University of Virginia Library

This feeling is reflected in Jones's preface: after noting that the reader will find no essays on collational formulas, he adds, "There need be none" (p. 5).

Yvonne Noble in American Book Collector, n.s., 2, no. 6 (November-December 1981), 47.

Donald Gallup, On Contemporary Bibliography with Particular Reference to Ezra Pound (1970), p. 8.

See, for example, his "On Method in Bibliography," Transactions of the Bibliographical Society, 1 (1892-93), 91-98, to which is appended a sample of his Oxford bibliography (pp. 99-102). The designation of line endings had not been uncommon in the nineteenth century, but it was not generally accompanied by attention to typography. Madan's use of lower case and two sizes of capitals was a direct antecedent of the practice of those who came to have greater influence. A later statement of his approach to transcription appears (p. 56) in Falconer Madan, E. Gordon Duff, and Strickland Gibson, "Standard Descriptions of Printed Books," Oxford Bibliographical Society Proceedings and Papers, 1 (1922-26), 55-64.

For Pollard and Greg, see their "Some Points in Bibliographical Descriptions," Transactions of the Bibliographical Society, 9 (1906-8), 31-52; see also Pollard's practice in the first volume (1908) of the Catalogue of Books Printed in the XVth Century Now in the British Museum and Greg's in A Descriptive Catalogue of the Early Editions of the Works of Shakespeare Preserved in the Library of Eton College (1909).

McKerrow's work was an expansion of his "Notes on Bibliographical Evidence for Literary Students and Editors of English Works of the Sixteenth and Seventeenth Centuries," Transactions of the Bibliographical Society, 12 (1911-13), 213-318—which, however, did not take up the question of transcription.

The third Excursus (pp. cxxxi-cxlviii) of Greg's introduction (1959) to A Bibliography of the English Printed Drama to the Restoration (1939-59) is entitled "Transcription and Quotation"; but aside from the association between the two implied by linking them in this title, no attention is paid to the relationship. Greg is primarily interested in establishing rules for transcription of title pages and for what he calls "bibliographical quotation" (which ignores typography, even italics), to be used in recording the contents of a book. His view that it is "pedantic" to use quasi-facsimile transcription in the contents paragraph has surely now been superseded (see Bowers, p. 291); but when quasi-facsimile is not used, there is a question whether his "bibliographical quotation" is preferable to ordinary quotation.

Madan foresaw this situation in 1893 (see note 4 above) and thought it should mean the elimination of transcription: according to the report of discussion following his paper, "The daily cheapening of photography made him hopeful that, in due time, the chief objection to the use of absolute reproduction in fac-simile would disappear" (p. 105).

Modern Philology, 39 (1941-42), 313-319 (quotations from p. 317); Bowers quotes a long passage from this review in his discussion of quasi-facsimile transcription (p. 135). Gallup (see note 3 above) also seems to believe that cost is the principal factor involved: "if the publisher can be persuaded to pay for easily legible photographic reproductions as illustrations, title-page transcriptions themselves would seem seldom to be essential" (p. 8).

Later in the series, as in A. F. Allison's Robert Greene (1975) and M. R. Perkin's Abraham Cowley (1977), the reproductions are supplied in microfiche in a pocket attached to the inside back cover.

If any argument on grounds of cost were now to be made, it would focus on the cost of paper, not photography. Photographic reproductions take up considerable space and can greatly lengthen a descriptive bibliography; there is little point providing illustrations of typographic matter unless they are the same size as the original. Even the editors of the Pittsburgh Series have decided to forego reproductions of title pages for certain categories of material (such as books contributed to by an author).

The bibliographer should know when such situations exist and annotate the transcriptions accordingly; but a bibliography should enable readers to identify variants unknown to the bibliographer.

In 1939 J. D. Cowley went so far as to say that, with exceptions for "special purposes," "we may now abandon quasi-facsimile, and adopt the rule that roman lower case, with initial capitals when necessary, is to be used for transcription" (Bibliographical Description and Cataloguing, p. 56).

This point should be underscored, as it is in Bowers's "Purposes of Descriptive Bibliography, with Some Remarks on Methods," Library, 5th ser., 8 (1953), 1-22 (reprinted in his Essays in Bibliography, Text, and Editing [1975], pp. 111-134): "Please notice that what is in question is the whole book, and not just the title-page as the sole source of identification. . . . If we place such confidence in our ability to identify books by their title-pages that we fail to describe other matters in the more important body of the book, we are misleading a reader by providing him with insufficient information to identify an imperfect copy lacking the title, or by recording in a bibliography the presence of such and such an edition in a library when in fact it is a dangerous made-up bastard copy. If we are sincere in desiring to record the true details by which to identify books, let us deliberately describe books as if they had no title-leaves . . . there is more to description than identification, and more to identification than the title-page" (see esp. pp. 9-15 [119-126]).

