University of Virginia Library

Although nine editions of Beaumont and Fletcher's The Maid's Tragedy appeared before 1700, critical attention is centered on the first two. Q1, which was printed in 1619 by Nicholas Okes for Francis Constable and Richard Higgenbotham, has been labelled "curtailed and frequently corrupt."[1] It is about eighty lines shorter than subsequent editions; yet it contains a few lines that do not appear elsewhere. Q2, printed in 1622 by George Purslowe for Constable alone, supplies the lines which are not in Q1, and, in addition, makes a number of other changes in single words and phrases throughout the play. The texts of all subsequent editions derive from Q2.[2]

All of the editors of The Maid's Tragedy, from Dyce to Hazelton Spencer, have based their editions on Q2 while occasionally introducing readings from Q1 on eclectic principles. But, as I have argued elsewhere,[3] such a procedure is not acceptable for an old-spelling critical edition, since it can be shown that Q2 was printed from an annotated copy of Q1.[4] Therefore, to the Beaumont and Fletcher critic the printing of Q1 is a matter of some concern. In addition, the shop which produced Q1 also printed within a period of a few years Othello Q1 (1622), The Duchess of Malfi Q1 (1623), and those monuments to editorial frustration, Philaster Q1 (1620) and Q2 (1622). It is hoped, then, that the insight into Okes's operations provided by this study will be of interest to students of those plays.

Of The Maid's Tragedy Q1, Greg notes, "The text was printed in two sections, B-G and H-L, in slightly different types: in the first the speakers'

names are set in small-caps, in the second in italic, but they are not indented in either. This might suggest that the copy was divided between two compositors, and that there had been an earlier edition. It is, however, more likely that composition was interrupted, and that on resumption the original type was not available, or possibly that the work was completed at another press."[5] Since the compilation of the Bibliography, many investigators have found that copy can be divided between two compositors, and as a result of this division different parts of a printed book can exhibit different characteristics, without there having been an earlier edition. No edition of The Maid's Tragedy seems to have appeared before 1619 and it is not necessary to assume that there was one to explain the differences between the two sections. That the second section of the book was printed simultaneously with the first is a possibility that must be considered; however, if it turns out that this was the case, Okes could not then be the printer of the second section because it is very unlikely that he had more than one press in 1619.[6]

The two sections are clearly set apart from each other not only by the differences in the typography of the speech-prefixes but also by a variation in the speech-prefix abbreviations: the tag for Calianax is predominantly Cal. in the first section and Call. in the last and that for Aspatia is invariably Asp. in the first section and Aspat. in the last.

In addition, there is a clear break in the running titles between the two parts: the skeleton forme (I) that was used to impose B(i) was also used for C(o), D(i), E(o), F(i), and G(o), whereas a second skeleton forme (II) was used for B(o), C(i), D(o), E(i), F(o) turned, and G(i). At sheet H, however, two new skeleton formes (III and IV) were constructed and were used as follows: III H(i), I(o), K(i), and L(i); and IV H(o), I(i) turned, K(o), and L(o). The only irregularity in the transfer of the formes from sheet to sheet occurs when skeleton I was moved from D(i) to E(o): here the same running-titles that were used in D(i) reappear but their arrangement in the forme has been changed.

Additional information about the printing of the book can be obtained from typographical evidence, derived principally from a shortage of the small capitals used in the first section for the setting of the speech-prefixes and an occasional stage direction. Okes ordinarily reserved this font for such purposes as the composition of chapter headings, subordinate lines in title-pages, and proper names in stage directions, and in these instances the supply of type was adequate for the demand placed upon it. But when the small-capital font was used for speech-prefixes, as in The Maid's Tragedy Q1, Albumazar Q2 (1615), and Lingua Q1 (1617), the frequency of occurrence of identical speech tags was so great that certain sorts ran low. In

order to make up these deficiencies, letter from a smaller font of small capitals was introduced in all three plays. In Albumazar Q2 there is a mixture of l's, a's, r's, and e's, whereas in Lingua Q1 a mixture of e's and m's is found. In Albumazar Q2 the substitution of the small for the large small capitals was systematic enough in some instances to permit an insight into the method of composition.[6a] Similarly, the pattern of shortages in The Maid's Tragedy Q1 reveals a good deal about the way in which its compositors worked.

