University of Virginia Library

III

The more substantial matters discussed in the preceding section — workmen's output, edition sizes, and the relationship between composition and presswork under conditions of concurrent production — must now serve as a prelude to notes on a number of bibliographical procedures that imply quite different productive conditions. If the evidence of part II withstands challenge, it must I think be held to falsify several current hypotheses. It is not easy to summarize these but the ones I have in mind relate particularly to compositors' measures, cast-off copy, skeleton formes, proof-correction and press figures.

If one assumes that a compositor usually worked on only one book at a time he would have had no need to alter the measure to which he had set his composing stick. Changes in line measurement within any one book might therefore be taken to indicate an abnormal interruption, after which the stick was reset to a slightly different measure, or

the presence of a second compositor.[33] Professor Bowers cites, for example, Bellon's The Mock-Duellist (1675) in which sheets B-F are set with a 120 mm. measure and sheets G-I with a measure of 121 mm. "In such a book the inference is probably that with sheet G another compositor . . . took over the work. In general, one is likely to conjecture that any interruption of the printing sufficient to cause a single compositor to adjust his stick again after working on some other book would most likely have been sufficient to cause the skeleton-formes to be broken up . . ." ("Bibliographical Evidence," p. 157). These conclusions must, however, seem misplaced if one begins from a different premise. If we assume concurrent production, for example, then the likelihood of measures reflecting the division of work among compositors will be small. In the first place, production times were too long; and in the second place, compositors working on several books at a time, often in quite different sizes of type, would have had to change their measure constantly. As it happens, these propositions are consistent with the Cambridge evidence, whereas analysis of a few of the Cambridge books suggests that it would be impossible to judge how compositorial work was divided on any of them from the evidence of measures. Not only do the widths of type-pages set by the same compositor vary, but different compositors are often found setting to an identical measure, and interruptions are routine. The general practice inferred from limited physical evidence and the underlying assumption about work method remain mutually consistent, but in most cases they are likely to be quite wrong.

One of the more delicate exercises in advanced analytical bibliography is tracing the pattern of skeleton formes as evidenced by running titles in order to determine the order of presswork and, it might be claimed, the number of presses used. This pattern may be related by a time scale to another showing compositorial stints, or it may of itself be taken to imply a certain number of compositors at work on the book. The relationship indicated between composition and presswork may then be employed as an analytical tool in determining such things

as nature of copy, methods of setting, edition size and proofing procedures. Implicit in all such analyses is the fundamental assumption that composition and presswork on a particular book would normally seek a condition of balance. This is a difficult area in which to order the work done while being fair to the arguments of those who have used such evidence; nor am I confident that I fully understand the analytical principles used. But the subject is important and even at the risk of misrepresentation demands discussion.

The pioneer study in the use of headlines, as in much else, was written by Professor Bowers over thirty years ago.[34] The association of sets of headlines with skeleton formes is now so well evidenced that it may be taken for granted, and, as Professor Bowers has also remarked, "the basic principles of the printer's use of headlines did not differ markedly in any period when books were printed by hand."[35] Where a single skeleton was used for both formes of a sheet,

The phrase 'the press was idle' is perhaps misleading since under conditions of concurrent printing the press would not be 'idle' at all but employed on another book. It has however had considerable repercussions and a great many bibliographical arguments have been constructed on the assumption that this inferred idleness could not have been the norm and must have been avoided in order to secure a balanced relationship of composition and presswork. So Professor Turner: "In order to effect the minimum press delay, the formes . . . would have had to go through the press in the following order . . .".[36] ". . . in one-skeleton work the press was forced to stand idle . . .".[37]

"There is every reason to believe that press delays were abhorrent to the 17th-century printer" ("The Maid's Tragedy," p. 217). ". . . a compositor would not change from setting by formes to seriatim setting without risking a press delay unless he was ahead of his press . .."[38] ". . . the adoption of one-skeleton printing for several formes, and the resultant press delays . . ." ("Philaster," p. 28). "If we assume that two skeletons would have been employed in the most efficient manner . . .".[39] ". . . working on the assumption that composition and presswork could stay more-or-less in balance . . ."[40] ". . . two formes . . . would not have been machined concurrently, for had they been, a delay in presswork would have resulted."[41]

Professor Hinman, however, extended the argument by pointing out that if a book were printed in a very small edition, printing would be so well ahead of setting that a second skeleton would be of little use.

Professor Hinman then suggests that if the edition were a very large one, however, the reverse might be true. Printing would take longer than setting and the pressmen might well seek to avoid delays and restore the balance by using two skeletons. Hence Professor Bowers has subsequently stated that "certain assumptions can be made about the rate of compositorial to press speed and thus about the number of copies printed";[42] and Professor Williams has remarked that "in a small edition press time would be briefer than composition time and the compositor would always be concerned lest he fall behind and so delay his press."[43]

Yet another application was indicated by Professor Hinman when he noted that skeleton formes "have an intimate connection with various possible methods of stop-press correction" ("New Uses for Headlines," p. 222). Applying this principle in a re-examination of the proofing of Lear, Professor Bowers wrote:

At the same time he offered a succinct restatement of the basic position:

The temporal relationship between composition and presswork here assumed is however capable of many permutations. One might start with evidence of presswork and seek signs of, or infer, compensating adjustments in composition; alternatively, one might begin with some knowledge of the speed of composition and then try to trace evidence of presswork to match. In the first case the evidence of presswork

will almost invariably be in the shape of skeleton formes, although their interpretation may not always be straightforward.

