)  2r   1v   4r   3v  

For formats with more pages per forme, it is particularly desirable to have along the top a readily accessible reminder of the order in which the headlines are listed; and for long books (whatever the format), if there are recurrences that are not easily noticeable, some discursive commentary on the tabular record would be in order. In any case, it seems clear that some version of Blayney's basic system deserves to become standard.[29]

In cases where running-titles on rectos are different from those on versos, the number of places where a given headline could be reused is of course reduced; and Blayney, as a way of informing the reader of the wording of running-titles, transcribes a sample (identified by an underlined letter) of a recto and verso title (as shown above).[30] Descriptive


Page 24
bibliographers often include this information (with details about the type face) in the paragraph on typography, and thus in a full-scale description following this practice there would be no need to repeat it here. What one might decide to do (as Blayney does not) is to present another list recording the evidence that was employed in identifying the headlines: either before or after the kind of table described above, one could simply run the letters (identifying headlines) down the left margin and for each one print a photographic reproduction of one of the occurrences of that headline. An alternative would obviously be to describe in words the identifying features of each headline—and, indeed, such a description might be helpful in any case, since those features might not be readily noticeable. A list of this kind may not always seem worthwhile; but there is no doubt about the importance of including a table like the ones illustrated above whenever headline recurrences have been located.

(b) Point-holes

The tympan of the hand-press (at least after the early years)[31] was furnished with adjustable pointed pins that could be fastened into position along the two longer sides of the tympan; when the paper to be printed was placed on the tympan, it was pressed over the points of these pins, which thus pierced the paper. These “points” served three purposes: they held the paper in place; they helped to assure proper register of the perfecting forme if the same holes were slipped over the points when the paper was turned over; and, through their particular placement, they made it impossible to turn the paper over the wrong way—if, again, the same holes in the paper were reused.[32] Philip Gaskell's A New


Page 25
Introduction to Bibliography (1972), in its section of imposition diagrams (pp. 88-105), shows the location of points in relation to the printed sheets in various formats and thus where one should expect to find point-holes in books of differing formats. Point-holes are of course frequently not visible in bound books, for in folio they fall in or near the fold, where a tight binding can make them inaccessible, and in quarto and octavo they come in the upper and outer margins, respectively, and are frequently lost in trimming. But they are often enough available (in untrimmed copies or unbound sets of sheets) that they have been a subject of discussion regarding their use in bibliographical analysis. David Foxon once suggested, for example, that the distance between point-holes might distinguish the work of different presses; but Kenneth Povey and K. I. D. Maslen found contrary evidence in which point-hole distances did not correlate with press figures.[33] Indeed, given the fact that the points were adjustable and that formes from duodecimo jobs might be interspersed between formes from jobs of other formats on any given press, it is an open question whether consistency is to be expected. Nevertheless, when point- holes are visible, they at least tell one how the points were set for that particular sheet (or some copies of it), if not necessarily for the whole book; examining their positions relative to the sheet can sometimes help to distinguish different impressions, and noting the leaves on which they appear can reveal the particular imposition used (what kind of half-sheet imposition in duodecimo, for example, or which of the alternative whole-sheet impositions in octavo and duodecimo). Their positions should therefore be recorded in a bibliographical description—in turn adding to the store of recorded instances of observable point-holes.

Vander Meulen, in his Dunciad bibliography, reports the locations of point-holes in relation to the leaf (viewed from the recto side), using a spatial arrangement of the distances from the hole (designated “x”) to each of the four sides of the leaf (with the figures in millimeters and “%” signifying a deckle edge), as in this example for a quarto (adapted from p. 251):


Page 26

B1,2  C1,2  D1,2 
126 × 120%  136? × 107%  136 × 110% 
301%  296%  294% 
B3,4  C3,4  D3,4 
163 × 80%  123 × 123%  152 × 89% 
301%  296%  300% 

This system is admirably clear but requires a great deal of space (several times in Vander Meulen's bibliography the point-hole record occupies a whole page or more), and in some cases one might wish to use a more concise form that conveys the essential information. One could, for instance, run the figures on in linear fashion, always giving the measurement from the hole to the top of the leaf first, followed by the other measurements in clockwise order:

B1,2: o|120%|301%|126. B3,4: o|80%|301%|163. [and so on]

But since the edge of the leaf at which these holes occur is known for a given format, and since a bibliographical description records elsewhere the format and the leaf dimensions, one could argue that as a general rule only one set of measurements need be given (along the edge where the hole appears), even though such a system would ignore the fact that the holes do not always occur exactly at the edges of the leaves as folded[34] and that not all leaves have precisely the same dimensions. This approach would turn the example above into some such form as the following:

Point-holes appear along the top edge of the leaf, spaced horizontally as follows: B1,2: 126|120%; B3,4: 163|80%; [and so on]

One may well ask whether the second measurement in each pair (or, in general terms, any measurement to an outer edge) is useful if the copy has been trimmed, since such a measurement would probably vary in different copies even if the position of the hole on the full sheet was in fact the same (and if the measurement does not vary, that fact would not necessarily mean that the hole was in the same position). Even so, measuring the distance from the hole to the cut edge(s) in every copy examined is useful, so that the largest measurement can be used as a minimum figure for the distance to the original edge(s).

One might also record the distance between the point-holes on each sheet (if both holes are visible), by adding together the intervening distances on the relevant leaves. These figures by themselves would not help readers visualize where the points can be found and might even be misleading, since identical distances would not necessarily mean iden-


Page 27
tical placement of points in relation to the two edges; but the measurement could be helpful for some analytical purposes and could be a useful supplement to the hole-to-edge figures, especially in instances of cut edges. The two approaches could be combined in the following way, for example:

Distances between point-holes vary, and the holes appear at the top edge of the leaf, as follows:

B 289 mm apart (B1,2: 126|120%; B3,4: 163|80%)

C 259? mm apart (C1,2: 136?|107%; C3,4: 123|123%)

D 288 mm apart (D1,2: 136|110%; D3,4: 152|89%)

If the distances between the point-holes do not vary, then one could of course make a much more compressed report, as in this hypothetical example:

Point-holes 290 mm apart ($1,2: 135|115; $3,4: 155|95), except in D, G, and R, where they are not present in the examined copies.

Recognizing the difficulties inherent in orienting point-holes to leafedges, one would be well advised (as Vander Meulen has suggested to me)—especially in formats where the measurements would be to two cut edges, top and bottom—to state the position of the holes in relation to the type-page, for that relationship should be constant even if successive sheets had been laid on the tympan in slightly different positions. One possible notation for this purpose would be to substitute for the vertical line a bracketed indication of line number (for vertical measurements) or underlined position within the top line (for horizontal measurements):

192 [line 33] 28  126 [were] 120 

Alternatively, one could place between the brackets a measurement from the top margin of the type-page (for vertical measurements) or left margin (for horizontal measurements); such a measurement would be somewhat more precise but would be less convenient for the user—since a great advantage of the former style is that it enables one to detect variation without measurement. Of course, when a hole is too near the corner of the leaf to be keyed to a line or word in the text, one would have to use a measurement from some point in the type-page in any case.

