University of Virginia Library

The Printing of the Second Quarto (1619) of King Lear
by
Richard Knowles

To date, three theories of the composition of the second (Pavier) quarto edition of Shakespeare's King Lear have been proposed. William S. Kable, following the example of D. F. McKenzie,[1] has argued from spelling changes between Q1 (1608) and Q2 (1619) of Lr. that the whole play was set by Compositor B of the First Folio.[2] In his dissertation of 1971[3] John F. Andrews, also arguing largely from spelling evidence, has revived Madeleine Doran's suggestion[4] that the first eight sheets were set by one compositor and the last three by another; he calls the first man Compositor F (but later G)[5] and identifies the second as Folio Compositor B. Subsequently Peter W. M. Blayney has questioned the reliability of Kable's spelling analyses and has hypothesized, partly on the basis of patterns of skeleton alternations, that all of the Pavier editions except Sir John Oldcastle were set by two compositors working concurrently throughout.[6] On the grounds of additional evidence of the recurrence and migration of distinctive types,[7] I wish to propose a still different theory: that the second quarto of Lr. was set from two type cases by two compositors working concurrently throughout sheets A to H, and thereafter from one of these type cases by a third compositor who by himself set the last three sheets, I to L.

New evidence that Q2 plays some part in determining the Folio text of Lr. makes such a proposal as this potentially more useful than it would have

seemed to be a year or two ago. Although previous attempts to establish the presence of Q2 Lr. behind the Folio text, most notably that of Andrew Cairncross, have not been convincing,[8] in a forthcoming article Gary Taylor produces important evidence that the pages set by Compositor E in F1 Lr. more closely agree with Q2 than with Q1 in punctuation and a number of unusual spellings.[9] If compositorial preferences and habits helped to shape the text of Q2 and thereby the text in F1, one naturally wants to be able to distinguish the Q2 compositors and their characteristics. But even if Q2 proves to have had little or no influence on the received text of Lr., a study of its printing may still have a general interest for what light it may shed on the working methods of William Jaggard's men; the compositors who worked on Q2 Lr. almost certainly worked on other Pavier quartos and very possibly on the First Folio.

Thomas Pavier's quarto of Lr. was printed in Jaggard's shop in 1619, from many of the same types used some three to four years later to set F1.[10] It was evidently set by formes from a copy of Q1 in which sheets D, G, and H were in the uncorrected state and formes E(o) and K(o) were corrected (Doran, p. 3). Though Q2 is not a page-for-page reprint of Q1, accurate casting-off of Q1 copy must have been relatively easy. The collation of Q1 is A2 B-L4 the text begins on B1 and ends on L4, and it is printed 38 lines to a page. The collation of Q2 is A-L4; its text begins on A2 and ends on L4, and it is printed 37 lines to a page. Thus 79 pages of text in Q1 are stretched into 85 pages in Q2. Printing of one fewer lines per page in Q2 achieves a little more than a third of the stretching; the rest is accomplished through the style of the setting. A line of white space is added before most entrance directions; stage directions in the margin and stage directions or speeches doubled up in the same line with another speech or part-line in Q1 are usually lined separately in Q2; a few turnovers and turnunders are carried down and indented to make a separate part-line; and there is some deliberate relining.[11]

The result of such clean copy and a regular house style was very accurate casting-off for Q2. Despite occasional inconsistencies, all three compositors were able to adhere fairly closely to style and needed to make few adjustments in spacing and lining. Only one page is a little short, and a small handful of

others show a little crowding, none of it very serious. Except in cases of deliberate relining, the length of the prose line in Q2 follows that in Q1 very closely (though in long passages some differences occur). In fact, in the five instances where Q1 divides and hyphenates a word at the end of a page (B4v, C2v, D1, G2, and I4), the same end-of-line word-division occurs midpage in Q2 in four cases and differs in the fifth by only two letters. Obviously the Q2 compositors adjusted their measures to allow close copying of the Q1 line and thus to facilitate their working from cast-off copy. Perhaps it was to approximate the Q1 line more accurately that the two compositors who set sheets A and B both seem to have reset their measure from 94 to 95 mm. starting with sheet C. The only evidence of press-correction that has appeared to date is a pair of variants in forme A(o), where in A3 of Folger copy 1 begot (l. 26) and marry (l. 33) read bgot and matry.[12]

As the following chart shows, two skeletons were used in printing Q2 Lr.