According to Foxon, "After Capell's death, quasi-facsimile seems to have vanished from English scholarship for a century" (p. 13)—and Madan could therefore "put forward his system as novel" (p. 15). Foxon does, however, allude to Henry Harrisse's Bibliotheca Americana Vetustissima of 1866; and other instances could be cited. The question is whether these occurrences are isolated innovations or whether they reflect a continuous tradition.

Title pages with illustrations are another matter: they are usually more expensive to reproduce well.

They are often less desirable anyway, especially when the types involved are large and heavy, for areas of solid black do not reproduce well by xerography. (In making adjustments, one should note that reproduction ratios do not always remain constant on a single machine.)

For example, see some of the small reproductions (as on pp. 96-97) in R. A. Sayce and David Maskell's A Descriptive Bibliography of Montaigne's Essais, 1580-1700 (1983), a bibliography in which reproductions supplant transcriptions.

Any spot that is touched up, to conceal a perceived blemish, is itself a blemish. The problems produced by retouching (as well as the peculiarities of individual copies) are mentioned by James Hardin in his able restatement of the importance of transcription, in "Descriptive Bibliography and the Works of Johann Beer," Wolfenbütteler Barock-Nachrichten, 4 (1977), 2-6.

Bowers, in his discussion in the Principles pointing out some of the advantages quasi-facsimile transcription has over photographic reproduction (pp. 135-137), mentions poor inking, flyspecks, and imperfections in paper as some of the characteristics of originals that could be misleading in photographs, and he warns readers of the dangers of retouching. Philip Gaskell, objecting to this argument, wrote a letter published as "Photographic Reproduction versus Quasi-Facsimile Transcription," Library, 5th ser., 7 (1952), 135-137. Gaskell says that it is the responsibility of the bibliographer to find copies for reproduction that are without blemish (and he maintains, incidentally, that "reproductions can be greatly reduced without losing their value"). One must naturally agree that the bibliographer has the responsibility to find unblemished copies, when they exist and are locatable; but Gaskell's discussion shows no awareness of the problems that still exist after one has obtained "a good photographic reproduction." Bowers replied to Gaskell in "Purposes of Descriptive Bibliography, with Some Remarks on Methods" (see note 14 above), arguing that photographic reproductions are "no such cure-alls that we can set up their utilization as a necessary or even as a generally desirable principle." He examines "the curious fallacy that the title-page is the most important single feature of identification that exists for a book," shows that photofacsimiles of titles provide some help for identification only when a reader has a copy that "differs from any that the bibliographer has seen and reproduced," and concludes—after describing the scholarly standards for descriptive bibliography—that "no mechanical process can act as a substitute for such standards." In A New Introduction to Bibliography (1972) Gaskell sets forth the established rules for transcription but also asserts that "there is often a case for dispensing with the laborious process of quasi-facsimile for any but short extracts from title-pages, etc." (p. 323).

Some bibliographies print what appear to be defective reproductions without explanation. For example, Joel Myerson's Emily Dickinson (1984)—which, as one of the Pittsburgh Series, uses title-page and copyright-page reproductions in place of quasi-facsimile transcription—contains a reproduction of a title page in which (in the copies I have seen, at least) the date is not legible (see p. 38); the copyright page in the facing illustration is difficult to read because of defective inking, and one is not told whether this defect is in the original or only the reproduction.

The function of facsimiles and standards for their production and annotation are sensibly discussed by Franklin B. Williams, Jr., in "Photo-Facsimiles of STC Books: A Cautionary Check List," Studies in Bibliography, 21 (1968), 109-130.

Illustrations have a different status, of course, in a bibliographical catalogue of a collection, where each entry does describe a single copy. But even there the distinction between a scholarly verbal account and a depiction still holds. Philip J. Weimerskirch, in "The Use of Title-Page Photography in Cataloging," in Library Resources and Technical Services, 12 (1968), 37-46, surveys the history of the idea of using photographs of title pages and describes instances of library cataloguing (or temporary cataloguing) in which such photographs have been employed; it is not his purpose, however, to pursue the distinction between photographs and verbal statements, or between photographs of single copies and generalized descriptions.