Do they, for example, indicate one press or two? Professor Bowers long ago remarked that "the evidence of running-titles to determine the number of presses is often dubious in the extreme and its application hazardous."[46] And Greg expressed some doubt about the equation of skeleton formes with presses.[47] Yet such equations have been made. Professor Price, writing of Your Five Gallants, claimed that "in 1607 [Eld] had at least two presses, as the running-titles . . . show." And again, writing of Michaelmas Term, "the series of running-titles seem to imply that at least four presses worked on the book."[48] Professor Bowers: "Since regularly alternating two-skeleton formes produce maximum efficiency for one-press work, the staggered appearance of three skeletons . . . suggests the use of two presses." On this assumption, it becomes possible to observe a "mathematical regularity of transfer between the presses according to a fixed and efficient system"; hence "the three-skeleton pattern . . . is proof of two-press printing" (my italics).[49] And again: "The analysis of running titles reveals that two presses printed Q2 [Hamlet]," each press being served by a different compositor.[50] It is not surprising then to find others writing of "a normal pattern for two-compositor work in which each man serves a different press."[51] And writing of The Revenger's Tragedy, Professor Price noted that the four skeletons present suggest two presses, adding that elsewhere "Eld's pressmen clearly revealed their use of two presses by printing on different stocks of paper."[52]

The attractive simplicity of Professor Bowers' initial proposition about skeleton formes is no longer easy to discern. Nevertheless it has been repeatedly put to use in order to determine also the number of compositors engaged on a book. In an article on the printing of Romeo and Juliet Q2, for example, we are told that "variant compositorial characteristics suggest the presence of two compositors" and are assured that "the mechanical evidence of presswork corroborates that suggestion" (my italics). The quarto was printed from two skeletons recurring in regular sequence. The writers continue: "This evidence from running titles can be explained only with great difficulty as accompanying the work of one compositor; but a reasonable explanation may be offered by resorting to the hypothesis of a second press, and thus of a second compositor" (my italics).[53] Earlier, Professor Bowers had remarked that "printing by two presses must necessarily require the services of two compositors" ("Bibliographical Evidence," p. 166 n. 13). Again that a "general alternation involving the use of four skeleton formes is inexplicable for printing with one press; yet if we hypothesize two presses it follows that there must have been more than one compositor."[54] In another case, where only one skeleton-forme was used, "the running-title pattern indicates no second workman."[55] Professor Turner has written: "One skeleton ordinarily means one compositor; two may mean two setting simultaneously . . .".[56] But the clearest example of skeleton formes in relation to composition is offered by Hamlet Q2, in which "compositor X served one press and

imposed his formes for that press, whereas compositor Y served a second press and, correspondingly, imposed his own distinct formes for that press." These observations led Professor Bowers to remark that "when, as in Hamlet, the spelling tests for compositors equate so precisely with what one may conjecture to have been their stints from the evidence of running-titles, we may be somewhat more confident in the future about roughing-out two-compositor work in books on this running-title evidence" ("Printing Hamlet Q2," pp. 41-42).

There would appear to be enough flexibility in the principle to allow its reverse application, for, as in the case of Romeo and Juliet, two skeletons may become two presses if there is some slight evidence of two compositors (and hence "corroborate" the suggestion that there are two compositors). But since compositors have left no skeletons they are less easy to detect than headlines, and there are therefore fewer cases of presswork conditions being inferred from the prior evidence of composition.

By now I hope I have, at the very least, made clear by selection and juxtaposition the multiple and often confusingly diverse general statements inferred from the number and order of skeleton formes,[57] and laid bare the fundamental assumptions about desirable ratios between compositors and press-crews. It simply remains to ask how reliable such analyses would turn out to be if tested by analogy (a fair enough procedure, since their authors imply extended application by analogy).

Two Cambridge books may serve: Beaumont's Psyche (1702), a folio in fours, and Newton's Principia (1713), a quarto (Cambridge University Press, I, 126-7, 219-21). To take Psyche first, and in particular the quires 2E-2Z which can be related to full work-flow charts,[58] we may observe that the edition was 750; that only one compositor worked on it at a time (Bertram set 2E-2F, 2K-2Q½, 2T½-2Z; Crownfield 2G-2I, 2Q½-2T½); that four skeleton formes were in regular use; and that setting and printing of these 19 quires, or 38 sheets, took about 20 weeks in all (from mid September 1701 to 31 Jan. 1702). All I wish to establish now is the futility of attempting to infer any direct

correlation of presswork with these conditions from the four skeleton formes. Two quires set by different compositors (2F-2G), and with identical forme patterns, were both printed at the same full press; two quires set by the same compositor (2K-2L), but with utterly different forme patterns, were both printed by the same half press; two quires set by the same compositor (2O-2P), and with identical forme patterns, were both printed at the same half press; all four skeletons appear in quire 2T, one sheet of which was printed at one full press and the other sheet of which was printed at another full press; all four skeletons appear in quire 2U, one sheet of which was printed at full press and the other sheet of which was printed at half press. And yet if this evidence were not available it would be perfectly respectable to infer that this regular use of four skeletons might mean either (a) a large edition; or (b) two compositors, if not three; or (c) at least one full press in continuous operation.