Before this matter is examined further, the effect of type shortages on quarto printing practices should be clarified. Professor Bowers has made the point that the supply of type available to the compositor could dictate the choice between one- or two-skeleton printing.[7] When the latter method was employed, and when proof was to be obtained, it was necessary under normal conditions to have a minimum of fifteen type pages standing at any time during the course of printing (four on the press, four on the distributing bench, and seven on the imposing stone), but when the former was used the minimum number of type pages standing could be reduced to eleven (four on the press and seven on the stone). To a compositor setting from a low case, the one-skeleton method would be attractive for this reason, but in one-skeleton work the press was forced to stand idle not only while corrections were made in the type of the forme being machined, but also while any forme was being imposed. When two skeletons were used, however, one was always free to impose the next forme to be machined, and the interval occasioned by the correction of the forme being machined could be used to proof the next forme.[8] When proof was to be taken, two-skeleton printing was therefore more efficient and more desirable than one-skeleton, if the supply of type was adequate to permit its use.

Professor Bowers' count of fifteen standing pages in two-skeleton proofing supposes that the compositor would not ordinarily distribute type from a wrought-off and rinsed forme before he had a new forme ready for the press, that is, until type page $4 had been set. Indeed, in order for any advantage to be gained in proofing by two skeletons rather than by one, the perfecting forme of a new sheet had to be imposed by the time the corrected proof of the first forme had been returned. But if the supply of type was insufficient to allow fifteen pages to be standing, special measures had to be adopted. Professor Bowers indicates that the compositor of Lear Q1, when faced with a type-shortage problem, stripped the quarters of the wrought-off forme before composition of the next sheet was completed, imposed the old furniture and running-titles about the new type, and then

distributed the type from the stripped quarters of the wrought-off forme. If this method was used and the composition was in perfect balance with the presswork, there would be four pages of the new sheet ($1, 1v, 2, and 2v) on the stone at the end of the machining of the first forme of the preceding sheet; therefore, it would be possible to reduce the number of standing pages to twelve.[9]

However, there was a further refinement of the two-skeleton method which enabled the printer to reduce the minimum number of standing pages to this number while retaining all of the advantages of two-skeleton printing. Professor Bowers' analysis was based on the assumption that the compositor would set seriatim, but, if the copy was cast off so that pages could be set out of order, the pages of the new sheet on the stone at the end of the machining of the first forme of the old could be either the inner or the outer forme of the new sheet ($1, 2v, 3, 4v or $1v, 2, 3v, 4) rather than the first four pages. In this case the pages on the stone would correspond in their position in the forme to the pages of the wrought-off forme, and all four quarters of the wrought-off forme could be stripped and the type distributed immediately. The standing-type pages could in this manner be reduced, before distributing, to a minimum of twelve, and the amount of type in the boxes could be kept at a maximum.

Of course, before the compositor could set by formes it was necessary that his copy be cast off with some accuracy.[10] Moxon described two methods for casting off manuscript,[11] but implied that neither permitted very accurate fitting of the copy, particularly if the manuscript was in an irregular hand. In addition, both methods were complicated and must have required a considerable amount of time to perform for a book of any size. However, it is obvious that Moxon described the casting off of prose; the entire procedure must have been much less complex when the copy consisted substantially of verse lines. Once the copy was cast off, the compositor could fall to setting at any convenient point, and one imagines that this fact alone would have been enough to recommend setting by formes in instances where there was no question of a type shortage. Not only could the first forme to be machined be delivered to the press at the earliest possible moment, but also the relationship between the speed of composition and the speed of the press became more flexible since an alert workman

could take advantage of a choice between setting more or less difficult matter if such a choice existed.[12]

The compositor of the first section of The Maid's Tragedy Q1, perhaps influenced by a shortage of the small capitals chosen for setting the speech-prefixes as well as of certain sorts in the roman font used for the text of the play, apparently elected to set by formes. In order to show that this method was adopted, it is first necessary to consider the difficulties that would have arisen if the book had been composed seriatim.