An explanation of the whole system, whatever form it takes, would appear in the preface or introduction to the bibliography, and one could at that point explain the “%” (or whatever symbol one chooses)[35] and


Page 28
the fact that horizontal leaf measurements are taken from the recto side (though if all relevant leaves have deckle edges, the latter point would be obvious from the appearance of the deckle symbol with the second of the two figures every time). Explaining the system fully in every affected description might seem cumbersome, but some minimal introductory explanation, as in these examples, should probably be used each time as an aid to the reader.

(c) First- and second-forme impressions

When sheets are printed from relief surfaces—that is, from metal types or relief plates, rather than by planographic methods such as photographic offset—it is frequently possible to tell which side was printed first and which second, for the side printed first may display an embossed effect resulting from the indentations produced by the perfecting forme. (The indentations made by the first forme, causing an embossed effect on the unprinted side of the sheet, are of course largely flattened by the perfecting forme, which thus leaves the primary embossed effect on the first-forme side.) When sheets were pressed between hot metal plates before folding (as Baskerville did), this evidence for the order of the formes was destroyed; but in other cases one can often uncover it by holding the paper in a parallel beam of light, which causes the embossed hillocks to cast shadows. The eye, however, often needs assistance in making the required discrimination; sometimes a magnifying glass is sufficient, but at other times one needs to use a special lamp of the kind described by Kenneth Povey in 1960.[36] Since the perfecting forme does not flatten every hillock produced by the first forme, and since one therefore can sometimes have difficulty judging which side of the sheet is rougher, an additional helpful technique is to search for hillocks that have ink on them, for these can only be on the side printed first.

It is worth the bibliographer's time to determine, when possible, the first- and second-forme impressions of each sheet[37] of the books being


Page 29
described because the order of formes is obviously one of the fundamental facts about the presswork for a book—and is thus valuable information in its own right, aside from whatever analytical uses it can support (such as the detection of cancels in instances where the paper evidence is not conclusive). The bibliographer must remember, however, that (as with other evidence) multiple copies have to be examined before conclusions can be drawn. If one copy of a book shows the first-forme impression of a given sheet in the inner forme, and another copy shows it in the outer forme, one may be able to conclude (despite Gaskell's doubts in A New Introduction to Bibliography, p. 132) that concurrent printing on two presses had been employed (each press printing some of the white paper with different formes and then exchanging the half-printed sheets for perfecting), although other explanations are possible (such as an interruption in the printing of a sheet, or the reprinting of a sheet). When one is dealing with a half-sheet gathering, variations between copies as to which side of the half-sheet shows second-forme indentations probably signify that the printing was done by the “work-and-turn” method (in which the same forme is used for both sides of the full sheet, producing two copies of the same half-sheet text).[38] Establishing such facts as these, even if the first-forme evidence is not used in any other way, supplies important details of printing history (relating not only to the books involved but also to the practices of the shops where they were printed).

Vander Meulen (in his Dunciad bibliography) records first-forme information as follows (this example adapted from p. 324, with “I” standing for inner formes and “O” for outer):

First forme] I: B-D; O: E; Mixed: F

The only addition one might wish to make to this sensible plan is a provision for indicating which copies show each pattern in the “mixed” category, as in this hypothetical example:[39]


Page 30

Inner: A-C, F-G, K. Outer: D, H, J. Mixed: E (inner, copies 2-5, 8, 11; outer, 1, 6, 7, 9, 10, 12).

Or, for books with multi-sheet gatherings,[40] one could append to each relevant signature (that is, the signatures for any multi-sheet gatherings in which all the sheets do not exhibit the same first- and second-forme pattern) either (1) the leaf numbers (joined with a period to show conjugacy) or (2) a bracketed figure identifying the sheet (numbering from the outermost sheet to the innermost—as in the following examples for a folio in sixes (where three sheets are quired together):

Inner: A1.6, A3.4, B, D2.5. Outer: A2.5, C2.5, C3.4, D1.6, D3.4. Mixed: C1.6 (inner, copies 3, 4, 7; outer, 1, 2, 5, 6, 8).

Inner: (A[1,3], B, D[2]. Outer: A[2], C[2,3], D[1,3]. Mixed: C[1] (inner, copies 3, 4, 7; outer, 1, 2, 5, 6, 8).

One might sometimes prefer to have a record arranged in sheet order, which in some instances might facilitate studying the sequential progress of the sheets through the press (if indeed the sheets were printed in signature order):

A(i), B(i), C(i), D(o), E(i,o) [2-5, 8, 11; 1, 6, 7, 9, 10, 12], F(i), G(i), H(o), J(o), K(i).

A1.6(i), A2.5(o), A3.4(i), B(i), C1.6(i/o) [3, 4, 7; 1, 2, 5, 6, 8], C2.5(o), C3.4(o), D1.6(o), D2.5(i), D3.4(o).

Povey in fact used this approach and combined with it an indication of which forme was printed on the felt (rather than the mould) side of the paper:[41]


Page 31

First forme 

One could of course rearrange this information in the style suggested above, subordinating the felt/mould record to that of first formes, by appending “felt” and “mould” to “Inner” and “Outer,” or “f” and “m” to “i” and “o”:

Inner felt: D, F, G. Inner mould: H, I, M. Outer felt: N, O. Outer mould: B, C, E, K, L.

B(om), C(om), D(if), E(om), F(if), G(if), H(im), I(im), K(om), L(om), M(im), N(of), O(of).

These unlabeled examples of course assume some such heading as “First-forme impressions,” since that is what is being listed. One might question the logic of emphasizing (or seeming to emphasize) first-forme over second-forme impressions, when the primary point is simply the order of the formes, whatever it is. But a heading like “Order of formes of each sheet” would result in a less efficient listing (since specifying first— or second—formes conveys by inference, but unambiguously, the information about the unnamed formes); and thus, in practical terms, making first formes the focus of the record seems acceptable.

(d) Typographical variations

Variations—in what is printed at a given place—among copies of the same impression (printing), or between impressions of the same edition, fall into two obvious categories: those following from deliberate alteration of the printing surface and those caused by the action of the press or the process of handling and inking the type or plates. The former may result from attempted textual correction or revision (and concomitant unintended changes) or from the effort to rectify some flaw in the visual appearance of the type images or their placement; when such variations occur within an impression, they are generally called stop-press alterations or, more simply if less descriptively, press variants. The other category—variations produced by the action of the press, by the inking process, or by handling between impressions—includes progressive damage to types, ornaments, woodcuts, or plates and shifting positions (or even loss) of types. (Study of the progressive damage to ornaments, ornamental initials, and woodcuts—as opposed to the damage to basic letterform-types—is one of the oldest techniques of bibliographical analysis and has long been used as evidence for dating or timing; a good