• Skeleton I
• Running title I: A2v, A3v, C1, C2, E1, E2, G1, G2, I1, I2
• R-t II: A3, A4, C2v, C1v, E2v, E1v, G2v, G1v, I2v, I1v
• R-t III: C3, C4, E3, E4, G3, G4, I3, I4
• R-t IV: A4v, C4v, C3v, E4v, E3v, G4v, G3v, I4v, I3v
• Skeleton II
• R-t V: B1, B2, D2v, D1v, F2v, F1v, H2v, H1v, K2v, K1v, L2v, L1v
• R-t VI: B2v, B1v, D1, D2, F1, F2, H1, H2, K1, K2, L1, L2
• R-t VII: B3, B4, D3, D4, F3, F4, H3, H4, K3, K4, L3, L4
• R-t VIII: B4v, B3v, D4v, D3v, F4v, F3v, H4v, H3v, K4v, K3v, L4v, L3v

Skeleton I was used to print both sides of sheets A, C, E, G, and I, and skeleton II was used to print both sides of the alternate sheets B, D, F, H, and K, and of the final sheet L. It is a somewhat unusual pattern for a two-skeleton quarto; ordinarily one skeleton would be used in printing the first side of a sheet and the second skeleton would be used in perfecting the sheet. But what is particularly interesting about this pattern is that it runs through the first nine of the ten Pavier quartos: The consistency of this pattern from one play to another suggests a regular plan of printing decided upon from the beginning of the series and closely adhered to up to the last volume.

Of the eight running titles in Q2 Lr., all of which read The History of King Lear, four suffered damage in the course of printing. The f in title V is whole in B1, damaged in B2; in some copies (e.g., Folger copy 2, British Library copy C.34.K19, Newberry Library copy) its top is just beginning to bend backward in B2, in all other copies that I have seen it is bent sharply up and back, and in all its appearances in D, F, H, K, and L the top has been imperfectly bent back into its original shape. The letter f also degenerates in titles I (whole in G1, topless in G2 and in both formes of I) and VI (whole in H1, topless in H2 and in K and L). The final r of title VII is whole in K3 but lacks part of its curl in K4 and in L. Titles V and VI exchange places between B and D and remain reversed in F, H, K, and L. All of this skeleton evidence indicates two things: since the type breakage always appears in the inner forme, it suggests that here, as apparently elsewhere in the Pavier quartos, the outer forme was normally printed first; and since the type breakage and reversed titles reappear in normal sequence in eight of the eleven sheets, one may safely assume that all of the sheets were printed in alphabetical succession.

Although the type breakage in the running titles suggests that the outer forme was normally printed first, it is impossible to know that for certain except for the four sheets (B, G, H, and K) where such type damage occurs. Neither with scanning light and loupe nor with a three-dimensional microscope (kindly made available by Mr. John Aubrey, Supervisor of Special Collections) was I able to find type impressions in the pages of the Newberry copy of Q2 Lr. which could decide the question. Since in all known instances here and elsewhere in the Pavier quartos the outer forme is printed first, I simply assume it to be the usual practice, lacking any evidence to the contrary.

The recurrences and migrations of distinctive types show very clearly that the first nine sheets of Q2 Lr. were set from two cases of type, which for the sake of convenience and custom I call case x and case y.[13] Types from case x appear in sheets A, C, E(o), F(o), G(o), H(o), and I(o), and those from case y in sheets B, D, E(i), F(i), G(i), H(i), and I(i); a few types from case x also appear in G(i). In the last two sheets, K and L, types from both cases appear, but it is evident from the distribution patterns and from residual or linking types from earlier pages that the single compositor now completing the book is working from case x. The flow chart in Figure 1 indicates the chief and most immediate sources of the types in each forme; individual linking types, which show up in a later forme after having lain in the case during the setting of intervening formes, are not represented in this chart, important though they are as confirming evidence. As is usually the case in this kind of analysis, I have had to discount a few appearances of individual types in the wrong

places, assuming that they represent misidentification on my part, or confusion of two different types that have sustained similar damage or malformation, or that they are subject to one or another of Hinman's explanations of such anomalies (Hinman, I, 126-138).