The text of the play begins on sig. B1. In sheets B and C the following distribution of small capitals is found (the number to the left of the slash represents capitals of the larger size; that to the right the smaller):

B

     

	1	1v	2	2v	3	3v	4	4v
e	4/0	3/1	6/0	3/0	3/1	2/1	0/7	0/3
l	4/0	4/0	5/1	3/0	4/0	5/3	2/10	0/9

C

     

	1	1v	2	2v	3	3v	4	4v
e	0/1	2/0	6/0	0/0	2/0	1/0	0/0	0/0
l	0/2	0/0	4/1	1/0	0/2	6/4	0/8	0/2

This pattern may have resulted either from seriatim setting or from setting by formes. If seriatim setting is posited, it must be argued that the one small e on B1v and the small l on B2 were used accidentally rather than deliberately substituted. Then, while setting B3 the compositor noticed that his supply of large e's was running low, and he attempted to eke them out by the occasional substitution of small e's in B3 and B3v. However, his stock of the large letter was gone when he began B4, so that B4, B4v, and C1 were composed with the small e's exclusively. About the same situation obtained for the l's. The substitution of small for large letters commenced on B3v, but by the time half of B4 was set all of the large type was used. Small l's then were employed exclusively from about the middle of B4 through C1. At C1v, however, large e's reappear and no more small e's appear in sheet C. Large l's reappear at C2 (there having been no occurrence

of l on C1v) and throughout the rest of the sheet show a pattern of depletion similar to that of sheet B. It appears that between the setting of C1 and C1v a new supply of large e's and l's was obtained, and it seems possible that the source of these letters was the first forme of sheet B to be sent to the press.

And indeed several pieces of type which were used in B(o) reappear in C: the "S" of "Scœn." (B1) is found in "Secundus" (C3v); the "m" of "time" (B1, l. 21) is found in "Madame" (C4v, l. 9); and the "y" of "your" (B2v, l. 17) is found in "my" (C3, l. 15). The forme distributed after the setting of C1 must, then, have been B(o), which, it follows, would have been delivered to the press before B(i). However, if the outer were the first form of B to be printed, problems arise in connection with the hypothetical seriatim setting of B unless a long delay is posited between the setting of B4v and the time of distribution of B(o). Under ideal conditions a compositor should have been able to set and distribute about four type pages in the same amount of time required by the press to machine one forme, but in this case the interval that elapsed during the machining of B(o) would have been occupied only by the setting of C1.

Such a long press delay (the time required for the compositor to set three pages) was, of course, possible, but another hypothesis which accounts for the evidence while eliminating the delay would be preferable to one including it. It may also be argued that sheet B was set by formes, in which case the pattern of small caps is

B

     

	1	2v	3	4v	1v	2	3v	4
e	4/0	3/0	3/1	0/3	3/1	6/0	2/1	0/7
l	4/0	3/0	4/0	0/9	4/0	5/1	5/3	2/10

Here it must be assumed that the type used in another book[13] was distributed between the setting of B4v and B1v. If this assumption is allowed, it is seen that a clear pattern of depletion and resupply of the large letter emerges.

It should be mentioned that there is no clear line of division between the use of the large type and the small on pages where the two are mixed. On B3, for instance, the small e is the next to last of the four e's used on the page, and on B3v the small l's are numbers three, seven, and eight of the eight l's appearing. It has been noted earlier that the two sizes do not appear mixed except in the three plays where the demand for particular sorts was abnormally high; the large font appears unfouled in several books

produced in the years between the printing of the plays. Moreover, if the sizes were kept mixed in a single case one would have expected to find at least an occasional small r in the speech prefix Stra. of The Maid's Tragedy since small r's had appeared previously in Albumazar Q2; but Stra. appears with large r's throughout. One concludes that after the earlier plays were printed the two sizes were sorted out and returned to their own cases. When the small type was needed once more for The Maid's Tragedy it was again added to the case of the larger type a few sorts at a time. The more-or-less random mixture of the two sizes on certain pages would indicate only that the compositor did not completely exhaust his supply of large type before replenishing his stock with the small.