Page 32
recent example is Blayney's section on “Ornament Damage” in his Texts of KING LEAR, pp. 486-491.) Any attempt to report such details obviously requires full collations of the texts of multiple copies of each impression involved. Even with the aid of a Hinman, Lindstrand, or McLeod collator (one of which should certainly be used), this task is time-consuming, and many descriptive bibliographers have felt that it was not part of their responsibility to collate the texts of the books they describe, leaving this job to editors, who have no choice but to do so. Such bibliographers have thus placed themselves in the odd position of believing that the pages making up the bulk of a book demand less of their attention than the pages that come before and after the main text.[42]

This attitude is manifested in extreme form in those “bibliographies” that give quasi-facsimile transcriptions of title pages and not much else, as if the title page is somehow the most important page of a book. There is no question that title pages and other preliminaries, as well as such concluding pages as integral advertisements, can offer important clues to the differentiation of impressions and issues; but so can details discovered on the text pages, details that can be systematically searched for only by close scrutiny of every page. Variations found in this way are conceptually no different from those found on preliminary or concluding pages, and they are potentially more significant. Whether they reflect the course of presswork on a single impression or result from events between impressions, they are equally part of what a physical description ought to report. The former indicate states of pages, and the latter are sometimes the only features distinguishing separate impressions from each other. Even those bibliographers (largely of the past) who conceived author bibliographies to be simply guides to the identification of first printings recognized the importance of including such information; and certainly any bibliographer who understands that to be truly descriptive a bibliography must account for the entire book ought to see, as a corollary, that details found through a careful examination of the text belong in a description, whether or not they are the sole means for identifying impressions.


Page 33

The reluctance that bibliographers have felt toward textual collation does not stem merely from the amount of work involved, though that is often an element (for descriptive bibliography is demanding enough without this additional burden). A more fundamental reason that has allowed bibliographers to feel justified in regarding textual collation as dispensable, irrespective of the labor, is that texts are not thought of as part of the physical makeup of books. Yet they are physical, of course, consisting of inked letterforms and punctuation marks on paper or parchment. Verbal works, it is true, are intangible,[43] but texts of those works, when placed on material surfaces, become physical entities. Presumably an interest in an author's works, along with a recognition that the texts claiming to represent those works are affected by the process of their material presentation, underlies bibliographers' decisions to embark on descriptive bibliographies in the first place. But then, having made that decision, many bibliographers proceed as if physical descriptions and textual considerations were wholly separate matters.

Moreover, to compound the incoherence, they sometimes report “points”—textual or type-damage variations that are offered as identifying features to be checked by collectors, cataloguers, and others making use of the books—on a seemingly random basis, without the benefit of full collations to determine whether these are the only points, or even the only ones required for distinguishing impressions of books, or states of pages. They are thus admitting the relevance of textual variants while rejecting the necessity of a systematic search for them. Some of the “points” that have been cited, particularly in unsophisticated authorbibliographies of the years between the world wars, either are typographical errors that happen to have been noticed (and that in fact appear to remain constant in all copies) or are randomly chosen instances of variation that are misleading when used in isolation. (As an example of how they can be misleading, a single variant may be noted as the distinguishing mark of a second impression, when in fact there may be other variants in other gatherings, and different copies may show different combinations of these variants.) By the late 1920s, complaints about the abuses of “point”-evidence began to be heard, and the criticisms have continued, off and on, to the present.[44] Although the abuses


Page 34
were real, the attacks have sometimes had the undesirable effect of causing wholesale contempt for “points.” But typographical variations, when they exist, are important features to be aware of and must be called attention to; the problem is not in the “points” themselves but in the methodology sometimes employed for locating them and assessing their significance. And a defensible methodology must begin with thorough textual collations between copies of the same edition.

How extensively the results of textual collations need to be reported in descriptive bibliographies is a complicated question. In the abstract, there is no way to rule out anything discovered in either intra- or inter-edition collations as irrelevant to the aim of accounting for the physical objects that have attempted to transmit a given work. As a practical matter, however, there may in some cases be responsible ways of condensing material that would otherwise be unwieldy (and if it would not be unwieldy, there is no point abridging it). The basis for abridgment could well be the question of what is necessary for identification, even though that question does not determine the coverage of a descriptive bibliography; in other words, the matter of identification could dictate the degrees of detail with which various bodies of evidence are reported but would not cause any of them to be totally eliminated.

For example, textual variants (that is, variations in wording, spelling, and punctuation) between editions (separate typesettings) are unlikely to be the sole means of distinguishing those editions because resettings of type can hardly be so exact in every respect that they do not leave other, and more easily detectable, signs of their difference. If the textual variants are numerous, and especially if they have been (or are likely soon to be) reported in a scholarly edition, these variants could be handled in a descriptive bibliography by a general statement characterizing them, giving some examples, and citing any fuller account (such as the apparatus of an edition). Most major authors have been (or are likely to be) accorded scholarly editions, and descriptive bibliographies of these authors need not include full lists of variants between editions.[45] But


Page 35
when textual variants are relatively small in number or when they have not been (and are perhaps not likely to be) recorded elsewhere, they should be reported more thoroughly, preferably in their entirety.[46] What descriptive bibliographers should remember, however they finally decide to handle inter-edition textual variants, is that these variants are part of the story that a descriptive bibliography sets out to tell.

Textual variants and other typographical variations between copies of the same edition (intra-edition variants) sometimes vary between separate impressions and sometimes vary among copies of a single impression. In the first case (which rarely occurs in books of the fifteenth, sixteenth, and seventeenth centuries, when editions usually consisted of single impressions), the variations—resulting from textual alterations made between printings,[47] or type and plate damage that occurred in the handling and storage between printings—may or may not be the sole means for identifying the impressions. But in any event a full list of the variations is desirable in order to define thoroughly what each printing consists of. Furthermore, the possibility always exists that copies were made up of sheets from different printings, and an abridged list of variations may not allow one to determine whether a given copy consists of sheets from the same printing throughout. When a full list of such variations has been published in a scholarly edition or elsewhere, the descriptive bibliographer may wish to argue that an abridged list is defensible. But, at a minimum, it ought to include all textual variants and at least one variation (of whatever kind) for every sheet containing variations. And once this much is recorded, there would seem to be little point of inconveniencing readers by referring them elsewhere for the remaining items. Variations between printings of an edition (which in any case are not likely to be as numerous as variations between editions) should therefore normally be listed in full. Such lists of typographical differences are likely to be one of the main features of entries for second and later printings (along with any title-page and binding differences). They could take some such form as the following, which gives at the left the readings


Page 36
of the first of the printings being compared (keyed here to page and line, though signature notation could be added), followed by those of the next printing, the one being described:[48]

16.12  nations] matrons 
78.19  home;] semicolon missing through plate damage along right edge  
117.8  whenever] wherever 
207.21-23  many / efforts / not] the an, ff, and ot obliterated  

If the list were long, one might separate it into two sections, the first recording intentional textual alterations and the second the results of type and plate damage or other accidents.