The problem of producing a facsimile of an entire book is analogous: if one wishes to make available to scholars a "facsimile of the first edition," how does one go about selecting a copy to photograph? Charlton Hinman, in preparing his Norton facsimile (1968) of the Shakespeare First Folio, recognized that no single copy would serve the purpose and selected pages from many different copies so as to represent the "corrected" text at points of stop-press alteration. The individual pages are of course those of particular copies, but the resulting facsimile as a whole is not identical with any surviving copy. In attempting to represent "the First Folio"—that is, a whole edition—it could not be limited to the characteristics of any individual copy. See also Fredson Bowers, "The Problem of the Variant Forme in a Facsimile Edition," Library, 5th ser., 7 (1952), 262-272.

See Bowers, Principles, pp. 113-123; and G. T. Tanselle, "The Concept of Ideal Copy," SB, 33 (1980), 18-53.

David Vander Meulen, in one of the most judicious assessments of the strengths and weaknesses of quasi-facsimile transcription and photographic reproduction, makes the same point: quasi-facsimile transcription, he says, "by reducing a physical object to words as the rest of a bibliographical description does, provides a standard form of the title page that can be more easily re-used and transmitted. Moreover, by presenting an ideal copy rather than a specific one, a quasi-facsimile can better convey the bibliographer's interpretation of ambiguous portions." See A Descriptive Bibliography of Alexander Pope's Dunciad, 1728-1751 (diss., University of Wisconsin, 1981), p. 44. Vander Meulen provides several telling examples of the way in which photographs (as well as transcription that falls short of quasi-facsimile) can be misleading. (His own practice in transcription, by the way, includes an indication of the measurement, to the nearest ⅓ millimeter, of the type faces employed.)

For one view of this shift, see G. T. Tanselle, "Physical Bibliography in the Twentieth Century," in Books, Manuscripts, and the History of Medicine: Essays on the Fiftieth Anniversary of the Osler Library, ed. Philip M. Teigen (1982), pp. 55-79. See also note 34 below.

This view of title-page quoting as part of a historical account reinforces Bowers's point that the evidence from a title page is only one element in the total body of evidence supplied by the book as a whole (see note 14 above).

Even if what is quoted is a variant known in only a single copy, for the point of the quotation is not to represent this copy but to report that some portion (however small) of the edition or impression in fact contained the variant in question.

Bowers uses single, rather than double, quotation marks to distinguish quasi-facsimile transcription; the device is less effective for British readers, however, because of the more common use of single quotation marks in Britain for ordinary quotation.

For further discussion, see G. T. Tanselle, "Tolerances in Bibliographical Description," Library, 5th ser., 23 (1968), 1-12. Foxon says that if reproduction is rejected for transcription, "we ought to aim at scientific precision and record the size of the type in each line and the amount of space between one line and the next" (p. 18)—a statement that implicitly recognizes various levels of quotation but regards only one as defensible.

The practice of reproducing passages of text photographically is not new; what seems to be increasing is a tendency to think of such reproductions not as illustrations but as part of the text itself. The former approach is suggested in Hinman's The Printing and Proof-Reading of the First Folio of Shakespeare (1963), for example, by the fact that the many reproductions of passages from the First Folio set into pages of Hinman's text are labeled as "figures," sequentially numbered and with captions. In contrast, some of the essays of Randall McLeod—to cite one scholar who has recently employed photoquotation extensively—incorporate reproduced passages, or even phrases, directly into the text in the way that reset quotations have always been used. There is no question about the appropriateness of the reproductions, for the typography of the original printing is in each case part of the subject under discussion. And one could say that it is only a stylistic matter whether one constructs sentences so that they can stand independently of illustrative quotation or whether one makes the quotations an integral part of the sentences. But a significant distinction is lost sight of if one believes that a reproduction—regardless of its importance for the discussion—can fit into a new text in the same way that a transcription can. The point is dramatized by the title of McLeod's essay in The Division of the Kingdoms: Shakespeare's Two Versions of King Lear, ed. Gary Taylor and Michael Warren (1983), pp. 153-193. His title (which may be rendered here as "Gon. No more, the text is foolish.") is photoquoted at the head of his essay but is converted to the standard typography of the volume in the table of contents and in the running heads. If McLeod photoquotes the title because he believes the original typography to be essential to the meaning, he has at the same time guaranteed that the title will generally be misquoted, because references to his essay in checklists, footnotes, and other essays will not normally accommodate the reproduced typography. (Of course, resetting the title would not be misquoting, according to the usual conventions of quotation, if the presence of the reproduction did not suggest that the typography is an inseparable element of the title.) The fact that photoquotation, whatever its extent, is not appropriate for some contexts may seem only a practical difficulty, but it clearly reflects a deeper issue.

If it does, the problem posed by the variation is itself likely to enter into the discussion, and specific copies in which the different readings can be found should of course be specified.