The distinctive types also establish the place of Q2 Lr. in the whole series of Pavier quartos. When William Neidig of the University of Wisconsin proved in 1910 that the ten Pavier title pages, despite their different dates, had all been printed in 1619, he also established their order through the press: The Whole Contention (two plays), A Yorkshire Tragedy, Pericles, The Merchant of Venice, The Merry Wives of Windsor, King Lear, Henry V, and Sir John Oldcastle; the place of A Midsummer Night's Dream was left in doubt. It has generally been accepted that the plays themselves were printed in this same order.[14] On the evidence of watermarks Peter Blayney has recently placed MND just prior to Lr. (pp. 196-197). Although its title page was printed immediately before that of Lr., the body of Wiv. bears no typographical relation to Q2 Lr., being printed in a larger and different font, while on the other hand a number of distinctive types in the first three sheets of Lr. appear in the last three sheets of MND.[15] The pattern of migration of types from one book to the other is somewhat complicated, and the model of it that I have constructed in Figure 1 will require confirmation from a study of the types in the whole of Q2 MND. For case y at least there is an unmistakable link: a large number of types from MND H(i) reappear in Lr. B(o). The complication is that types from both formes of MND sheets G are found in both case x and case y. That fact might suggest that the printing of MND was interrupted after sheet G was printed and the two cases of type used on another job, where they became mixed; but the fact that only one sheet of MND remained to be printed and that forme E(i) seems to have remained as standing type until the printing of that last sheet seems to argue against an interruption at this point. It is also possible that F(i) and G(i) should reverse places with their forme-mates F(o) and G(o) in the chart, as having been set at the opposite cases; but I think the patterns of distribution suggest otherwise. I have assumed that most of G(i) was originally set from and mostly distributed back into case x because about twice as many recognizable types from MND G(i) reappear in Lr. A(o) (set at case x) as show up in B(i) (set at case y), and similarly that G(o) was originally set from and mostly distributed back into case y because about three times as many recognizable types from MND G(o) reappear in H(i) (set at case y) as in Lr. A(o) (set at case x). I assume, that is, that most of the types were distributed back into the cases that they originally came from, and that the total number of types distributed into each case will be roughly in proportion to the number of individual types

that can be recognized in each case. Confirming evidence comes from linking types from the formes previously set: types from F(i) show up in A(o) (case x) but not in B(i) (case y); conversely, a number of types from F(o) reappear in H(i) (case y), but only two or three from F(o) reappear in A(o), and I assume that they simply came directly from G(o), where I failed to recognize them, rather than from the sort boxes of case y. Forme H(o) of MND shows no trace of types from MND sheets F and G; since several of its distinctive types are readily recognizable in E(i), E(i) seems for some reason to have been left standing during the setting of F(i) and G(i) and was then finally distributed before the composing of H(o); therefore H(o) was likely to have been set by the same compositor and at the same case (case x) as were F(i) and G(i).

On these facts and assumptions one may tentatively advance a hypothetical explanation of the typographical relations of Q2 MND and Q2 Lr. Since the last three sheets of MND seem to have been set from two type cases

and, as Peter Blayney has shown (p. 198), with two skeletons used for the most part in the same unusual pattern of alternation as is found in Lr., one may safely suppose that the last three sheets of MND were set by two compositors working more or less concurrently; since the type cases are the same for both plays, one may even think it likely that the same two compositors who set the last sheets of MND continued at their cases to begin setting the early sheets of the next Pavier quarto, Q2 Lr. It looks as if Compositor 1, after setting MND G(i), distributes the long-standing type pages from E(i) and begins to set H(o) sometime before his fellow compositor at case y begins to set H(i), the last forme for this play. He finishes while Compositor 2 is distributing types from G(o) in order to set H(i). Apparently Compositor 1 begins work right away on the first forme of the next play, A(o) of Lr., before G(i) has been unlocked. Since both G(i) and H(o) are locked up, and since one of Jaggard's type cases, as Hinman has shown (I, 96, n. 1) held about 16,000 ens, or less than three quarto formes, Compositor 1 found himself somewhat short of type, and "poached" (to use Hinman's word [I, 133]) some types from the forme, G(o), that his mate was presently distributing, very probably from the same distribution frame or bench.[16] He may have begun setting Lr. A(o) then, but before completing it he distributed most of the now available G(i), whose types appear only in the last two pages (3 and 4v) of A(o). Meanwhile Compositor 2 has finished MND H(i) and is about to begin Lr. B(o), and, although only two or three rather doubtful types from MND G(i) seem to appear in Lr. B(o), I think it likely that at this point Compositor 2 poaches back from G(i), then being distributed, a quantity of type roughly equal to that recently borrowed by Compositor 1; the reason so little shows up in B(o) is that it is buried by types soon distributed from H(i). It begins to show up in the first page of the next forme, B1v, which Compositor 2 apparently set before beginning to distribute any types from B(o). The precise details of these speculations will have to await further confirmation from the study of types throughout the whole of Q2 MND, but I think that there is already sufficient evidence from types to support Peter Blayney's assignment of MND just prior to Lr. in the Pavier series.