The other kind of intra-edition variants, those among copies of single impressions, also sometimes requires full listing in a bibliographical description for a similar reason—to allow for the detection of different states of individual pages. Changes that occur to particular formes during the course of a single printing create states of the pages printed from those formes. Because a sheet printed from the first state of its inner forme and the second (or third, or fourth) state of its outer forme may be combined in a given copy with another sheet consisting of the first state of both its formes and with other sheets displaying various other combinations, one cannot apply the term “state” to entire copies of books but only to the affected parts.[49] As a result, there cannot be separate entries for states in a descriptive bibliography; states must instead be taken up within the entries for individual impressions, and some account of the variations involved is an essential element in a description of an impression, for without it an important physical characteristic of the impression as a whole would not be accounted for. In books from the first two and a half centuries of printing, the most common form of intra-impression variation results from stop-press textual alterations. In later books, from a time when textual changes were more likely to be made by cancel leaves and gatherings or between printings, the intra-printing variations are apt to be those caused by progressive type or plate damage, variations in inking, and shifting type during the course of printing. Stop-press alterations, which create textual variation through the deliberate in-


Page 37
sertion, deletion, or other adjustment of pieces of type (and sometimes through inadvertent type-changes that are by-products of the process), should be listed in full; but the instances of variation that result from the action of the press may usually be reported in the form of selected examples, with some comment generalizing about the extent of such variation.

The actual manner in which all these kinds of typographical variation should be recorded cannot be prescribed in advance for every situation, but in general it can be said that a formulaic listing is probably the clearest and most easily usable form—always supplemented, as necessary, by discursive prose. Stop-press textual alterations are regularly listed in scholarly editions in a form similar to that used for other categories of textual evidence.[50] And though this form of course differs somewhat from edition to edition, a plan that was made prominent by virtue of its use in Fredson Bowers's editions of Dekker (1953-61), Beaumont and Fletcher (1966-96), and Marlowe (1973) is the following (condensed from the first volume of the Dekker, p. 94):

This form establishes the kinds of information that any such listing should contain: arrangement by forme, identification of the order of the variants (if possible), and specification of which copies contain each reading.

Since Bowers's listing was designed for an edition, it is understand-


Page 38
able that he placed the “corrected” reading—the one in his text—first; but in a descriptive bibliography, the readings should be placed in chronological order, as Bowers in fact does in the bibliographical descriptions appended to his Fielding edition (1967-).[51] (Although chronological order generally means listing uncorrected readings before corrected ones, it is important to remember that attempts to correct the text do not necessarily result in actual corrections, and the bibliographer should probably call attention to instances of this sort; furthermore, the sequence of the variants is not always clear, and the order of the listing can then be only arbitrary.) In addition to rearranging Bowers's entries in this way, one could make the record much more concise, with no loss of clarity, by using numbers to stand for each copy and placing all the information relevant to each entry on a single line:[52]

C(i) 2r I.iii.13  be] 1; be: 2-5 
4r I.iv.98  how] 1; hoe, 2-5 
4r I.iv.111  readie,] 1; readie? 2-5 
4r II.i.1  farre] 1; far, 2-5 
E(i) 4r III.ii.56  how,] 1; how now, 2-5 

In this sample I have retained the act-scene-line numbers because doing so would enable readers to use the list in conjunction with editions other than the one being described; but normally in a descriptive bibliography one would give only the page-line numbers of the book under examination (whereas in an edition it is always helpful to provide a reference to the newly edited text as well). Often there is evidence of more than one stage of stop-press alteration, as when some copies show one correction in a given forme, and other copies have another correction in addition


Page 39
to that one. The above plan could be adapted to accommodate this situation by arranging the lines in the order of correction-stages and placing a superscript figure after the forme designation—as in “C(i)2”.

As for the recording of variations produced by the action of the press or the handling and inking of the formes, there is less precedent to follow, but the same general plan can generally be employed. Shifting types, for example, could be recorded as follows, using empty brackets to signify spaces between types.[53]

B(i)5v: 26.5 handing] copies 2, 6, 7; han[]ding 3, 5, 9; hand[]ing 1, 4, 8

G(o)4v: 104.8 bookstore] 2, 7, 9; book[]store 1, 3, 6; boo[]kstore 4, 5, 8

The amount of space represented by the empty brackets may vary among the copies cited; it is not the purpose of the listing to record every slight movement of type that has been noted but only to call attention to the fact that shifting does occur at certain places on the text pages. Progressive damage to the letterforms is more difficult to represent. Xerographic or photographic reproductions of the damage in particular copies could of course be incorporated into the listings, but the value of such precision is seldom enough to justify the trouble involved (in contrast to the precision required for identifying types in the study of typesetting). Noting the presence of progressive damage in the following manner is probably adequate:

C(i)2r: 35.10 descriptive] 3, 8-11, 13; the p shows progressive damage from each of the following groups to the next: 2, 12; 5; 1, 4, 6, 7

E(o)8v: 80.17 bond] 4, 7, 10, 12, 13; the o is damaged so as to appear an e in copies 1-3, 5, 6, 8, 9, 11

K(i)6r: 155.3-17 no damage at right edge] 1, 3- 6, 8, 13; plate damage along right edge, causing comma at end of line 14 to vanish, in 2, 7, 9-12

For entries of all these sorts, such a style is primarily useful when more than two or three examples are given; otherwise they can simply be presented in ordinary sentence form.

Not every copy examined need be cited in such entries, and nothing need be implied about the readings in any copies not cited (though of course one can, if one has the information about every examined copy, use an inclusive system, such as saying that all copies not cited conform to the reading given at the left, in instances where such a procedure would save enough space to make it worthwhile). The other way in which these lists can be selective is that they need not record every spot


Page 40
at which shifting type, inconsistent quad impressions, and progressive type or plate damage occur. These occurrences are generally of no significance individually, but collectively they constitute a fact about the printing history of the impression; a selective list accompanied by a generalization about the extent of such occurrences is normally sufficient. But whenever the shifting or damage creates a textual variant—as when a space appearing within a word creates two separate words (see the example labeled 104.8 above) or when a letter or mark of punctuation develops damage that makes it appear to be a different letter or mark (see the example labeled 80.17 above)—the individual instance does have significance because it would cause anyone using a copy containing such an instance to misconstrue the text. All instances of this sort should therefore be included in a listing, or indeed the listing could be limited to them, with the attached commentary noting the extent of other shifting and damage that do not create textual problems.[54]

References to type and plate damage, whether it occurred between printings or during the course of the press run for a single printing, serve to remind one of a requirement for the collating of copies from the same edition: the necessity—whether one uses a single control copy for all collations or different pairs of copies in order to multiply more rapidly the number of collated copies—for noting not merely all the places that differ in any way but also some of the anomalous occurrences that are identical, such as instances of type or plate damage, wrong-font letters, odd spacing, and shifting type, which could be spots that vary in other copies.[55] Unless one makes note of these places, even though they are the same in the two copies being collated at any given time, and then checks them in numerous other copies, one may never discover certain variations, for one can easily have the bad luck never to bring together for full collation two copies that differ at particular places. Many of these notes made during collation will probably not find their way into one's published lists, but they are useful both for providing a guide to problem spots to be checked in copies that one is not going to collate com