The role of the incunabulists in establishing modern descriptive bibliography is touched on in G. T. Tanselle, "The Evolving Role of Bibliography, 1884-1984," forthcoming in the proceedings of the Grolier Club Centennial Convocation.

See "Michael Trevanion," "Thomas J. Wise's Descriptive Formula," Book Collector, 13 (1964), 355-356; and Simon Nowell-Smith, "T. J. Wise as Bibliographer," Library, 5th ser., 24 (1969), 129-141 (esp. p. 132).

In Madan, Duff, and Gibson, "Standard Descriptions of Printed Books" (see note 4 above). Gibson recommended that the collation of modern books be treated like that of books of the "middle period" (1558-1800), for which Madan specified signature collation in the case of "important" and "interesting" books; for "ordinary" books he felt that signatures need not be noted except "in case of any complication" (p. 58).

Henry Bradshaw's Correspondence on Incunabula with J. W. Holtrop and M. F. A. G. Campbell, ed. Wytze and Lotte Hellinga (1966-78), p. 30 (other examples on pp. 31, 43-44). Bradshaw's "long" use of this system went back more than a year, for on 18 February 1863 he had explained it to Winter Jones: see "Letters of Henry Bradshaw to Officials of the British Museum" (ed. A. W. Pollard), Library, 2nd ser., 5 (1904), 266-292, 431-442 (esp. pp. 277-278; it is also illustrated in a letter of 8 January 1866 on pp. 288-292).

In the long subsequent history of this convention, there has been repeated debate as to whether these superscript figures can ever be odd; but there scarcely can be any doubt today that the formula serves best its function of showing a book's structure if such figures refer only to the leaves (necessarily even in number) that make up (or originally made up) the folded quires, with insertions and deletions to be handled separately. Greg sums up this debate in "A Formulary of Collation" (see note 40 below), p. 372 (309); and Bowers discusses it thoroughly, pp. 225-235. Gaskell (New Introduction, p. 329) and Bowers (p. 434, speaking of nineteenth- and twentieth-century books only) would make an exception for 18° in 9s, since the format results in nine-leaf gatherings; but even here, one leaf must be disjunct, and writing "A9" does not show where that leaf is located. It would be preferable, I think, to write, for example, "A-M8($4 + 1)." There have also been suggestions for incorporating further information into the superscript figures. One of Pollard's is mentioned below in note 45; another is the use of formulas like "A-G8/4" or "A-G8.4" to show a regular alternation between eight-leaf and four-leaf gatherings (a device still often employed). Alistair Elliot has more recently built on this device—in "Duplicated Signatures," Library, 5th ser., 25 (1970), 354—to suggest superscript figures separated by commas ("A-K2,2") as a shorthand way of indicating regularly paired signatures in successive gatherings, as when A2 is followed by *A2, B2 by *B2, and so on. The difficulty with this practice is that, unlike Pollard's, the superscripts are here made to refer to a peculiarity of signing, not of gathering; yet information about gathering is the function of the superscripts, and a further note would be required to explain the situation. (A clearer device might conceivably be an ellipsis ["A2 * A2 . . . M2 * M2"], which would distinguish this situation from that where a dash is conventionally used.)

Bradshaw himself moved to this system (which he was already using for references to format that were separate from signature formulas): see Francis Jenkinson's note on p. 278 of Pollard's edition (note 37 above).

As in Pollard's "The Objects and Methods of Bibliographical Collations and Descriptions," Library, 2nd ser., 8 (1907), 193-217, and the British Museum catalogue of incunabula (see note 5 above); McKerrow's An Introduction to Bibliography for Literary Students (1927), pp. 155-161; and Greg's "A Formulary of Collation," Library, 4th ser., 14 (1933-34), 365-382 (reprinted in his Collected Papers, ed. J. C. Maxwell [1966], pp. 298-313), and the "Introduction" (drafted in 1942) to A Bibliography of the English Printed Drama to the Restoration (see note 7 above), esp. Excursus IV, "Formulas of Collation," pp. cxlviii-clviii.

Elementary digests of the system also conveniently appear in Gaskell's New Introduction, pp. 328-332, and in Terry Belanger's "Descriptive Bibliography," in Book Collecting: A Modern Guide, ed. Jean Peters (1977), pp. 107-113.

Bradshaw, describing his system of signature notation in 1863, said, "It looks rather appalling at first, but it is so much the most convenient way of collating books—as you can see directly where the defect is in your copy" (Pollard's edition, cited above in note 37, p. 277).