As I have already suggested above, the evidence from types in Q2 Lr. indicates that the composition of that play was at first divided between two typesetters, each working at his own case, until at some point in sheet I or K the work was assumed and completed by one compositor at one case. Although not proven absolutely from the evidence of types alone, this hypothesis is far more likely than the alternative supposition of Kable and Andrews that one compositor working alone set the first eight sheets, for it is hard to imagine why throughout most of a book one compositor should shuttle from one case to another as he set each new sheet or forme. Such a method of working is contrary to the normal procedures in Jaggard's shop during the printing of the Folio, if not indeed hitherto unknown in any shop. It may be barely

conceivable that one compositor completely set sheets A through H from alternate cases while a companion compositor, perhaps an apprentice, distributed types from the wrought-off formes to replenish the other case just used to set the previous sheet or forme; but such an arrangement seems very unlikely. Hinman has shown that the normal pattern for Jaggard's compositors when they set the Folio was for the same compositor to work at the same case. The clearest example that he offers of the exceptional instance when a compositor moves to another case to allow a fellow compositor to distribute at the first case has since been discredited. Hinman thought that Compositor B moved to case x to set Timon Gg3 while Compositor E distributed into case y, but Trevor Howard-Hill has recently demonstrated that Gg3 was set by E, not by B.[17] Since composition, requiring justification and constant reference to copy, would take much more time than distribution, a division of labor between them would not have been an efficient arrangement throughout the printing of a whole book (unless perhaps we may imagine the apprentice regularly doing some other job as well). The much more likely and normal method of work to be inferred from the evidence of types is that the composition of sheets A through H was divided equally between Compositor 1 working at case x and Compositor 2 working at case y.

Furthermore, the distribution patterns and sequence of formes in the first four sheets seem clearly to indicate two men rather than one. Between the printing of the white-paper and perfecting formes of sheets A, B, and D there was clearly a period of time long enough for the white-paper forme to be machined and for its types to be washed, unlocked, and distributed before the next forme was set at the same case. If during these interims the supposed single compositor continued to work on Q2 Lr. rather than on another book, he would in all probability have begun work on the perfecting forme of the same sheet. What we find instead, if we accept the premise that one compositor is shuttling from case to case, is a sequence of formes that makes no sense: A(o)—B(o)—A(i)—B(i)—C(o)—D(o)—C(i)—D(i). Such a sequence would double the number of sheets being worked at one time and multiply the possibilities for confusion, delay, and error without any evident compensations. The hypothesis of one compositor working alternately at two cases would imply, for Q2 Lr., a pattern of composition that is unaccountable, disadvantageous, and hitherto unknown. Far preferable is the conclusion more obviously indicated by the evidence, that in setting the first four and probably the first eight sheets two compositors worked concurrently, each following the normal practice of working at his own case and distributing his own types, and each, in sheets A through D, setting the perfecting forme for his own sheet.

Once each of these two compositors had set two initial sheets and had disposed of all the standing type from MND, they seem to have settled down

to a firm pattern of collaboration on the next four sheets. Compositor 1 at case x set the white-paper forme of each sheet, and Compositor 2 at case y set the perfecting forme. The only aberration is signalled by the appearance of some types from case x in forme G(i). These seem mainly to come from F1 and F4v and to appear mainly in G2 and G3v. Two explanations suggest themselves: Compositor 2 poached some types from the just-unlocked forme F(o), perhaps because his case seemed a little low, or Compositor 1 set a page or two of G(i) at case x to help out his fellow. The admittedly sparse evidence seems to rule out the latter possibility: the presence in all four pages of G(i) of a few linking types from E(i) indicates that those pages were all set at case y; and the absence from G(i) of unmistakable linking types from case x (from sheets A and C) makes it improbable that any parts of the pages of G(i) were set at case x. For some reason Compositor 2 at case y began to set G(i) before distributing the previous forme. Since he had always, prior to this, distributed the previous forme first (unlike his fellow at case x, who twice had set new formes, C(i) and F(o), without first distributing the prior ones), and since Compositor 2 earlier had apparently been prompt to reclaim some type from MND G(i) to replace what his fellow had poached from MND G(o), it is possible that case y had a somewhat smaller stock of types than case x. If so, that would explain why, with F(i) of Lr. still locked up, Compositor 2 now found it advisable to appropriate some type from a couple of standing pages of F(o); and it may have been partly to repay this borrowed type that later on at least some of the type from G(i) was distributed into case x.