Newton's Principia is a little easier to deal with since it is a quarto, and although there are four skeletons in all, only two of these were in use at any one time. The first printed most outer formes in sheets C-2P, the second most inner formes; in 2Q-2V their roles were reversed. New skeletons were constructed for 2X, one printing all inner formes to 3P, the other all outer formes. Under these very straightforward conditions, we might normally infer one of the following:

I trust that this example is thought to be no worse for its approximations than most such arguments, but it seems to me to point up once more the nature of our guiding assumptions about skeleton formes and the relationship of composition to presswork. In doing so it also indicates the likely error in our general statements on these matters since their claim to represent the truth can be falsified by contradictory case-studies. If, as for many books, there were no external evidence to control speculation, any of the explanations given above, suitably

sophisticated, could be employed in a publishable account of the printing of the Principia. The facts of the matter are: the edition was 700; only one compositor at a time worked on the book (Pokins set B, 2Q-3R, a-d; Délié C-2P); the book was printed at both full press (49 sheets) and half press (17 sheets) without these conditions in any way being reflected in the number or order of skeletons; when there were changes in skeletons, press conditions remained constant; a delay of some months in 1712 is unmarked by the skeletons; the mere reversal of skeletons at 2Q was preceded by a delay of 11 months; the creation of entirely new skeletons for 2X may have been related to a delay of about 3 months; the first pair of skeletons were in use for over 2 years, the second for 1½ years; printing of the 66 sheets in the book extended from October 1709 until May 1713.[59]

I have not examined the skeleton formes in Bowyer books, but the fact that the sheets in them were often printed at one press and perfected at another must render very complex indeed any analysis seeking to relate compositors, formes and presses — even with the help of press figures. It cannot be assumed that other and earlier presses did not do likewise (without the figures); it is just that we happen to know for certain in some cases what the Bowyers did.

I wish now to broaden the argument a little by adverting to the Shakespeare First Folio and by offering yet another case-study. When Professor Hinman writes:

When Professor Hinman makes this point, he is deducing the probable size of the edition from an hypothesis about timing. The tentative nature of this deduction is made very clear and Professor Hinman's scholarship is of such excellence that it is seldom possible to offer views that he has not already entertained. Yet there can be no mistaking the main import of the above passage: two compositors and one press working on the Folio alone yield 1200 edition-sheets a day. Obviously, without full information about all other work on hand, one cannot falsify Professor Hinman's argument or its implication that the printing of the Folio was, by and large, a self-contained operation. Nor, without such evidence, can one prove it. But it may be salutary to consider its status as a general proposition which is likely to be true for other books of the period. For, as Professor Hinman himself says,

Now if roughly comparable books show quite different conditions of production, the above hypothesis about timing and the deduction from it about edition size will be weakened. More, its proven inability to predict the other possibilities will severely limit its standing as a statement of general application.

Volume I of the 1705 Cambridge-printed edition of the Greek lexicon Suidas is perhaps a book that is "roughly comparable" (The Cambridge University Press, I, 224-33). It is a folio in fours, with some 954 pages, about 8,500 ens per forme set double column in English Greek and English Latin with Long Primer footnotes; being started by 1 Nov. 1701 and finished by 4 Sep. 1703, it took some 22 months to print: the edition size was 1500 (150 large-paper copies, 1350 smallpaper); it was set throughout by two compositors working simultaneously on each forme; 166½ sheets were printed at full press and 72 at half press. The Shakespeare volume is a folio in sixes, contains just under 900 pages, has about 10,600 ens per forme set double column in Pica English; according to Professor Hinman's table it took about 18½ months to print but if we add the 2½ months given as the possible length of an interruption, the total would be 21 months; and Professor Hinman suggests an edition of about 1200 on the assumption that two compositors and one press were working on it more or less continuously. It would be foolish to think of these two books as being any

more than only very roughly comparable; the Shakespeare Folio is in English, not Greek and Latin, and has 22 fewer sheets than the Suidas volume. Yet the Folio has slightly more ens to the sheet and Suidas took slightly longer to print, so that in the quantitative matters of bulk, relation of composition to presswork, printing time, and edition size, the two books are not perhaps so very different.

When we discover, however, that throughout exactly the same period as the one in which the first volume of Suidas was being printed the identical one-and-a-half presses that printed it also served three to four other compositors, two of them often working simultaneously, to print another 1500 copies of the second volume of Suidas — yet another Shakespeare Folio as it were — as well as 20 other books whole or in part and at least 23 smaller jobs, then we might be forgiven for thinking that Professor Hinman's estimate is badly at fault. Nevertheless my point is not that his equation (two compositors and one press yield 1200 edition-sheets a day) is wrong — indeed, under some conditions it might well be exact — but that it seriously misrepresents the general conditions of book production.[61]

If I am right, and there is miscalculation somewhere, the reason for it probably lies in an inference drawn from skeleton formes, and its consequences return us to the subject of concurrent printing. For the purity of Professor Hinman's argument virtually commits him to the view that the Folio was printed on one press, "the Folio press." Apart from 18 quires near the beginning, the Folio is a one-skeleton and therefore, it is claimed, a one-press book: ". . . throughout most of the book, indeed, two-press work was manifestly impossible, the same skeleton having been used in successive formes . . .".[62] Professor Hinman here means that two presses cannot have been in simultaneous use, but as he says at another point:

Moreover, two compositors working together and serving this one press are thought to represent an ideally self-sufficient relationship for an edition of about 1200 copies, and since apparently this is the condition more frequently found,

Professor Hinman's main stress here, and his concession, are the crux of the matter. At times of course most analytical bibliographers working in this field have to confess an imbalance of composition and presswork on a book, either implicitly by failing to pursue the point or explicitly by marking breaks in an otherwise apparently consistent pattern. When this happens, some odd jobbing at case or press is a likely and convenient suggestion to restore the ratio and avoid idleness. This opportunistic resort to a theory of concurrent printing need not be documented at length but it is important to note its circumstantial origins. For much of the Shakespeare Folio, set by one man, there is persistent evidence that the economic considerations behind the plan (at least in the form suggested) did not apply; and to explain the apparent imbalance Professor Hinman allows that 'the Folio press' must have engaged in some concurrent printing:

And again:

If I now seem to labour a point it is simply because Professor Hinman's account of the printing process reflects and therefore lends massive authority to the erroneous assumption that a book was normally put into production as an independent unit. The single skeleton forme, its association with 'the Folio press', the suggested edition size, the 'economic' balance between compositors and press-crew, all combine to reinforce this view. What is offered as exceptional — occasional concurrent printing — other evidence would suggest to be normal; what is offered as normal — a self-contained operation — is elsewhere exceedingly rare. Neither the Cambridge nor Bowyer papers would permit such inferences to be drawn from skeleton formes; neither would permit such assumptions to be made about the operations of a single press; neither would be consistent with the general economic argument put forward. Nor is it, I think, a matter of proven historical difference, as though the early 17th century were doing something that the 18th century no longer found necessary. For no differences have been constated that cannot be seriously questioned by exposing the primary assumptions. Noting at one point that most of the Folio was printed by a single press, Professor Hinman revealingly adds:

With all respect, one is obliged to say that 'for all practical purposes' it would not. The moment we admit the possibility of two presses we halve the work of one of them on the Folio and concede that each is concurrently printing other books as well. The problems of calculating the ratio between compositors and press-crews are doubled, for the ratio must be assessed for each crew and the assessment must take full account of all other work on hand. The pattern of work becomes far more complex as the various edition-sheets for different books are printed off one with another at the two presses. For some limited theoretic purposes the ratios abstracted by Professor Hinman may be sound, but his evidence, as he concedes, is consistent with normal conditions of concurrent printing at press, and much of the time with concurrent work at case.

The implications of assumptions which seem to be so much at odds with usual printing conditions do not end here:

Under the latter conditions 'there would be a gross imbalance between composition and presswork', under the former 'a highly efficient ratio'. Professor Hinman makes much of the economic reasons for simultaneous setting, and hence setting by formes: Again, an economic ratio of composition and presswork for the Folio By displacing type-shortage as the primary reason for casting off copy, and substituting an economic relationship dependent on assumptions about timing, Professor Hinman not only diminishes the classical status of his own major demonstration but starts a bibliographical hare. The constant factor throughout the Folio is shortage of type because of the method of quiring, and Professor Hinman himself makes it clear that this "may have made it more or less mandatory to set the Folio by formes." His attempt to give an extended generality to his brilliant particular and practical proof from type-shortages misrepresents the general conditions of work, not only in Jaggard's shop but in the period as a whole.

For this again is my immediate concern: the encouragement given to the view that even where there is no conclusive evidence of type-shortages, revealed by the presence of identical sorts in both formes of a sheet or in the first half of a quire, we may have setting by formes. In such cases reliance is usually placed on a 'pattern of distribution' — evidence which is used with most admirable insight and control by Professor Hinman but which, in lesser hands, and in quarto printing,

may prove very tricky indeed. In a folio of course the major production unit was the quire and (now that Professor Hinman has pointed it out) it is obvious enough why in a folio in sixes type should be inadequate for normal page-by-page setting. In a quarto, however, the reasons are far from clear, and despite Professor Turner's assurance that "it begins to appear that [Elizabethan play quartos] were more often than not composed by formes" (my italics) ("Printing King and No King", p. 255), a certain scepticism ought perhaps still to be exercised. Professor Williams, for example, arguing that the quarto Epicedium (1594) was so composed, implied considerable concern on the compositor's part at setting vv for w, an assessment by the compositor of the number of w's required for the work, a count of those available to him in the case, and a decision to set by formes for this one reason despite attendant complications. In another case "Random mixing of roman and italic forms of 'k', 'K', 'S' and 'Q' . . . are common in the quarto and are without significance. The shortage of lower-case 'w', on the other hand, discloses a pattern throughout the quarto."[67] But one may fairly ask whether it is safe to prove a case by accepting only such limited 'patterned' evidence. The idea that a 'pattern' must be significant because it appears to indicate a regular method of work is one of the most perniciously seductive presuppositions of current bibliographical analysis. The conflicting evidence of 'k' is disregarded in part because it "violates the order of imposition and printing as disclosed by the running-title evidence." For "evidence from running titles indicates that outer B preceded inner through the press." This of course is merely a further assumption given the status of proof (and then applied as such) because the skeleton formes can, but arbitrarily, be ordered in a pattern.

In the Shakespeare Folio, Professor Hinman noted, "As a rule . . . no forme has types in common with either the forme immediately preceding or that immediately following it" (Printing and Proof-Reading, I, p. 81); and the sequential relation of setting, printing, and distribution here implied has been adapted for the quartos. Professor Turner had earlier given it shape when, writing of Philaster, he observed

Mr John Hazel Smith, discussing Much Ado, offers a similar argument: "The precedence of A(i) is proved by three italic types (B1, B2, d1) from that forme which are then divided between the formes of B." And again: And Professor Turner's subsequent formulation gives the principle a usefully definitive form: Mr Smith's study is probably the least fortunate example of an attempt to prove setting by formes in a quarto by "applying scientific bibliographical methods," for Professor Hinman has since pronounced it wrong.[70] But the irony is that the methods used by Mr Smith were those "illustrated by the work of Charlton Hinman on the Shakespeare First Folio and by George W. Williams and Robert K. Turner, Jr, on other Renaissance quartos" ("Compositor of Much Ado", p. 10). And some of these methods are not, inherently, very reliable.[71]