Page 41
pletely and also for giving one a fuller basis for making generalizations about the presswork for a particular printing. In every case the reader should be told what degree of thoroughness was used in checking copies and what principle of selectivity was followed in reporting the results.[56]

(e) Impressions from materials not meant to print

Occasionally quads (and other material meant to hold space between types or to keep the type in place) work up to the level of the printing surface and leave blind or inked impressions on the paper. Similarly, bearers (type, wood, or other material used to fill in areas without type in order to keep the platen from tilting when it makes contact with the forme) sometimes leave tell-tale traces on the paper (for two examples, see the fourth and fifth illustrations in Roger E. Stoddard's Marks in Books, 1985). In one sense, such occurrences might be thought of as a subset of the category of typographical mishaps, just discussed. But the difference (aside from the fact that they are not, strictly speaking, typographical) is that they are an aspect of presswork whether or not they vary within an edition. When the unintended typographical events discussed above are identical in all copies of an edition, they are not the result of the presswork process but rather reflect the state of the type when it was composed. But impressions from quads and bearers, even when they remain the same, are often the result of faulty make-ready and locking up of the forme or the action of the press itself. (It is true there is the possibility that some flaw in the setting of the quads was ultimately responsible; but since this possibility is relatively unlikely and cannot in any case be determined, it is best to treat all such instances as aspects of presswork.) There are thus three kinds of situations to report—(1) when all copies of an edition are identical:


Page 42

D(o)3r: 53.7 inked impression of a quad visible between and and when

M(i)4r: 183.2 row of blind impressions of quads fills out this line

(2) when there is variation between printings:

D(o)3r: 53.7 and when] inked impression of a quad appears between these words

M(i)4r: 183.2 no blind impression visible] row of blind impressions of quads fills out this line

and (3) when there is variation among copies of a printing:

D(o)3r: 53.7 and when] copies 9, 13; blind impression of a quad is visible between these words in 2, 5, 8, 11; inked impression of the quad appears in 1, 3, 4, 6, 7, 10, 12

M(i)4r: 183.2 no blind impressions visible] 1, 3-10, 13; a row of blind impressions of quads fills out this line in 2, 11, 12

When types are the objects serving as bearers, they often consist of units of composed types that had been, or were going to be, used for printing elsewhere; thus a whole line or passage of text may be visible in blind impression in areas of a page that were supposed to be blank. In these cases an effort should be made to identify the text whose types were used as bearing material (with the help of raking light if necessary), and the record should indicate where it occurs in inked form (generally on a page not far away, but sometimes in another book that was being produced in the same shop). Having identified the text and noted the relationship (in terms of formes) between the locations of its inked and blind impressions, one may then be able to draw some conclusions about the order of printing of certain formes. Randall McLeod, in his 1999 Rosenbach Lectures, has demonstrated—more fully and clearly than anyone else—what this kind of analysis can accomplish. At times combining this evidence with that from identifiable recurrent types and headlines, as well as from the embossing that reveals the order in which the two formes of a sheet were printed, he shows, for example, that one gathering in the 1528 Aldine edition of Castiglione was printed from the inside out and another from the outside in. He also discovers, in the case of two other Aldines, that a headline in a book dated 1504 was used as a bearer in a book dated 1503, and he then offers an account of the overlapping production of the two, recognizing that still other books were no doubt being worked on at the same time. His approach is clearly complementary to such techniques as those developed by Adrian Weiss for identifying fonts and “fill-in” jobs: all help to build up a picture of the concurrent activity in a shop and thereby to add precision to the production history of individual books. After McLeod's work,


Page 43
bibliographers must carefully scrutinize any “blank” areas on the pages of the books they examine, for those areas may contain blind impressions (less distinct in some copies than in others as a result of repressing during rebinding) that furnish valuable information about the presswork process.[57]

(f) Press figures

The numbers and other symbols now known as press figures—notations that are clearly not signatures but that are found in the lower margins of some pages in many eighteenth-century British and American books (and were occasionally used in the decades just before and after


Page 44
that century)—have been extensively studied only since the middle of the twentieth century.[58] Before then, the few bibliographers who paid attention to them were primarily interested in the variant figures that sometimes occur in different copies of the same edition.[59] But beginning with two works of 1949—an article by Walter E. Knotts and the important dissertation of William B. Todd—there has been a growing recognition of the value of recording all the press figures in a book under investigation, not just the variant figures that happen to have been located in the copies examined.[60] It was Todd's dissertation, in fact, that caused Bowers to make a last-minute addition to the Principles stating that these figures should be “minutely recorded in a shorthand manner for every forme


Page 45
and sheet of a book in an eighteenth-century bibliography” (p. 321).

In the discussions that followed in the 1950s and 1960s, with their scrutiny of a great many figured books, certain puzzling questions were repeatedly addressed and remained unanswered—most notably whether the figures refer to pressmen or presses and how to interpret the absence of figures in some formes of books that contain them in other formes.[61] Despite the fact that the precise printing-house function of figures is not yet entirely understood, it does seem clear that they are a product of the presswork phase and have something to do with accounting for the work done by different press teams. Some of the ways in which bibliographers have used them, such as considering them independently sufficient clues for distinguishing impressions or determining the order of formes through the press, have displayed the same incautious optimism that characterized certain other efforts of bibliographical analysis in the years following World War II. But there is no question that figures can sometimes help to determine imposition, as when two figures show up in a half-sheet (ruling out the work-and-turn system) or when a figure appears on a duodecimo page that in one possible layout would be in the inner rank of pages in the forme (making more likely an alternative layout in which that page could more easily be reached).[62] Whatever use a descriptive bibliographer puts figures to, the fact remains that they are prominent vestiges of the book-production process and that there can be no excuse for failing to record them in descriptive bibliographies.

How best to do so, however, has never been agreed on, and several different systems have been tried. Knotts's 1949 article used a table in which horizontal rows represented different copies and vertical columns the signatures; in the appropriate squares page references and figures were entered, in the form “B1v-1 (inner).” Though clear enough, this system would not be practical for dealing with a large number of books or with large books, for two-thirds of a page is occupied by the table for eight copies of The Beggar's Opera, a book with only six gatherings.


Page 46
Philip Gaskell, the following year,[63] set up charts with figures along the left and signatures across the top; whenever a particular figure appeared on a given sheet, the fact was indicated by a diagonal line in the appropriate square (or crossing lines when the same figure occurred in both formes). Again, this notation is not economical of space, since each chart serves only for a single copy; in addition, there is no indication of the page (or the forme) figured—essential information for a study of pressmen's habits or shop policy—although of course a page reference could be substituted for the diagonal line.