Two generalizations can be made about the work of the two compositors in the first eight sheets of Q2 Lr. The first is that they seem to be working together: for the first four sheets they work in parallel, each setting both formes of two sheets at his own case and using his own skeleton; then for the next four sheets they work in close collaboration, one man setting the white-paper and the other man the perfecting forme of each sheet, sharing both skeletons between them; and then finally, as we shall presently see, they are both taken off the job together. In the transition between MND and Lr., a quantity of types borrowed by one man is immediately reclaimed by the other. Although D. F. McKenzie has forcefully taught us to doubt that apparently regular patterns in a book's physical features necessarily imply a regular method of work,[18] such parallelism as we find here is strong temptation to invoke Mill's inductive method of concomitant variation. When each

compositor's variation in working method is mirrored by the second compositor, one may reasonably suspect a causal relation, such as a close and perhaps continuous collaboration between the two men. The second generalization to be made about the work of these two men is that it was obviously not confined to Lr. at this time. The fact that Compositor 2 consistently and Compositor 1 more than half the time distribute the previous forme before setting another one implies a regular delay of at least half a day of press time between the completion of composition of one forme and the beginning of the next. Hinman estimated that the setting of one Folio page (which at 132 lines is roughly comparable to the 148 lines in a forme of Q2 Lr.) would take about a half a day, and the presswork to print 1200 perfected sheets would take a whole day (I, 44-45), or a half-day for one forme. During that half-day between setting a forme and beginning to distribute its types back into the case to set the next forme, each compositor must have been employed at other jobs. A tempting suggestion is that one or both were employed in setting the Pavier edition of Wiv., which was composed at about this time in a different type face.

At sheet I the typesetting of Q2 Lr. is taken over by a third compositor and becomes a one-man operation. That this compositor is a different person from either of the two previous compositors is a certainty. Madeleine Doran (pp. 36-38) noted fifty years ago that from sheet I(o) to the end of Q2 Lr. there is a marked change to more carefully lined verse, more freely emended readings, more careless mistakes, heavier punctuation, and a sudden change from the spellings Glo(u)cester and Glost. to Gloster and Glo. John Andrews's studies of prose and verse alignment, substantive and metrical alterations, punctuation, and preferred spellings have further confirmed that the habits of the compositor of the last three sheets of Q2 Lr. are markedly unlike those evident in the first eight sheets. I have applied one further test, the frequency of spaced commas in short lines, introduced by Trevor Howard-Hill (op. cit., 1973). The average percentage of spaced commas in the first eight sheets is 9%; in the last three sheets the average is 41.5%, a significant difference.

This new compositor phases out case y by distributing its two standing formes of type and setting only one more forme at that case. After setting I(o) at case x he distributes G(i), from case y, into case x and uses its skeleton to impose I(o). He appears also to have unlocked and distributed parts of three pages of H(i) at about this time, but since so few of its types—only four that I can identify with any degree of certainty—reappear in I(i), and since its skeleton is not used until after K(o) is set, H(i) seems to have passed mainly into the hands of another compositor, perhaps distributing it into another case for use on another job. Compositor 3 abandons case y after setting I(i) there, whose types and Skeleton I disappear from Q2 Lr. after its printing. Linking types from case x appear in all of the remaining formes, making it clear that Compositor 3 is working at that case from now on.

I do not automatically assume that because K(o) was printed before K(i)