The comment called for here can only be a very general and cautionary one. Neither the Cambridge nor Bowyer records offer much positive evidence of setting by formes; although their combined testimony does demonstrate the rarity of such a practice for books other than page-for-page reprints and must therefore give us pause. We must recall too that neither Moxon, Stower, nor any other early grammar mentions casting off as a means of enabling work to be set by formes. In every case it is, as Stower puts it in his index, a "manner of calculating in order to ascertain the number of sheets a manuscript will make, the size of the letter being fixed on"[72] — that is, a device for costing, and for determining the paper required, not for organizing work. Nowadays we call it estimating. As Professor Hinman observes, actual casting off for setting would not have been undertaken without good reason, although it is true that the difficulties may have been overestimated for verse plays as distinct from full prose works (Printing and Proof-Reading, I, 73). But if, as is claimed, "the practice was by no means uncommon" and "is to be seen in first quartos that issued from many different printing houses, over a wide stretch of years" ("Shakespeare's Texts," p. 31), it is to be hoped that firmer controls will be applied in its demonstration than have hitherto been evident. In particular, arguments heavily reliant on time-schemes will rarely command that ready assent which was given to Professor Hinman's initial proof that the Folio must have been set by formes.[73] On the face of it, the most important reason for setting by formes in quarto is unlikely to have been urgency, nor even an unusually small fount, but a fount depleted because of concurrent printing — for if work overlapped on two or more books using the same fount of type, setting by formes would offer a method of making some progress with all. Professor Hinman has again led the way in showing how, in Jaggard's shop, concurrent setting of other books, reduced the supply of type for the Folio.

If copy is cast off for a quarto text, there is no compelling reason why any sheet should not be printed in any order — say, H, F, A, C, D, B, E, G. One might expect and assume a straightforward progression through the book, but there is no compelling reason for it. But order of formes through the press is an important ingredient in much bibliographical work. Where there is detectable damage in the course of printing (whether to types, headlines, rules or ornaments) it may be quite possible to prove order, and in some cases a precedent forme, at least within the same sheet, may be determined by using the Martin-Povey lamp. I am not sure whether it is evidence of this kind that led Professor Turner to write that "information about presswork, specifically the order of the formes through the press, is relatively easy to obtain and is based on evidence that is the least controvertible" ("Beaumont and Fletcher Folio", p. 36), but, so far as I can tell, order has usually been determined, not according to such evidence, but according to a pattern of headline recurrences. "Evidence from running titles indicates that outer B preceded inner through the press" (Williams, "Setting by Formes", p. 43) is a familiar form of wording; or "on the evidence of running titles, it is clear that B(o) was machined before B(i)". I confess that I have never understood what was meant when I have read such a phrase, and again I suspect that priority is based on assumptions about timing, and inferences drawn from variants, from a pattern of alternating skeletons, or from reappearing types which permit a hypothesis about distribution. In any case, whatever the internal patterns which some physical features may take within a book, there is little reason to elucidate them by constructing a time-scheme or by supposing the successive printing of all formes of the same book. I know of no evidence that obliges us to think of one sheet (or forme) being followed immediately on the press by another of the same book. There is some case for it when perfecting, none between sheets. There is too much evidence in the Cambridge books of perfectly regular patterns sustained under the most diverse conditions of concurrent printing. It is not always easy to tell when an apparently general statement is really only a singular one made of a particular book, but if it is generally true, as Professor Turner says, that "to prove the order of printing is usually to prove the order of composition of the formes" ("Beaumont and Fletcher Folio", p. 37), important textual consequences may follow from the initial assumptions.[74]

It is perhaps worth looking briefly at one Cambridge book, Bennet's Answer to the Dissenters Pleas (2nd ed., 1700). Its testimony is not all that important, since it is a page-for-page reprint, yet it does show quite vividly that, once copy is cast off, any sequence of setting and printing might be followed. The sheets were composed as follows: Bertram set E, K by 13 Jan. 1700; B, H, S by 20 Jan.; Knell set C by 26 Jan.; Bertram D, L, U and X/* by 10 Feb.; Knell completed F, G½ between 20 Jan. and 17 Feb.; G½, I by 24 Feb.; Bertram P, Q, R by 24 Feb.; N, O, T by 2 Mar.; Knell A, M by 9 Mar. The order of printing appears to have been: E, H, K, B, L, S, U, C, D, F, X/*, G, I, R, Q, N, O, P, A, M, T (Cambridge University Press, I, 192-3).

The Cambridge papers, if not those of the Bowyers, provide very clear evidence of regular proof-correction of all books printed. Such a practice may have been slightly unusual as many of the books were classical texts and the press prided itself on its accuracy, yet I think not, for London houses in the 18th century, like Cambridge, regularly set as their price for proofing one-sixth of the rate of composition.[75] There is considerable doubt, however, about the validity of applying 18th-century evidence to Elizabethan books; even Moxon's testimony from the later 17th century has been rejected as irrelevant to the earlier period. If this is so, then the 'norms' used to introduce some measure of probability into analytical accounts of the proofing and printing of earlier books will themselves be only inferential. Moxon, we may recall, notes that:

Having corrected it, the compositor

And before continuing printing, the pressman will check

Professor Bowers has remarked, however, that

To make this point is to stress again the primary importance of continuous printing at press. Professor Hinman would doubtless agree, for he says that

But Professor Hinman's basic reason for rejecting Moxon's account as in any way relevant to the 1620's has little to do with timing. It is rather the many self-evident errors that survive in the printed text. Discussing — and dismissing — in a footnote the idea that regular proofing may have preceded that established by a collation of the variants he has observed, Professor Hinman notes:

It is a view that in general Professor Bowers would probably — and reciprocally — endorse, since he has observed that

It may seem singularly fool-hardy not to follow such authority, but I am constrained to persist in a certain incredulity. Professor Hinman's failure to list the 'many obvious errors of all kinds' at least makes one's task of qualification a little easier since he has not, in this case, sufficiently illustrated, let alone proved, his point. If Moxon, and proofing practices so well evidenced elsewhere in the century and beyond it, are to be displaced as the 'norm', the question would seem to demand rather fuller discussion than I have yet seen devoted to it.