At about the same time, Todd began his important series of articles following from his dissertation[64] and recommended a simple and concise notation that has been frequently used by others since. It is similar to Knotts's in consisting of page-dash-figure, but it does not involve a chart, only a list run on in paragraph form. Todd has used both page numbers (“18-6”) and signature references (“D2v-2”); though the second form takes slightly more space, it conveys much more information conveniently (sheet and forme), and either method makes it possible to list all the figures in a book, with their precise locations, in the space of a compact paragraph (or in one column of a composite chart).[65] This condensation, however, is achieved by the sacrifice of convenience in extracting certain other pieces of information: thus it is difficult to tell, without resorting to a tally of one's own, just how many figures appear in a given book, whether or not particular pressmen had preferred pages for figuring, and what kind of division of labor the figures suggest. The method is not readily adaptable to recording variant figures, and Todd later used a table in which each horizontal line represents the agreement of specified copies on the figures in a given sheet (both inner and outer formes), so that variant copies require additional lines.[66]


Page 47

Still other methods have been tried. Kenneth Povey in his extensive 1959 survey, “A Century of Press Figures” (see note 33 above), examined 111 figured books and recorded the figures for individual books by placing the page references (in terms of signatures) in horizontal rows with the corresponding press figures immediately below; separate pairs of lines were used for outer and inner formes. This system provides for a great deal of information but is somewhat cumbersome and space-consuming. Povey supplemented it in one instance by an extremely convenient summary table showing how many times each figure had appeared in each possible location (p. 255), an invaluable aid when the intention, like Povey's, is to demonstrate something about the nature of press figures and not simply to identify different printings. Two years later Allan Stevenson, in his comprehensive introduction to the second volume (1961) of the Catalogue of Botanical Books in the Collection of Rachel McMasters Miller Hunt, included a thoughtful discussion of the whole problem of recording press figures (pp. clxiv-clxvi) and recommended grouping forme or page references according to figure, as “1:iA iB oC,” or “1:A7v B1v C8v.” He recognized that such additional space as is required by the second style may be well worth while for the additional information about figuring practice that can be derived from it. Philip Gaskell, in his Soho bibliography of the Foulis Press (1964), used the same system but condensed it by specifying that all leaf numbers referred to versos unless rectos were indicated: “`1' on Vol. 1 A2 B1... D1 F1 G1 I1 L1 N6.”

It can be seen, from this sampling, that the two basic approaches have been (1) a sequential list of signatures, formes, or pages with the corresponding figures indicated for each, and (2) a list arranged by figures, with the signatures, formes, or pages on which each figure appears grouped by figure. The first lends itself more readily to a study of the process of printing, the second to an examination of the division of labor and of individual figuring habits. Both are relevant to any bibliography conceived of as a record and not simply a guide to identification (which would in any case require some kind of full listing to aid in the detection


Page 48
of variant figures). Furthermore, the greater the body of data that is accumulated about press figures, the more reliable will be hypotheses about their nature; and it should be part of the descriptive bibliographer's duty to help build up the body of printing-practice data by recording figures as fully as possible when they are encountered. Only if the statistics are clearly presented by each person who has occasion to deal with a figured book—and conveniently arranged for extracting several kinds of information—can the practical difficulties of assimilating complicated data about thousands of books be eventually overcome. For these reasons, I suggested in 1966 a system that involves two kinds of table: one, which could be called a “sequence table,” listing figures (and their locations) in the order of their appearance in the book; the other, a “summary table,” bringing together such information as the number of inner and outer formes figured and the number of times each figure appears in each location.[67]

The sequence table records, for each signature, any figures that appear within the gathering labeled by that signature. In the case of books printed in the usual fashion, there are two columns for each signature, one representing the inner forme and one the outer; in the case of half-sheet imposition by the work-and-turn method, only one column is required. Whatever page normally carries the figure is indicated at the top of the column, and all figures appearing in the column may be assumed to occur on that page except when a different page is entered in brackets after the figure. Thus, in the table below (for about half—the half that contains all the figures, as it happens—of Evans 35106, Transactions of the American Philosophical Society, vol. 4, Philadelphia, 1799), inner formes are usually figured on $4r, but inner I has its figure on 1v instead:

(i)  (o)  (i)  (o)  (i)  (o) 
4ra   4va   4ra   4va   4ra   4va  
—  *4  —  *2  Bb  *2  — 
—  *4  —  *4  Cc  *2  — 
C2   *2[2va —  *1  Dd  —  *2 
—  *4  —  Ee  —  *4[2va
—  —  *4  Ff  —  *1 
—  —  *4  Gg  — 
—  —  —  Hh  —  *2 
—  —  *4  Ii  —  *1 
*2[1va —  —  *2  Kk  *4  *2 
—  *4  —  *4  Ll  *2  — 
—  *1  —  *1  Mm2   *1[1va
—  *2  Aa  *4  *2  Nn2   1[2va


Page 49

Information about size and position of figures is also included in the table: an asterisk marks those figures that are larger in size than the others in the same volume, and superscript “a” and “b” indicate the left and right halves of the lower margin, whether or not the letterpress appears in two columns. If the figures recorded here had been reported in one of the more usual compressed forms, the results might look like one of the following:

8-4 16-4 20-2 28-4 36-2 43-4 52-4 60-4 62-2 76-4 84-1 92-2 100-2 108-4 116-1 123-4 132-4 140-4 156-4 164-2 172-4 180-1 187-4 188-2 195-2 203-2 212-2 216-4 228-1 236-2 244-2 252-1 259-4 260-2 267-2 270-1 276-1

A4v-4 B4v-4 C2v-2 D4v-4 E4v-2 F4r-4 G4v-4 H4v-4 I1v-2 K4v-4 [and so on]

A8-4 B16-4 C20-2 D28-4 E36-2 F43-4 G52-4 H60-4 I62-2 K76-4 [and so on]

“1” on pp. 84, 116, 180, 228, 252, 270, 276; “2” on pp. 20, 36, 62, 92, 100, 164, 188, 195, 203, 212, 236, 244, 260, 267; “4” on pp. 8, 16, 28, 43, 52, 60, 76, 108, 123, 132, 140, 156, 172, 187, 216, 259

1: L4v P4v Z4v Ff4v Ii4v Mm1v Nn2v. 2: C2v E4v I1v M4v N4v X4v Aa4v Bb4r Cc4r Dd4v Gg4v Hh4v Kk4v Ll4r. 4: A4v B4v D4v F4r G4v H4v K4v O4v Q4r R4v S4v U4v Y4v Aa4r Ee2v Kk4r.

Although the last of these makes the basic pattern somewhat more apparent than the others, it is only with considerable difficulty that patterns can be discerned in any of these rather bewildering lists: it would take concentrated study to observe, for example, that the only sheets figured in both formes are Aa and Kk, that the only unfigured sheet is T, or that inner formes are figured only eight times—facts easily observable in the table given above.