In 1949, the same year when Bowers's book appeared, James G. McManaway stated the need for standardization of the formulary of collation, in Curt F. Bühler et al., Standards of Bibliographical Description, pp. 73-74. He was overstating the case when he said that "many variants are now in use, with the result that the printed collation of a book may be intelligible only to the man who wrote it"; but there is no doubt that Bowers's book has played a crucial role in bringing about widespread acceptance of a single standard.

He takes for granted the practice of referring to leaves by attaching regular (non-superscript) figures to signatures. (He is naturally not concerned with later books—generally American—in which the signatures are numbers; in such cases leaf numbers must be subscript figures to avoid confusion. Greg notes [Bibliography, p. cxlviii] that some bibliographers find subscript leaf numbers, even with letter signatures, "more distinctive or more elegant"; but one must agree with him when he says, "I see no advantage in the practice"—though one could mention the insignificant advantage that leaf numbers would be treated the same way whether signatures were letters or numbers. See also "A Formulary of Collation" [note 40 above], p. 368 [300], where Greg says that using subscripts "suggests something recondite and obscures the plain meaning and origin of the notation.")

Pollard had earlier, in the British Museum catalogue of incunabula (see note 5 above), used plus signs attached to superscript fugures—as in "a-m10 n10+1" (p. 59)—but this approach does not specify the location of the insertion.

As the following formula suggests, an irregularity in a gathering necessitates separating the gathering from the regular sequence; in Bowers's formulary, such a gathering "must not be linked with another gathering by a dash. A-D4(D4 + D5) is completely incorrect" (p. 237). Some writers have since questioned this point. Stevenson sees no ambiguity in such linking and uses it in the Hunt catalogue (e.g., entries 409, 417, 548). The objection to it is one of logic, for the parenthesis applies only to the last-named gathering, not to the whole sequence. Margadant (p. 23) carries Stevenson's point a step further and gathers all irregularities together in a single parenthesis at the end of a sequence of gatherings, as in "A-K4(±D2,—E4,—K4)." If the purpose of the formula is to show the structure of a book, sequence is all-important; and I do not believe one can claim that, for this purpose, such an arrangement "can be read more easily."

Although even here there is an alternative: the replacement in E, if unsigned, could be rendered "χ2" (see Bowers, p. 247). (But rendering it as "E3.4" does not necessarily mean that it is signed. See note 49 below.)

This use of quotation marks in effect provides an equivalent for "[sic]." Emphasizing that quotations are accurate, however, may seem a superfluous task for the formula.

This situation in regard to "E4(—E3,4 + E3.4)" is explained by Bowers as follows: "when there is no possible conflict between an inferred signing of a complete cancellans and the signing or completed signing of the original quire, the inference can be freely made in the formula without the necessity to indicate (except in the note on the signing) whether the cancellans signing was inferential in whole or in part" (p. 247).

Both Stevenson (p. clix) and Margadant (p. 23) express a preference for "χ"; but Margadant unwisely prefers to designate a single leaf with a superscript "1" (p. 22).

Such notation is purely positional: Bowers says, "The leaf of the regular gathering preceding the plus sign is used only to establish the position of the succeeding insertion and is never quoted, nor is any indication ever given whether it is signed or unsigned [or missigned]" (p. 237).

The conservative treatment advocated by Bowers (pp. 235-239, 241-242) for leaves between gatherings is certainly to be recommended: they should be regarded as independent members of the sequence of gatherings and not associated with the gathering before or after except when the link is definitely established, as with the same signature.

This system of noting the total number of inserted leaves is used only when the leaves are disjunct. (See Bowers, p. 240, where he warns against confusing such totals with quoted leaf numbers.)

If it were signed as a later leaf reference, such as "*F2", the superscript could not be used, for one cannot have such a notation as "*F22", which seems to combine reference to a leaf and to a gathering.

Failing to be consistent might create ambiguity by implying that the two usages had different signification.

This doubt would be eliminated if the reader remembered, and had reason to think the bibliographer followed, Bowers's rule that signatures of inserts "are ordinarily inferred only when an unsigned leaf follows a signed leaf which is itself an insert and when the signature does not conflict with or duplicate the signatures or completed signatures of the original gathering" (p. 459).

The fact that "* E5" is an actual signature should be evident even without quotation marks.

Writing "* Q1.2" and "s1.2" instead would make the signing of the second leaves no clearer. On the analogy of "e3.1" above, presumably one could write "* Q1.1", with the same attendant problems.