Since no other explanation for the distribution of the two sizes of type in sheet B presents itself, it seems that, on the evidence, one must decide in favor of the setting of that sheet by formes rather than seriatim. The method of composition was, of course, intimately related to the order of imposition of the formes and the order of the formes through the press, evidence relating to which is available from the pattern of running titles. If B(o) were the first forme to be machined, the pattern of running titles indicates that the order of imposition of the remaining formes of the first section was

Such a regular alternation between the prior imposition of the outer and then the inner formes of succeeding sheets can hardly have been fortuitous, and, since there was a shortage of type, one can be reasonably sure that the order of imposition was also the order of composition. If the first sheet were set by formes, it is reasonable to suppose that there was some connection between this procedure and the rather curious order of composition of the remaining formes.

It must be remembered, first of all, that the printing of a play was a commercial venture in which there was probably a fairly narrow margin between profit and loss, and that one of the major items of expense in such a job was the paper required. If one edition-sheet could represent the difference between making money and losing money, we would expect the master printer or compositor to cast off the copy with some care so that the exact amount of paper needed for the book could be purchased or otherwise obtained. Only in this way could the risk of wasting paper be minimized.[14] One supposes that the whole book would have to be cast off for this purpose; therefore, when typesetting began, it is probable that a typical manuscript was marked in such a way as to show what portion of the handwriting was to "get in" a specific page.[15] When the casting off had been done properly,

it then became the compositor's responsibility to make sure that the typeset text stayed within the estimated total number of sheets. The marks on the manuscript would serve to warn him at the end of every page of the degree of conformity between his actual and estimated progress. If he was running over or under, it behooved him to conserve or waste space in some way so that the actual space consumed would be brought back into line with the estimate.

If the book was being set by formes, however, special problems would arise. Assume that composition began with the outer forme of the first sheet. The text of sig. $1 of the outer forme would, of course, have to adjoin the text of sig. $1v of the inner forme, and sigs. $2v and $3 would have to adjoin respectively the text of sigs. $2 and $3v of the inner forme. Sig. $4v, in its turn, would adjoin sig. $4. By setting in the order $1, 2v, 3, 4v the compositor established very definite limits for the beginning and ending of the pages of the inner forme. Later, in setting consecutively from the top of $1v to the bottom of $2 and from the top of $3v to the bottom of $4, he was forced to adjust the white space, make double lines of type of single MS lines or vice versa, or take other steps to come out exactly on the line, or in prose passages, the word, dictated by the previously set outer forme pages.

Opportunities to adjust space could be created by a good workman in a number of places, but these opportunities were not limitless. Obviously, if an adjustment had to be made, there was a much better chance of making it smoothly in two pages than one. Therefore, it would have been to the compositor's advantage to arrange, wherever possible, to set consecutively pages the text of which would adjoin one another in the finished book. If the priority of the setting of the inner and the outer formes of succeeding sheets were alternated, every third sheet the compositor would find an additional opportunity to set adjoining pages consecutively (e.g., C4v-D1). The following graph shows how this procedure would work out (adjoining pages that would be set consecutively are underlined):

On the other hand, if the outer (or inner) forme were always set first, this arrangement could be produced: When the priority of the setting of the outer and inner formes is alternated, it is seen (as Professor Bowers has suggested to me) that pages the texts of which adjoin are set consecutively ten times every three sheets; when the outer (or inner) forme is always set first the number of adjacent pages set consecutively is reduced to nine. Thus, in six sheets alternate prior setting of the inner and outer formes of succeeding sheets would produce twenty such opportunities as opposed to eighteen produced by the regular alternation, in nine sheets thirty as opposed to twenty-seven, and so on. For the first section of The Maid's Tragedy, which consisted of only six sheets, the advantage given by this method was slight, but the pattern of the running-titles argues strongly for its use.