The advantages of this proposed sequence table are that one can see at a glance the deviations from a dominant pattern; one can easily study the patterns for outer and inner formes separately or in conjunction; and one notices where the unfigured formes fall in relation to the figured ones. It seems essential, if one is to have an overall view of a complex arrangement of figures, to record the figures in a concise fashion that, at the same time, does not sacrifice clarity of sequential and spatial arrangement to compression. Such tables as the one above not only reveal whatever pattern there is in the figures but also are economical of space and effort—for one table can serve as well for ten volumes as for one, simply by having separate columns for each book with only one set of signatures down the left side (similar to the arrangement Todd uses occasionally in his Burke bibliography). If one is trying to investigate the possibility of


Page 50
concurrent printing, for instance, or any other aspect of the relations among a number of volumes printed in the same shop during the same period, these multiple tables facilitate comparisons. The leaf references seem preferable to page numbers since they show immediately the position of each figure in terms of the structure of the gathering. (Since the length of individual books recorded in such a table will vary, dashes can be used to indicate unfigured formes or half-sheets, while signatures that do not appear in any given book can simply be left blank.) Similarly, the record of variant figures in multiple copies of the same title can be neatly displayed by assigning columns to various copies.

The second kind of table in my suggested system is part of a “summary of figures” for each book. In long books containing several figures it becomes especially helpful to have statistical totals easily available. This summary, then, indicates the number of sheets figured in both formes, figured in only one forme, and entirely unfigured (or alternatively, whenever appropriate, the number of half-sheets figured). It includes a table (like the one of Povey's mentioned earlier) that records the number of times each figure appears on any given page of a gathering. These statistics are totaled to provide data on the number of inner and outer formes figured, the number of times each figure appears in inner or outer formes, and the number of times figures occur in each possible position. In the table below (again for the sample of Evans 35106), figure 2, for example, appears on the first leaf verso of one full-sheet gathering, three times on the fourth leaf recto, and nine times on $4v; it appears on four inner and nine outer formes, or thirteen times in the full-sheet gatherings of the volume:

Summary of figures: 2 of 33 sheets figured in both formes, 30 figured in one forme, 1 unfigured; all 3 half-sheets figured. Figures appear near left margins.[68]

Full sheets  Full-sheet Totals  Half-sheets  Grand Totals 
Figs.  1v   4r   2v   4v   1v   2v  
0(i)  5(0) 
4(i)  9(0)  13  14 
11  4(i)  12(0)  16  16 
Totals  25  8(i)  26(0)  34  37 


Page 51

It can also be readily seen from the table that $4v is by far the most popular page for figuring in the volume (25 times) and that outer formes are figured more than three times as often as inner formes (26:8).

Varying totals resulting from the discovery of variant figures in different copies may be included in the table by separating the varying totals for the same figure and location with short dashes and arranging them in the same order as the columns for the corresponding copies in the sequence table; thus, in the previous example, if three printings had been discovered through variant figures, the full-sheet portion of the first row of the summary table might read as follows:

Figs.  1v   4r   2v   4v   Totals 
0-1-0  0-0-1  0-1-2  5-4-4  0-1-1(i) 5-5-6(0) 5-6-7 

The information in these summary tables is useful for discovering the standard practice in a given shop or the customary practice of a particular pressman; it brings further order into the content of the first table but is not a substitute for that table, which records the precise sequential arrangement of figures on which the generalizations in the summary are based.

I have provided further examples of the use of this system in two articles on American press figures;[69] it is also employed in Donald D. Eddy's A Bibliography of John Brown (1971) and in Vander Meulen's bibliography of Pope's Dunciad. Eddy uses only the sequence tables and attaches a page reference to every figure rather than placing the most common location in the heading; Vander Meulen makes several other modifications, such as designating the position of the figure on the page with greater precision than I had suggested (“A3r-a53” meaning that the figure is 53 mm from the left edge of the type page) and adapting the sequence tables to accommodate impressions in different formats (p. 154). Alterations may be necessary in any system as occasion warrants, but some such approach as this one seems to me to justify the space it requires by facilitating the comparison of copies and the discovery of additional varieties of figuring; it also makes patterns in figures readily detectable and thus may assist the further investigation of the nature and function of figuring.

(g) Variation in furniture width

When standing type or plates are reimposed for a second or later printing, any or all of the formes may be made up with an amount of


Page 52
furniture between the type-pages that is slightly different from the amount that was present in the imposition for the previous printing(s). Thus a measurement of the spaces between type-pages that were adjacent in the forme can be a way of distinguishing printings that do not seem to be otherwise distinguishable. This technique is potentially usable for analyzing books of all periods; but since it was not usually feasible in the early centuries of printing to keep large quantities of standing type, the technique clearly has a wider application for books of the past three centuries, and in fact it has so far been used primarily in connection with nineteenth- and twentieth-century books. Even in instances where variations in furniture width do not furnish the only way of distinguishing impressions, the fact of such variation is a physical characteristic deserving of record in a description.

In 1957, Matthew J. Bruccoli (in Notes on the Cabell Collections at the University of Virginia) measured the inner margins (gutters) of facing pages at the centers of gatherings as a way of recognizing reimposition. In the case of Jurgen, for example, he wrote, “the 2nd impression may be identified by the gutter measurement: at pp. 176-77 it is 31 mm., while in the 1st impression it is 36 mm.” (p. 33). Six years later he used gutter measurements to confirm the identification (based on reset passages) of four 1851 printings of The House of the Seven Gables and to locate a fifth “1851” printing not otherwise recognizable. In this instance he cited measurements from two locations (on pp. 120-121 the difference was 5 mm, and on pp. 216-217 it was 4 mm), and he expressed the view that a difference of at least 3 mm would be sufficient to take into account the possibility of paper shrinkage.[70] Obviously gutter measurements on any two conjugate leaves would be relevant, but the conjugate leaves at the centers of gatherings were selected because it is easier to make relatively accurate measurements at that point. But even there, and with the help of an inserted strip of paper, it is not easy to get accurate readings in tightly bound books (or in any book where the folding of the paper has loosened its fibers and stretched the paper).

To avoid these problems, Peter L. Shillingsburg has suggested a system that eliminates the need for direct measurements of gutters, and in addition it enables one to consider differences in furniture between pages adjacent across what are now top edges and fore-edges as well as gutters.[71] The method is to hold a book so that light passes through a given leaf and to measure the distance (if any) between the outer (or inner) margin of the type-page on the recto and the outer (or inner)


Page 53
margin on the verso, noting whether the verso margin is to the left or right of the recto margin; then one performs the same operation on the conjugate leaf (or on any leaf containing the outer- and inner-forme type-pages that were adjacent in other directions to the recto and verso pages already measured). The figure that results from combining the two measurements (subtracting if both reflect shifts in the same direction or adding if the shifts are in different directions) should remain constant through all copies of an impression-sheet. (The actual register might differ in different copies, but this figure should remain the same.) When this “register-coefficient” (as one might call it) is different in two copies of a sheet, the width of the furniture was different in one or both of the formes from which the copies were printed, and therefore the two copies are probably from two different printings. Obviously the absence of such a difference does not necessarily mean that the copies are from the same printing; and this obvious point is of particular significance here, calling attention to a distinct limitation of the system. Since the measurements are only of relative differences, they would not reveal reimpositions in which the furniture, though indeed altered in width, was changed by the same amount in the outer and inner formes: the relation of the recto and verso type-page images would not reflect the fact that the actual amount of space between the type-pages in the forme had changed.