Bowers's discussion at various points makes this clear. On p. 246, for example, he speaks of the "essentially positional nature of formulary construction": in "C4(C3 + 'C4')", he points out, "the main feature of the construction is the indication of position"; thus the "identification by printer's signature is not essential but only a convenience for helping the reader to understand that the signed insert must not be mistaken for true [C4]." The same point can be made about inferential signing, which "is not of crucial importance in the collational formula" and which is usually adopted only if it provides "superior convenience for reference." He concludes, "The general trend of formulary notation towards reliance on positional notation makes for conservatism of inference." One might add that this trend carried to its logical conclusion would result in no inference at all (and no quotations of actual leaf signatures either). (See also p. 495.)

Square brackets and italics (suggested by Greg in "A Formulary of Collation" [see note 40 above], p. 369 [306]) are equivalent ways of designating inferred signatures. (See also note 66 below.)

The silent correction of missignings in reference to regular leaves in collational formulas is discussed by Bowers on pp. 222-225.

Except that, in Bowers's view, "it would seem to be unnecessary in [inserted] folds to indicate signing of later leaves, . . . although no positive objection could ordinarily be made" (p. 242).

Greg, at the beginning of "A Formulary of Collation" (note 40 above), comments on the printer's "register" in early books, which listed the signatures and the units in each, and he suggests that "there might be a good deal to be said for applying the term to our formulas of collation." In his Bibliography he does often refer to the formula as a "register," but he is aware that his practice is sometimes at odds with what the term implies: "Neither in reference nor in the register should the signatures given be regarded as in any strict sense quotations from the original, though it is true that the registrum may be said to have consisted of such in the first instance" (p. clii). Many of the complications that he analyzes in fact arise from a consideration of whether the bibliographer is best served by actual quotation and from attempts to accommodate such quotation—as when he writes "χC('Cc')", which perfectly epitomizes the urge to rewrite signatures for bibliographical use and the simultaneous reluctance to eliminate the actual form of a signature from the formula (p. cliv). The history of the development of the collational formula is dominated by discussions that have at their root either an indecisiveness as to the primary bibliographical function of the formula or a curious hesitation to push the formula fully and logically in the direction of a structural statement. The modern formula can no doubt be legitimately regarded as a descendant of the printer's register; but to do so seems to constrict thinking about the formula. (Even the standard term "signature collation" illustrates the point, for the bibliographer collates gatherings, and only incidentally their signatures.)

However, a distinction can be made between a single leaf and a fold, and there are times when it may not seem inappropriate to assign "χ" to a fold. See note 71 below.

Bowers glances at these possibilities (p. 238) and says, "No real objection can be raised to this theory, perhaps, except to point out that the method cannot be consistent itself, since in certain circumstances quoting proves impracticable, as for the shorthand notation of simultaneous cancellation and substitution" (see also pp. 244, 248). One can of course use an expanded treatment of cancellation and substitution instead of the plus-minus sign; but the undeniable problem is that it would be not merely impracticable but in fact impossible to quote the signature of the cancelled leaf if no copy of it had been located. (One convention might be to place the plus-minus sign after the quoted signature of the substitution—as "C4('C3'±)"—to show that the named leaf replaces whatever leaf was previously in that position, without directly citing that previous leaf.) Bowers proceeds, "Moreover, since the ordinary formula takes no notice of the signing of individual leaves in a gathering so long as the signature of the gathering is established as noninferential, it need not necessarily do so for the signing of inserted leaves where the position is in no doubt. Quotes, therefore, may be reserved the more simply for the occasions when it is definitely desirable to indicate signings." One may certainly declare that the signatures of inserted leaves are not to be quoted, but the discussion at this point is how to handle them if they are quoted, and some form of consistent quoting is recommended: "inferentially signed [inserted] leaves must ordinarily be differentiated from signed inserted leaves" (p. 237). If, on the other hand, one decides that the signatures of inserted leaves are not to be noted, then by definition there are no situations in which it is desirable to note them.

In these formulas, the notation for the second inserted leaf in C would more conventionally be preceded by a plus sign (a comma would be used if the second leaf were cited as "χ1"). But the comma seems preferable because it recalls the parallelism with the convention of the period for conjugate leaves and because it allows the plus sign to be reserved exclusively to locate the link between the regular leaf and the inserted leaf or leaves. One should also note here the distinction, emphasized by Bowers (e.g., pp. 238, 257), between italics and brackets. Italic type denotes an inferred signature for a whole gathering (as in A here) and cannot be used to refer to an unsigned leaf of a gathering signed on some other leaf (as in C), whereas brackets can serve both functions. Italic leaf references are permissible, therefore, only when the signature involved has been inferred for the gathering as a whole (as in A).

There is of course a third possibility: using quotation marks and brackets in the same formula. The result would be still clearer and considerably more awkward.