The view that Professor Hinman is concerned to question is, essentially, Greg's — that in the Folio "the printer was not indifferent to the accuracy of his text."[76] And it may well be that if we were "once possessed of a full record of the press variants in the First Folio" (Printing and Proof-Reading, I, 227), such a view might have to be altered. Professor Hinman's labours of collation leave him in no doubt that now "there is in fact considerably more evidence that the printer was largely indifferent to the accuracy of his text" (I, 227). Yet such a conclusion is scarcely judicious; there is a great difference between the truth and the whole truth, between "a full record of the press variants" and a full record of the surviving press variants.[77]

This is not just a quibble. Traditionally the stages of proof-correction have been at least three: galley (whether page- or slip-), revises, and, as a last resort, stop-press. And let us not forget that the manuscript copy precedes all three. Now it is incontestable that these several stages can be found in increasing frequency as one moves from manuscript (how much of that survives?), to page-proofs (very few of these), to revises (slightly more of these — if some of our surviving 'proof' sheets can be so considered), to stop-press (hundreds of these). Each successive stage supersedes the previous one; once the unique copy has been set and checked, it can be disposed of, once the single galley proof has been read and checked, it can be disposed of, once the revise has been read and checked, it can be disposed of, but once printing has started, the multiple copies are preserved and of course they are available for consultation in those portions of the edition still extant. It

only remains to note that the principle of increasing frequency persists even here, for, as Professor Hinman has observed in the case of the Folio, the earlier 'uncorrected' state is likely to be preserved in about ten per cent of copies, and the later and latest, press-corrected, states in ninety per cent of surviving copies; these last therefore will be the ones most frequently observed. We must of course work from what we have to what we have not, but our chances of going the full distance and thereby establishing 'a full record of the press variants' — if these are taken to include all stages of proofing — are very remote indeed. When such evidence (of its very nature) demanded to be discarded, it is difficult to see why one should assume that it never existed.[78]

But one may consider the point in another way: it is easy after repeated and intense scrutiny to discover 'obvious' misprints, and it is also very easy to miss them. Each year I put some four or five senior and intelligent students through the rudiments of type-setting and when they come to correct their work they almost invariably have to do it in two or three stages because these latter-day John Leasons have failed to correct all the 'obvious' errors the first time through. Yet there was a first time (see Printing and Proof-Reading, I, 233). It is true that the more experienced students make fewer mistakes, but it is again remarkable how many of these mistakes my latter-day John Shakespeares overlook in their first attempt at correction. Much the same point is made of course by Professor Hinman when speaking of sections of plays set by Compositor E, sections "which were subjected to much more proof-reading than others — yet only to very careless proof-reading, since a great many errors nonetheless escaped uncorrected in these plays" (I, 233). Errors, that is, persist through one or more stages of proof-reading; the much-proofed page from Antony and Cleopatra leaves errors uncorrected.[79] The existence of some formes in three or more states indicates that at one or more stages of correction errors were missed which were later thought serious enough

to alter. Greg's list of misprints in Q1 Lear, based on the corrected state of the sheets and therefore taking no account of the original errors that were subsequently altered, is most revealing in this connection (Variants, pp. 63-79). Professor Hinman's new evidence from variants introduced at a relatively late stage of the Folio's production does not dispose of Greg's judgement. The 'obvious errors' have been there since 1623 and Greg, who had as good an eye for them as anyone, still thought that "the printer was not indifferent to the accuracy of his text", and that "he took what were thought in his day to be reasonable precautions, and went to some trouble, to reach a moderate standard in the execution of what may not have been at all times an easy task" (Shakespeare First Folio, p. 464).

There is of course another way of looking at the problem — and I must repeat that I am really only concerned with questions of method and that like Troubleall I merely wish to ask 'by what warrant' certain inferences are given the standing of general statements. So, a priori, one might ask whether it is likely that the essentially trivial corrections noted by Professor Hinman would have been made at all if the printer were indifferent to the accuracy of his text? Or, to put it yet another way, is it likely that a printer who put up with so many bibliographically serious delays at press in order to correct minor blemishes would fail to observe routine correction procedures in order to avoid major infidelities and the prospect of really serious delays in the last stages of production? Which brings us back to Moxon.

It is not I hope gratuitously irresponsible to suggest that none of the evidence presented from the Folio demonstrates conclusively that the procedures which Moxon describes were 'essentially different' from those of the 1620's. At the very least, the "Proofe, and Reuiewes" pulled by Jaggard for Brooke's Catalogue testify to the currency of Moxon's terms at this time, and in Jaggard's shop (McKerrow, Introduction, p. 207). We must grant that the copy for the Folio has disappeared; we must grant that the foul proofs have disappeared; but what does remain in evidence corresponds exactly to that which we should expect to find at the later stages of correction as outlined by Moxon. And it is precisely at these stages of correction that copy is not consulted. That phase is well behind, and even if some errors have persisted it is not to be expected that substantive matters will now command 'painstaking' attention. But it is to be expected that typographical infelicities — the things that catch (and for long curiously avoid) a pressman's eye through the repeated pullings — will be picked up from time to time. Turned letters, lifting spaces, uneven inking, badly

defective letters — these are precisely the things which at this stage the beater, who "peruses the Heap" (Mechanick Exercises, p. 292), was deputed to look for. He takes care Rising letters, quadrats and furniture, and probably loose spacing, are fixed by the pressman, who has a bodkin for the purpose; but if letters are to be replaced, "he Unlocks the Quarter they are in, and desires the Compositer to put in others in their room" (p. 304). None of this is inconsistent with what we find in the Folio; indeed Professor Hinman's variants are clear evidence that Jaggard's beater was doing exactly what Moxon demands — turning out a book that was not marred by too many purely typographical blemishes.