it was also set first, but in fact any setting sequence other than K(o)—K(i)—L(o)—L(i) implies the disappearance of clusters of types from the first forme until their reappearance in the fourth forme. The one difficulty with this sequence is a single anomalous type, a capital T from K4.27 that immediately reappears in L3.18 without the prior distribution of K(i). Since the type begins a line in K(i), I assume that its presence in L(o) can be explained as the result of the kind of stripping accident described by Hinman (I, 128-130) in which a peripheral type displaced in stripping is tossed into the case and reappears in the next forme to be set. Some slight crowding in pages L3v and L4 may be evidence that L(o), with its blank page L4v, may already have been completed, but not necessarily. In L3v both stretching and crowding go on simultaneously: one part-line of verse is added by relining Q1 verse, and four speeches that in Q1 share a line with another speech are given separate lines, while at the same time the customary line of space before an entrance is omitted, six lines of verse are reduced to four-and-a-fraction lines of false prose, and two speeches stay doubled with another on the same line. In L4 three speeches doubled in Q1 are given separate lines, but another speech is left doubled and an entrance direction is not centered and spaced but is left flush right and doubled with another short speech in the same line. This slight crowding in L(i) does not necessarily mean, however, that L(o) had already been printed: in order that Q2 should pass for Q1 under casual inspection, Pavier or Jaggard may have intended to keep L4v blank even if it turned out to be "printed" last.

As well as I can determine from microfilm, types from both K(o) and L(o) appear in sheet A of Henry V, apparently the next quarto printed in the Pavier series. I had expected that types from case y would reappear in sheet B of H5, but it seems rather to contain a mixture of types from both case x and case y as well as a large number of new and unfamiliar types. Perhaps case y had been involved in another job in the meantime or had been replenished from some standing type. How the case x types got there must be answered by study of a live copy of H5; a likely place would be sheet L(i) of Lr.

The pattern of printing and distribution described above clearly implies, contrary to all previous views, that two men set sheets A through H and a single man set sheets I through L. One large question remaining is whether either of the compositors of the first eight sheets is the man who set the last three. To test this possibility I reclassified all of the data in John Andrews's dissertation about lineation, substantive and metrical alteration, and spelling in the first eight sheets, dividing it between the two men implied by the pattern of types. The result is a picture of two compositors so similar in their habits that much further study will be needed to distinguish one confidently from the other. The composite picture of both of them that Andrews gives in describing a hypothetical "Compositor G" applies very closely to either of them and distinguishes them both from the third compositor of sheets I to L. My own test of spaced periods implies the same thing: Compositors 1 and 2

of Lr. (and incidentally of the last three sheets of MND) have very similar spacing habits, markedly different from those of Compositor 3.

The main consequence of the foregoing analysis, then, is that all previous opinions of the compositors of Q2 Lr. will have to be revised. Kable's theory of one compositor is clearly untenable, Blayney's conjecture of two men working throughout is closer to the mark, and Andrews's distinction of a new compositor in sheets I to L holds up while his Compositor G turns out to be a composite of two men. Who these compositors were is a question that needs further study. Andrews's and Kable's identification of (at least) one of them as Folio Compositor B has been called into serious question by S. W. Reid: "The spelling practices of B in F1 are like those of Andrews' 'G' as much as they are like those of his 'B,' and B's spellings are more often like the spellings of both, combined, than they are like those of either."[19]

Causing still further doubts is Blayney's series of caveats about compositor identification: that two or more apprentices trained by the same printer or schooled by the same master might show almost identical spelling habits; that a compositor's habits in 1619 might not be those of 1622; that our knowledge that Jaggard bound no apprentices between 1614 and 1622[20] helps us not at all to know how many of the four apprentices that he freed (in 1607, 1614, 1617, and 1619) stayed with him, how many men he employed who were apprenticed elsewhere, or how many men he might sometimes have hired temporarily (as for such a project as the Pavier series of quartos). Blayney concludes, "The assumption that a Jaggard book of 1619 necessarily contains a single type set by any member of the Folio team" is "unsound" (p. 182). One further specific reason he gives for doubting that Compositor B had a hand in Q2 Lr. is that there the name of Lear's counterpart is spelled Gloster or Glocester, whereas in Compositor B's share of the Folio Lr. Glocester is never used, Gloster appears only five times, while in 21 other occurrences the name contains a u (Glouster 14, Gloucester 7).[21] The spaced comma test also casts some doubt: the highest percentage of spaced commas set by any compositor in Q2 Lr. is 41.5%, but Howard-Hill attests that Compositor B regularly prefers a space after commas and that in his work unspaced commas are relatively rare (op. cit., 1973, p. 67). The whole question of the Pavier quarto compositors needs further work; and if Q2 Lr. is any measure, that work will have to include the slow and laborious task of types analysis to determine whether and when two compositors of very similar habits are at work simultaneously.