Of course, there is a fundamental difference between references to leaves and those to whole gatherings; but the fact remains that to complicate the parenthetical references with indication of the signing of leaves suggests that such indication is an appropriate function of the formula—a function not then carried through with regard to the other leaves, those not associated with insertions. Before passing on to an approach that in my opinion is more attractive, one further point is worth noting. A system that aims to be uniform and explicit in its references to leaves should also use numbers in a consistent way to refer to leaves; and since numbers regularly specify individual leaves, they should probably not be used at the same time for totals. If, in C above, both inserted leaves had been unsigned, the conventional notation "(C3 + 2)" would seem less appropriate in the context than "(C3 + 1,2)"—or, in the second formula, "([C3] + [1],[2])"—in which the "1" and "2" denote the first and second of the inserted leaves (just as "C1" and "C2" denote the first and second regular leaves of the gathering). Whether this reasoning suggests that in the formulas above the second inserted leaf in C should be called "2" rather than "1" (or whether "1" can be thought of as suggesting the first unsigned inserted leaf in the gathering) need not detain us, since the approach illustrated by those formulas cannot be strongly recommended anyway. This question calls attention again to the problems associated with quoting in formulas, illustrated here also by the possibility of inferring a second leaf in some instances (as in B) but not in others (as in C, where an inferred "C4" would conflict in notation with the regular fourth leaf of the gathering). Such complications, though not serious perhaps, provide further reason for thinking that this approach does not achieve the desired combination of clarity with simplicity.

Bowers alludes briefly (pp. 240-241) to a system that goes part way in this direction: eliminating from the formula the signing of inserted leaves except when "the signing is related to the signing of the gathering"—thus the "1" in "K4(K3 + 1)" would be appropriate if the inserted leaf were signed with an asterisk or a paragraph sign but not if its signature included "K", which would necessitate, for example, "K4(K3 + 'K4')." He understandably finds this system inconsistent and sees "no really positive virtue in it."

There is a case for retaining the signatures: their presence would emphasize the distinction between regular leaves (or replacements for them), which would be noted by signature and number, and inserted leaves, which would be noted by number alone. Signatures would be especially welcome when the conjugacy of canceled leaves is not the same as that of the substitutions for them: if "G4(—G3,4 + G3.4)" were written without the signatures—as "G4(—3,4 + 3.4)"—the numbers following the plus sign might seem at odds with the "1.2" one would normally expect in this system; but the addition of "G" would made instantly clear that the leaves are replacements, rather than insertions that increase the total number of leaves in the gathering. Greg, in "A Formulary of Collation" (see note 40 above), rejects the idea of eliminating signature letters within parentheses because "it would lead to ambiguity in some of the more complicated cases"—though the case he cites could be handled by replacing a comma with a semicolon (pp. 379, 381 [310, 311]).

An argument could be made that in this system conjugate inserted leaves should be represented by the other standard notation for folds, the superscript "2", which would in turn necessitate the use of a signature (actual or assigned). Even though an inserted fold takes its identity from the gathering into which it is inserted, an inserted fold is analogous to a separate gathering in that a signature can apply to both its leaves without conveying any information about the specific signing of either leaf. The system would remain consistent, therefore, if one wrote, for instance, "B4(B2 + b2)" instead of "B4(B2 + 1.2)", because no information about the signing of individual leaves is being reported in either case; the difference turns solely on the question of whether the reporting of inserted folds is to follow the model for gatherings or that for inserted leaves, and a defense of either approach can be made. The whole issue might not seem worth bringing up if all insertions were as simple as the ones in this example. But when two or more successive folds, or even quired folds, are inserted at the same point, a far simpler formula would result if one regarded those insertions as in effect a subseries of gatherings, treating them as separate gatherings would be treated: e.g., "A4(A2 + a2 * B4 b2 2b4) B-E4."

An excellent demonstration of this point is provided by Bowers's discussion of the application of the formulary to incunabula (pp. 487-499).

There has been a tradition of omitting the "r" when the context makes clear that the recto page is meant rather than the whole leaf. But the practice of allowing the same reference to mean one side of the leaf and both sides of the leaf is illogical and potentially confusing; it obviously has no place in precise notation, as Foxon argues with some vehemence (pp. 8-9). There is also a tradition, especially among incunabulists, of using superscript "a" and "b" for the two sides of a leaf; but there is no justification for regarding this usage as the only defensible one, as Gaskell does on pp. 329-330 of the New Introduction.

Another possibility may be mentioned, building on Greg's suggestion for citing gatherings whose location is not obvious. If one wished, for example, to cite χ3 in "A-G4 χ4 H4", Greg would write "Gχ3", using the signature of the preceding gathering as a superscript index for the purpose of revealing immediately where the unsigned gathering is located. Similarly, one might indicate the location of an inserted leaf by using the notation for the preceding regular leaf as a superscript index—e.g., "B2†" or "E3e3". (Stevenson, in the Hunt catalogue, p. clix, suggests "Rχ" for an inserted leaf in R, but this notation does not show the exact location in R.)