Let us now recall too what Moxon says of revises and of correction at press: the forme, being now on the bed of the press, is left there, "and if any Faults are found in any Quarter of it [the Compositor] calls to the Press-man to Unlock that Quarter . . . that he may Correct those Faults" (pp. 238-9). There are several points here: the kind of corrections documented for the Folio are unlikely to have required removal of the forme from the press. At this stage, even after as many as three proofs, a revise is pulled — but now the forme is virtually ready for printing and the likelihood is great that printing will begin while the revise is being looked at. The single copy of the revise is likely to be a pull of the full forme; in the case of the white-paper forme its chances of being preserved are negligible, but in the case of the perfecting forme this single marked sheet has a greater chance of being placed on the heap and eventually bound. Although the revise will be of the full forme, Moxon suggests that it might be attended to in sections, or quarters, so that the pressman might unlock only so much of the forme as is necessary, perhaps only a page in the case of the Folio. In many cases in the Folio both pages must have been unlocked and corrected together, yet Moxon's wording does hint that the revise itself might be read in sections. I find it most interesting therefore that Professor Hinman should write: "Four actual proof-sheets for the Folio have survived — although . . . they ought perhaps rather to be called proof pages"; and "the essential proof-reading unit, so to speak, was rather the single page than the complete forme" (Printing and Proof-Reading, I, 233, 234). I wonder, however, whether

these four proof pages should not perhaps be referred to as revises to distinguish them from the first and substantial stage of correction as well as from the last and accidental one.

In any case, however cursory we may think the Folio by our own standards and in the absence of author-correction, it would seem premature to conclude that what so closely corresponds to revises and stoppress correction as described by Moxon was not preceded by the routine proofing procedures which he also outlines. These may have been deficient in execution, but I cannot think that Professor Hinman's inferences justify the view that "the method of printing and proofing adopted", whether in the Folio or beyond it in the earlier 17th century, "was essentially different . . . from the method described by Moxon some sixty years later" (I, 228).

Professor Hinman, in another context, also discounts the testimony of Ashley's translation of Le Roy (1594), "since Le Roy was not a professional printer" (I, 41). Ashley writes of the pressman who is pulling:

The late Mr Kenneth Povey found reason to believe that Ashley had expert help in making his translation; and it may be further noted that Jaggard's precise use of the word "Reuiewes", both as a noun and as a verb in the phrase "viewed, reuiewed, directed, corrected", suggests that Ashley's use of the word "reuiewed" was not idle. Ashley is at one with both Jaggard and Moxon in suggesting that, first, there might be two or three proofs and, next, a 'reuiew' or revise. Even the phrase "continew the rest" could relate to a process of continuous printing, stopped to make the late changes found in the revise, and then resumed. Mr Povey used Allde's 1624 edition of Massinger's The Bond-man as a test case, and found all the variants reconcilable with an orderly routine of proof-correcting and perfecting described by Moxon. He suggested, moreover, that Nowhere perhaps so much as in the consideration of skeleton formes and proof-correction procedures are modern constructions so crippled by the absence of primary evidence and so vulnerable therefore to the general objections that may be made to all inductive methods. One recalls Black: "a nice adaptation of conditions will make almost any hypothesis agree with the phenomena." It is doubly a pity, however, when writers adapt conditions to suit their theories and then find themselves obliged to discount the testimony of such an excellent palmer as Moxon.

But much the same may be said of many studies of 18th-century printing which have been conducted on the assumption that conditions then were essentially different from those of the preceding century. My own major argument in this paper is of course that productive conditions were constantly changing, not just from century to century in different houses, but from day to day in the same house, simply because concurrent printing has been the universal practice for the last 400 years. If I am right, this fundamental fact poses more problems for analytical bibliography than any minor period differences. These there certainly were, and they must be carefully charted, but we must beware of that ostensibly sophisticated historical relativism which insists on making fine distinctions between periods when virtually nothing certain is known about either element of the comparison. When, for example, Professor Todd writes that

the implication is that we are here dealing with quite distinctive conditions; but this, as the song says, ain't necessarily so. Whatever the variables, labour was always predetermined, controlled, and properly recorded, whether on piece rates or not. Crownfield's disposition of work, usually without figures, was no different from the Bowyers' with figures, and Stower's "Plan of a Book for checking Compositors' and Pressmen's Bills" assumes exactly the same conditions a hundred years later when press figures are on their way out (Printer's Grammar, p. 435). The procedures have always been the same — only the methods of recording them have differed.