One more problem remains to be touched on: the pattern of skeleton alternation in Q2 Lr., in which the same skeleton appears in both formes of

alternate sheets. To begin with, it is not a usual pattern either for a single compositor (Q2 Lr. Compositor 3 discontinues it when he is working alone) or for two men working together, as is the case in Lr. sheets A to H. The usual pattern would be for one forme to go into one skeleton and for the perfecting forme to go into the other. Yet in setting sheets A to D each man seems to hang onto his own skeleton from one forme to the next, and then in sheets E to H, when Compositor 1 is setting all the outer formes and Compositor 2 all the inners and we would expect to find the normal pattern of one skeleton associated with outer formes and the other with inners, we find instead that both skeletons are passed back and forth alternately between the compositors. An order of printing such as    

Skeleton	I	I	II	II	I	I
Forme	E(o)	E(i)	F(o)	F(i)	G(o)	G(i)

is perfectly possible, given that the half-day delay between formes that I have noted above would have provided plenty of time for Compositor 1 to retrieve the wrought-off and rinsed forme E(o), unlock it, and hand skeleton I over to Compositor 2 who, working in an alternating pattern, would subsequently use it to impose the perfecting forme E(i). The question yet remains why, having apparently held on to their own formes when setting A to D, the compositors should now pass them back and forth, and even why such a pattern of sharing required more than one skeleton. The alternative order    

Skeleton	I	II	I	II	etc.
Forme	E(o)	F(o)	E(i)	F(i)

makes even less sense if E(o) and E(i), F(o) and F(i), etc. are being set simultaneously. And yet this unusual pattern of skeleton alternation from one sheet to the next, which appears in nine of the ten Pavier quartos, continues unchanged in Q2 Lr. even though three different methods of printing (in A-D, E-H, and I-L) succeed one another and though each of the three compositors at one time or another uses each of the skeletons.

The distribution evidence in Fig. 1 suggests the most likely explanation: that by the time Compositor 1 came to impose F(o), he had pulled ahead of Compositor 2 by about a forme. He could easily have done so, since his formes A(o) and A(i) together contain only five full pages, a short first page, and a title page as against the eight full pages of sheet B set by Compositor 2, and since he could have set C(i) or F(o) or both immediately after imposing the previous formes whereas Compositor 2 always waited for the previous forme to come back from the press before composing the next one. The distribution evidence shows that no types from E(o) appear in F(o), and therefore that E(o) may still have been locked up in Skeleton I. Skeleton II, however, was now or would soon be vacant, since Compositor 2 distributed D(i) before setting E(i). Compositor 1 would then quite naturally have taken Skeleton II to impose the pages of F(o); and we know that he distributed E(o) before setting G(o), thus freeing Skeleton I for Compositor 2 to claim

for the imposing of E(i). Thus would begin the swapping of skeletons that continues throughout the rest of the collaboration of the two men; and the fact that they continue to swap them, rather than lapse back into their former practice of holding on to the same skeleton, can be explained also by the distribution evidence.

What results is a pattern that is regular but complicated because of the need to adapt an order of setting to a different order of printing of the formes. Assuming that Compositor 1 stayed about one forme ahead of his mate, the order of setting would be as follows:

The only modification that the distribution pattern suggests is that, since Compositor 1 distributes all or most of G(o) and most of F(o) before setting H(o) while Compositor 2 breaks pattern to begin setting his next forme, G(i), with only minimal distribution from F(o), G(i) might possibly have been set abreast of or even a little ahead of H(o). In any case the order of printing the formes would have been the regular, and different, one: E(o)—E(i)—F(o)—F(i) and so on. When the necessary adjustments are made between these two sequences, the skeletons migrate naturally to alternate sheets and from forme to forme-mate:

The intermediary steps are as follows:

• F(o) is imposed in vacant Skeleton II. Compositor 1 distributes E(o), freeing Skeleton I, and begins to set G(o).
• E(i) is completed and imposed in Skeleton I. It is printed before the already imposed F(o) in order to perfect sheet E before a new sheet is begun.
• F(o) is now printed. Compositor 2 distributes E(i), freeing Skeleton I, and begins to set F(i).
• G(o) is completed and imposed in vacant Skeleton I.
• F(i) is completed, and must be printed before the already imposed G(o) in order to perfect sheet F before a new sheet is begun. In order to vacate Skeleton II, both compositors strip and distribute F(o), sharing its types. F(i) is imposed in Skeleton II and is printed, while Compositor 2 begins to set G(i).
• G(o) is printed, and Compositor 1 strips and distributes its four pages before setting pages 3 and 4v of H(o), thus freeing Skeleton I.
• G(i) is completed and must be printed next whether or not H(o) has been completely set. Evidently Skeleton I is still available to impose G(i). Compositor 2 unlocks and begins to distribute F(i), vacating Skeleton II. H(o) is imposed in Skeleton II and printed, and Compositor 2 begins to set H(i).