If sheet B was set by formes, outer first, and the compositor maintained something near the ideal time relationship with his press, he should have just completed the setting of B(i) when the machining of B(o) was completed, provided of course that B(o) had been put on the press shortly after its imposition. He would have had time, then, to set only four more type pages before B(i) was returned from the press, and, if a press delay was to be avoided, these four type pages should have been the right ones to make up a new forme. In other words, once he started to set by formes the odds are that a compositor would continue to set by formes as long as the press kept pace with him. Hence, one has some reason to suspect that sheet C was also set by formes, and, from the running-title evidence, that C(i) was set before C(o). Thus, the assumptions previously made about the time of distribution of B(o) must be revised. With the summary rearranged by formes, the distribution of e's and l's in sheet C is

C

     

	1v	2	3v	4	1	2v	3	4v
e	2/0	6/0	1/0	0/0	0/1	0/0	2/0	0/0
l	0/0	4/1	6/4	0/8	0/2	1/0	0/2	0/2

If this table represents the correct order of composition, there is evidence, small though it is, for two distributions rather than one as previously supposed. The resumption of the use of large e's on C1v indicates that one distribution was made before the setting of that page; the appearance of a piece of type from B(o) on C3v, mentioned earlier, shows that the forme distributed must have been B(o). This forme contained fourteen e's, ten large and four small, and twenty l's, eleven large and nine small. All indications are that the compositor, seeing no point in keeping the two sizes separated, distributed both into his case on top of the small e's and l's remaining after the setting of B(i). Obviously, as this indiscriminate distribution of the two sizes continued they would tend to become more thoroughly mixed and hence would be of progressively less value as evidence. In sheet C, however, the pattern of depletion and resupply that characterized sheet B has not yet been entirely obscured, and one can see ten of the eleven large l's provided by the distribution of B(o) being used up before the small l's appear exclusively on C4 and C1. Moreover, the shift from small to large l's at C2v (reinforced by the appearance of large e's on C3) suggests that another distribution may have taken place before the setting of C2v.

But would these times of distribution be consistent with the progress

of the press during the time required to set the formes considered so far? As I have said, under ideal conditions the press could machine one forme in about the amount of time required to set one forme and distribute another. If B(o), the first forme composed, were delivered to the press soon after its imposition, it would have been printed and returned to the compositor for distribution by the time he distributed type from the unknown book which preceded The Maid's Tragedy through the press and set B(i). B(o), then, would have been available for distribution before the setting of C(i) began. While B(i) was being machined, the compositor would have been occupied with the distribution of B(o) and the setting of C(i). Under these conditions, B(i) would have come from the press at about the same time as the imposition of C(i), or, if the evidence of the small capitals in sheet C can be trusted, probably after the composition of C1 began. The relationship between the compositor and the press may be graphed:

In sheet D, which is nearly devoid of e's and which required but two l's (which are of the larger size), another type shortage is manifest which supports the method of composition advocated so far. In this instance the shortage appears in the roman capital I's of the text rather than the speech-pre-fixes. The following table shows the pattern of roman I's from the font used in sheets B and C (on the left of the slash and from a smaller font (on the right):

D

   

	1	2v	3	4v	1v	2	3v	4
I (text)	13/0	16/0	9/0	6/9	17/1	8/11	10/1	12/1

On D4v five large I's appear in the first eight lines of letterpress. The first small I's show up in lines 9 and 11 and are followed by one more large I in line 12. From line 12 through the rest of the page small i's appear exclusively. The text of D1v is set entirely with the larger sort; the one small I is found in the catchword. On D2 all of the large I's are clustered in the lines from 14 through 22; before and after that point small I's are found. On D3v and D4 the small I's appear near the bottom of the page.