Each system, therefore, has its advantages and disadvantages. When a book is loosely enough bound that the gutter measurement can be taken with relative confidence in its accuracy, that simple approach may be sufficient; but when direct gutter measurements are not possible or reliable, the other approach can be tried in case the situation is one in which it can be decisive. Indeed, it may be worth using both techniques, since it is possible for a variation to exist in one part of a forme and not in another, and the register-coefficient approach can result in discoveries involving leaves other than the center two in a gathering.

But whether one is recording actual distances across gutters or register-coefficients, it is important to be able to report on an examination of every sheet, not only as an aid in discovering mixed sheets but also to take into account the contrary possibility that the furniture was not identical for every forme of a single printing. When there is no variation throughout a book, a brief statement provides an adequate account: “This printing can be identified by gutter measurements of 3.8 cm at the center of every gathering,” or “... by a register-coefficient of 1 mm on $1.8 (in contrast to 0 in the first printing).” But when there are differences, a table may be more helpful, as in these two hypothetical cases:


Page 54

Gutter   $1.8   Register-coefficient  
A-C  3.6 cm  A-B  1 mm 
3.8 cm  C-F 
E-G  3.7 cm  G-K  1.33 mm 

If, in the second case, measurements on different pairs of leaves in different gatherings are required to locate variations, the leaf citations would have to be made following each signature rather than at the top of the column.

(h) Plating evidence

One of the basic questions to ask about the presswork stage of the production of a nineteenth- or twentieth-century book is whether the printing was done from type or from plates—and if from plates, whether they were relief or offset. The technique of relief plating—creating, from a type-page, a mould for producing a relief image of the type-page surface cast in a solid piece of metal—goes back at least to the early eighteenth century, but it did not become a practical method for commercial use until early in the nineteenth century.[72] Once available, the advantages that plates offered for handling, storing, and shipping meant that they would rapidly become popular. Stereotype plates were increasingly used by American book and magazine printers in the early decades of the nineteenth century, and by mid-century major American publishers ordered plates for nearly all their books and magazines. Electrotyping, another process that produced relief plates, largely replaced stereotyping in the last third of the century. In Britain, the use of plating grew somewhat more slowly, and some publishers had first printings made from type, reserving plates only for later printings. Although the majority of British and American books from the last two-thirds of the nineteenth century can be assumed to have been printed from plates, any given book may not have been so printed, and it is important for descriptive bibliographers to accumulate whatever evidence there is for deciding the matter, both as a fact of printing history and as a piece of information relevant to editors' assessments of textual variants.

Although clues for distinguishing standing type from plates are not always available in every edition, bibliographers dealing with nineteenth


Page 55
century books have developed a repertoire of evidence to search for.[73] These have been written about primarily by Peter L. Shillingsburg and Michael Winship, and from their work[74] one can derive the following list of distinguishing features: (1) Type movement. Evidence of shifting type obviously means that plates were not used; even when there is no gross shifting, there is often slight movement (detectable with the use of a collating machine) both within a single printing and more commonly between printings. Absence of shifting does not of course prove that plates were used; but if lines that contain textual variants show no movement in the invariant parts,[75] and particularly if the revised readings are not aligned precisely, one can conclude the probable use of plates. (2) Lightness or darkness of typeface-images. The general grayness characteristic of impressions from heavily used plates offers too imprecise and subjective a test to be relied on; but when textual alterations were made in a plate, the new letters often make a noticeably darker or lighter impression than the surrounding text. (3) Batter. Damaged letters do not in themselves signify the use of plates, but batter around the edges of pages frequently does come from the handling of plates between printings. (4) Vestigial or multiple signatures. When, as often happens in American nineteenth-century books, the signatures do not coincide with the actual gatherings or when multiple sets of signatures to accommodate alternative formats are present, the printing is almost certainly from plates. (5) Dimensions of type-pages. Since images from plates may be smaller than those from the standing type (because of the shrinkage of moulds as they dry and of plates as they cool), varying type-page measurements in different copies of an edition can signify that some copies were printed from type and others from plates (or some from first- and some from second-generation plates)—so long as the variation is more than can be explained by paper shrinkage (which is usually greater along one dimension than the other).

Descriptive bibliographers should check for all these potential clues and then state a conclusion with whatever degree of certainty the evidence warrants. Some of these categories of evidence, such as the move-


Page 56
ment of types, may be mentioned elsewhere in a presswork paragraph; the presence of signatures that do not match the gatherings might appropriately be taken up in the collation paragraph; and the dimensions of type-pages are a standard element in the typography paragraph. In reporting a conclusion about the use of plates, a bibliographer can simply refer to such evidence noted elsewhere and in addition should selectively cite any other evidence:

Printed from plates, as evidenced by the vestigial signatures reported above and the apparent absence of any shifting type, even in some lines (e.g., 112.7 and 250.16) where the last letter fails to print.

This first impression was probably printed from plates, since in the second impression some of the textual alterations (e.g., those at 105.17 and 195.8) are misaligned and have a dark appearance and since the type-page dimensions are identical in the two printings.

In both these hypothetical examples, the parenthetical citations of selected page-line references allude to fuller lists elsewhere in the descriptions, without suggesting that all the entries in those lists provide relevant evidence (but naturally implying that none of the entries offers conflicting evidence).

Books printed from offset plates, which came into prominent use in the twentieth century, can theoretically be distinguished from books produced by relief plates through the absence of indentations in the paper.[76] But the amount of indentation in some relief-plated books is so slight as to make this test one that it is hard to have complete confidence in. Fortunately there is at least one clue that is sometimes available: “offset slur,” the slight smearing of ink caused by improper contact between the offset blanket and the paper. Not usually visible without magnification, it consists of faint shadows pointing toward the trailing edge of the sheet as it went through the press (though it does not necessarily appear in every copy of an impression-sheet). Craig S. Abbott has explained how the directional nature of offset slur can be employed to work out possible impositions and to identify cancels and unmarked impressions.[77] He uses the words “inner” and “outer” to indicate whether the slur moves toward the inner or outer margin of a page and records the evidence in a table in which leaves and pages are grouped according to the units formed by their slur direction (adapted from p. 540):


Page 57

Leaves   Pages   Direction  
1-8  i-x, 1-6  inner 
9-16  7-22  outer 
17-24  23-38  inner 
25-32  39-54  outer  [and so on

When descriptive bibliographers discover offset slur, they should probably record its occurrences fully in some such form as this, rather than simply reporting that the book was printed by an offset process, in order to make available the evidence underlying any conclusions about imposition and in order to encourage the development of further uses for this technique of analysis.

All these proposals for reporting on typesetting and presswork are intended to be suggestive rather than prescriptive. They are meant only as examples of how one might handle certain production details that are basic to an understanding of the forms of books; but the precise method of treating these details in any given instance must emerge from the particular context. What I hope is that my comments will stimulate thought about the ways in which an account of these and other aspects of typesetting and presswork can be integrated into descriptive bibliographies— and thus will help bridge what is often perceived as a gap between descriptive and analytical bibliography. Descriptive bibliography is inevitably an analytical activity, and the results of bibliographical analysis should be regarded as a natural part of it.