In some respects, of course, the notation does indicate structure by distinguishing an inserted leaf from a regular leaf, and if one wished to show the conjugacy of insertions one could use periods, following the practice in the formula: "B2(1.)" and "B2(.2)" would refer to inserted leaves that belong to a fold, whereas "C3(1)" and "C3(2)", lacking periods, would refer to disjunct inserted leaves.

Bowers gives as an example the second inserted leaf in K4(K3 + 'K4',χ1), noting that a reference to it as "K('K4' + 1)" would require "prior analysis of the precise position of leaf 'K4'", whereas "K(K3 + 2)" states its position in relation to the regular leaf K3. In the standard formulary, however, "+2" would refer to two inserted disjunct leaves, not to a single leaf in a second position. Bowers comments on this difference in usage: "This method differs from the manner in which these leaves would be noted in a collational formula; but there is no anomaly since only positional reference is being made in the absence of a formula to which formulary reference could be constructed" (pp. 260-261). Nevertheless, it would be advantageous from a practical point of view if the practice in reference notation coincided with that in the collational formula and if this divergence in usage could be avoided.

Bowers points out (p. 269) that some bibliographers regard only irregularities as worth noting, allowing customary practices of signing for the period to be taken for granted. But it requires little additional space to be explicit, and that space would seem to be well used.

Except, of course, regular leaves in those gatherings having no signature at all, because the inferred signature in the formula constitutes a statement that no regular leaf in the gathering is signed. One might argue that, if the formula deals only with structure and not signing, there is no point italicizing an inferred signature. (Even then, assigned signatures such as "π" and "χ" would still indicate lack of signing.) But as long as one bases the formula on the system of signing present in the book (and it would be foolish to do otherwise), it seems reasonable to show which signatures employed in the formula actually appear in the book and which do not. The formula, in this view of it, would not be concerned with the signing of particular leaves but would incidentally, in the course of reporting the structure of a book, reveal which gatherings are entirely unsigned; the statement of signing would then be exclusively devoted to the signing of leaves (as opposed to gatherings as a whole).

Bowers explains that the explicit approach is essential for dealing with nineteenth-and twentieth-century books (p. 434); some possible confusion might be eliminated, therefore, by employing it for all books. Margadant also takes this position, pointing out (p. 25) the awkwardness in saying both "$3 signed" (meaning the first three leaves) and "$3 signed $ *" (meaning the third leaf). He goes on, however, to propose what seems to me an unnecessary elaboration: the Greek letter gamma to stand for the (variable) total number of leaves in a gathering, so that one can say "$½ γ signed", meaning that half the total number of leaves in each gathering is signed, regardless of how the bulk of the gatherings varies. When a book has, for example, a regular alternation of eight-leaf and four-leaf gatherings, I see no objection to the slightly longer statement "$1-4 in $8 and $1-2 in $4 signed."

The use of the equals sign in the formula to show the identity of a cancellandum at one point with an insertion or cancellans at another (see Bowers, pp. 250-251) would not, I think, conflict with this use of the equals sign in the statement of signing. (Some bibliographers prefer to use square brackets in the formula to enclose a statement involving the equals sign.) However, if one wishes to avoid the equals sign in the statement of signing, the word "as" (scarcely longer than the equals sign) could be used from the second instance onward, as in "C2 signed as C3, D3 as D, E2 as e2" and so on.

Bowers notes that "signing statements record only variations from normal" and that an unsigned title page, if it falls on a normally signed leaf in a gathering, need not be specifically exempted from the generalization about signing since "few title-pages are signed" (p. 270). However, the small amount of space required to say, for example, "—A1" is probably worth using in order to eliminate any possible doubt in a reader's mind (as Gaskell suggests in the New Introduction, p. 332).

It can be important, as Bowers explains (p. 271), to record details about the typography of signatures (size and style of type) and their placement (in relation to footnotes, for instance); and he suggests that this information, in a full description, might go in a separate paragraph on signatures. One could also justify placing in the statement of signing only deviations (such as italic signatures when most are roman), leaving the principal facts about the typography of signatures for the paragraph on typography, where information about other aspects of typography in the book is drawn together.

One should remember that books with the number of irregularities and insertions recorded in the hypothetical formula above are not common. Some real books are indeed quite complicated, but many are quite regular, and one can expect collational formulas generally to have a much simpler appearance than the one shown here.