I cannot here attempt to describe the thick web of theory spun around press figures.[81] In their incredible and perplexing variety they are eloquent witnesses to the customary conditions of presswork in any printing house, and perhaps only an imagination as fertile as Professor Todd's, and a mind as subtle, could have penetrated their mysteries. On their usefulness, let him speak:

Much bibliographical writing, like that in any new subject, has a strong proselytizing strain which is apt to show itself in a slight tendency to rhetorical overstatement and the premature elevation of particular observations to the status of general truths. One or two pieces of information that have become available since Professor Todd wrote the above account do call for its qualification; I present them now only to carry forward my general argument that the 'empirical' method, with its reliance on 'direct observation',[82] might lead us wildly astray. Press figures, as Professor Todd has indubitably shown, are of enormous value in revealing conditions normally concealed, but they still need theories to make them work and the theories so far applied have been largely without benefit of primary evidence from the printing house itself.

As I have indicated elsewhere, Cambridge pressmen in the early 18th century did not normally use press figures, and the first two volumes of Suidas are the only two books of the period in which they appear (Cambridge University Press, I, 128-32). This exception is wholly due to the employment of John Terrill who came up from

London towards the end of November 1701 and who left Cambridge again on 15 May 1703. Terrill's bills for presswork match the figures exactly and make it perfectly clear that, in this case, the figures represent a man, specifically Terrill, not a press. Terrill did not always use a figure, nor keep to the same one; and it is certain that here in the Cambridge house his use of a figure was a purely personal and optional matter. His main reason for using one at all would seem to have been that the first two volumes of Suidas were being printed concurrently and as they were independently signed 'Suidas the quire G' or 'Suidas G1' might refer to either or both volumes. So Terrill played safe by marking the sheets that he printed, although he thought it necessary to figure only one forme in each sheet. In this, his practice was consistent with that recommended by Savage in 1841 but not with that followed by the Bowyers in the 1730's.[83] Terrill's main concern seems to have been merely to use some idiosyncratic mark, and once the work had been paid for any other might serve as well. It is not surprising therefore to discover that Terrill used two different figures (* and ‡) in both volumes. The first (*) was used between the end of November 1701 and 28 Feb. 1702; during this period Terrill worked under markedly different conditions at different times — first with Brown at full press and then alone at half press — yet he used the same figure throughout. From 28 Feb. until 2 May 1702 he worked with Ponder without a figure. Thereafter, until his departure on 15 May 1703, Terrill used the second figure (‡), again in both volumes, to mark almost every sheet on which he worked. When he left, the figure disappeared.

Clearly the consistent use of one figure in one part of a book and of another figure in another part has in this case nothing whatever to do with simultaneous — or even successive — printing of each portion at different presses. Nor has the incidence of variously figured and unfigured sheets anything whatever to do with printing at full or at half press. If, even occasionally, a pressman was personally responsible for his choice of figure, as here, this would go far to account for the many idiosyncratic numbers or marks adopted in some books and their apparently haphazard arrangement. And if, even occasionally, a figure represents a man rather than a press, it is formally possible to argue that a sheet which shows varying figures in copy to copy simply reflects changes in the press-crew part-way through a single impression and not distinct impressions.

Indeed, whether a figure indicates a pressman or a press, such variation is in any case to be expected in books printed in very large editions. The London Magazine, for example, was printed for a time in 8000 copies. Since it comprised three and a half sheets, its printing would have kept three full presses wholly engaged for more than a week. Over such a time span — longer if the presses were required to do other work too, as Ackers' were — it is highly probable that changes would occur in the conditions under which the single impression would be completed. Changes of men, as well as changes of press, part-way through printing might well be reflected by new figures yet none of them be bibliographically significant — or at least no more bibliographically significant than the daily discontinuities incident to all printing in large editions. Naturally such evidence would rarely be left to stand alone; at the very least it would set one searching for new skeletons, partial re-settings, advertisements and so on; my point is the quite simple one that the relationship between variant states and distinct impressions must be very carefully assessed if the general conditions of work are not to be misrepresented.

But, as Dr Fleeman has already shown, there is quite conclusive evidence in the Bowyer ledger to associate press figures with a press not a man, evidence which can be corroborated by reference to the printed books themselves.[84] Bowyer numbered his presses and his accounts usually show, by their numbers and crews, the presses at which work was done. If a press-crew had a press of its own at which it regularly worked — and there is some evidence that this was so in Cambridge in 1740 — then the distinction between men and machines would virtually disappear;[85] but the Bowyer papers offer us no such simple resolution. In the examples of Bowyer books listed in appendix II (g), the figures and/or presses and/or crews can be lined up with a certitude unparalleled in any purely inferential construction. Yet it is most important to note, first, how many discrepancies there are between the records and the printed figures (especially in No. 5); second, the difficulty of assuming continuity of press-crews for any one figure; third, the irrelevance of the highest figure printed, although it designates a press, to the actual number of presses in use; fourth, that the occasional failures to figure a forme are in fact oversights and do not represent work done at a notionally blank press. It is another example of the by now familiar paradox: primary evidence definitely restricts

the generality of many statements hitherto made about the interpretation of press figures; yet it reveals such diversity of conditions in their use that almost any answer might well be true for any particular book. It is another warning that, as Professor Todd has put it, "any theory envisaging a uniform procedure in an unorganized, laissez-faire handicraft must be regarded with suspicion" ("Observations on Press Figures", p. 173). When therefore the writer of a review article in the Times Literary Supplement took Mrs Russell to task for not adequately listing press figures in her bibliography of Cowper, and suggested that the printers of certain editions might be identified by the pattern of press figures, or that because the figures 3 and 6 recur in Bensley's editions of Cowper these editions were always placed in the care of the same pressmen, the arguments may be much less 'advanced' than they seem.[86] In the light of complexities discovered by any primary documentation so far unearthed, such an ostensibly direct frontal assault turns out to be no more than a rear-guard action in defence of a much too simplistic and now obsolescent bibliographical method.