This ends the collaboration of Compositors 1 and 2. At about this point

Compositor 3 takes over. Either he or Compositor 1 distributes H(o), vacating Skeleton II for the imposition and printing of the perfecting forme H(i). Compositor 3 sets I(o) and strips Skeleton I from G(i) in order to impose I(o); he then distributes G(i) to replenish case x. Evidence that H(i) is unlocked and distributed at this point is slight and doubtful, and I have speculated that it may have been in the hands of another compositor; in any case neither the majority of its types nor its skeleton is used in setting and printing I(i). When I(o) returns from press its Skeleton I is stripped to impose I(i) and send it to press. At least by the time Compositor 3 finishes K(o), Skeleton II is freed from H(i), and he uses it to impose K(o). Henceforth he works only at case x, retrieving the same types and using only Skeleton II; Skeleton I and case y types do not reappear in Q2 Lr. He sets K(i), retrieves K(o) from the press and distributes it to free its skeleton and send K(i) to the press; sets L(o), retrieves and distributes K(i); and finally sets L(i).

This explanation of the pattern of skeleton sharing in Q2 Lr. is consistent with all of the facts of the distribution of types. Though it may seem complex when presented in all its detail, it all follows logically from the simple premise that Compositor 1 got ahead of Compositor 2 fairly early on. Thereafter each man keeps borrowing the other's skeleton simply because it is the only one vacant or is the one that can soonest be vacated. The one obvious objection to be raised against such a scheme is that, having finished setting F(o) before his mate had finished E(i), Compositor 1 might logically have proceeded to the next forme that would be printed, F(i), rather than start a new sheet with G(o) and so continue a forme ahead. That is certainly a possibility, but not a necessity; the two men may have simply decided to stick to their agreement to share the setting of each sheet from E onwards, one man setting all the outer formes and the other all the inners, perhaps in the expectation that Compositor 2 would catch up before long.

Nonetheless, it may be prudent to consider two alternative possibilities, both suggested to me by Professor Robert K. Turner. The simpler is that the two skeletons had some difference, such as gutter sticks of different widths, that necessitated each skeleton's backing itself if inner and outer formes were to register. The Newberry copy of Q2 Lr. is too tightly bound to test this theory with certainty, but a thin slip inserted down to the fold between conjugate pages of several sheets reveals no apparent differences in the gutter widths of the two skeletons.[22]

The other possibility is that Q2 Lr. was printed simultaneously on two presses. It is in fact possible to work out a pattern whereby two presses might have printed this book, one press printing all the outer formes and the other all the inners. Despite the changes in the number of compositors and in the ways they shared the work, if press 1 is continually one forme ahead of press 2, the two skeletons will alternate back and forth between the two presses with absolute regularity. One might be tempted, therefore, to think that the

two-press method itself might be solely responsible for the alternating pattern of skeleton use in Q2 Lr. The first four sheets offer no difficulty for this theory, since each compositor, shuttling from one press to the other, keeps retrieving his own skeleton and type. In all the remaining formes printed during their collaboration, however, each man retrieves his own type but, having unlocked and distributed it, then swaps skeletons with his mate. For this odd behavior the two-press theory seems to offer no explanation; it would have made much more sense for the men to hang on to their own skeletons, as they had been doing so far. In that case we would find, in sheets E through H, that one skeleton is always associated with outer formes and the other with inner formes; instead, they alternate. In order to explain the compositors' change of method of working we have to fall back on a theory such as one man's getting ahead of the other and imposing his pages in his fellow's skeleton because it was readily available and his own was not—exactly as we have already been able to do without resort to a theory of two presses. Two-press printing of Q2 Lr. remains a possibility, of course, but it is neither sufficient nor necessary as an explanation of the bibliographical facts. It may also imply a degree of regularity in method more ideal than practical in a busy shop like Jaggard's, which might handle a variety of overlapping jobs at once. Still, if Pavier and Jaggard had intended from the outset to prepare a collected works of Shakespeare, as seems likely, they might well have organized such a regular pattern of work to be adhered to from the beginning of the series to the end. It may at least be a hypothesis worth testing as further work is done on the other Pavier editions.

Notes