The pattern of I's suggests that two distributions took place during the composition of D, one between the setting of D4v and D1v and the other between D2 and D3v. However, if this were the case, a long and inexplicable

delay in the presswork would have occurred. Resuming the graph of the relationship between the press and the compositor,

one sees that C(i) would have come off the press at about the time of the imposition of C(o) and would have been succeeded on the press by that forme. Then, if composition and presswork stayed in balance, C(o) would have been machined during the time interval occupied by the distribution of C(i) and the setting of D(o) and should have been ready for distribution before the setting of D1v. Under these conditions the shortage of I's in D(o) can be explained without difficulty since C(i) contained only thirty-one pieces of the large sort, whereas forty-three I's were required to set the text occupying D1, 2v, 3, and the first eight lines of D4v. But since C(o) contained thirty pieces of the large sort and none of the small, it is hard to understand why there is evidence of short supply at the bottom of D1v when that page had required but seventeen pieces. Moreover, if C(o) were distributed between the setting of D4v and D1v, the next forme available for distribution would have been D(o). Had this forme been distributed between the setting of D2 and D3v, D(i) could not have gone to press before at least a part of D(o) had been distributed and D3v and D4 had been composed. All things considered, this is not a very encouraging picture.

It is possible, however, to account for the evidence on other grounds if it is supposed (1) that the distribution of C(o) was delayed until after the setting of the first thirteen lines of D2, and (2) that the small I's were added to the compositor's case in two steps (the first at the time of setting the ninth line of D4v and the second at the time of setting the catchword of D1v) rather than one. In this case the graph would appear:

C(o) should have been off the press and available for distribution at about the time of the imposition of D(o). However, the textual matter contained in D(o) and D1v would probably have presented little difficulty in composition, and indeed a number of short lines are found on the five

pages. At any rate, there is nothing in these pages which would have precluded the possibility of the compositor's gaining on the press to the extent of a page and a third. Probably a page or two of C(o)[16] was distributed as soon as the forme came from the press and during the setting of D2, thereby accounting for the cluster of large I's on that page, and the remaining pages were stripped and disposed of before the setting of D3v. The distribution of C(o) before the setting of D3v is further substantiated by the fact that two pieces of type which had been used in C(o) reappear on D3v: the "m" of "Madame" (C4v, l. 9) is found in "me" (D3v, l.9) and the "h" of "lighten" (C3, l. 24) is found in "hot" (D3v, l. 18).[17]

After sheet D the e's, l's, and m's of the speech-prefixes appear in a random mixture, as do the I's of the text except upon one occasion which will be mentioned later. However, in sheet E several more letters from the small-capital font make their debut. Their arrangement is

E

           

	1v	2	3v	4	1	2v	3	4v
I	3/0	9/0	7/6	4/13	0/0	6/6	1/6	10/1
N	1/0	3/0	9/0	16/0	1/3	7/2	6/0	9/2
P	3/0	6/1	0/5	0/0	3/2	0/3	0/1	0/0
S	3/0	1/0	0/0	0/0	2/3	0/0	0/0	0/0
T	1/0	2/5	0/2	1/0	0/3	0/0	0/1	0/0

Here it is once again possible to see a gradual depletion of the large letter and a resultant increase of the small until type was distributed after E3. However, there is evidence indicating that D(o) was distributed before this point: the "s" of "seeme" (D4v, l. 20) appears in "spare" (E2, l. 32); and the "y" of "you" (D1, l. 23) is found in "my" (E3, l. 25).

If the graph of composition and presswork is continued, it is seen that D(o) was probably distributed either just after E1v was set or perhaps just after the setting of E began, and that it was D(i) which was distributed between E3 and E4v:

From D(o) the following small capitals would have been distributed into the case: twelve i's, eleven n's, three p's, four s's: (three large and one small), and one t. If, as the time-schedule indicates, the distribution of D(o) took place at about the time that E1v was completed, the shortages that appear on E2, E3v, and the remaining pages of the sheet can be explained by the fact that an insufficient number of pieces of the large type was made available by the distribution.

However, these data also indicate that the workman did not wait until the last piece of type was gone before he replenished his stock. The only sorts that should have been in short supply when the composition of E began were i's and n's: thirty-four large i's and thirty-three n's had been used in the composition of D(i) (at this time on the press) and E1v. This was a far greater number of both sorts that required in sheets B and C. But there should have been a fair number of large p's, s's, and t's on hand. Fifteen large p's had been used in sheet C and only three were in D(i) and E1v; fifteen large s's were used in sheet B and again three were standing; and eighteen large t's were used in sheet C and once more three were standing. Nevertheless, small letters seem to have been added to the p, s, and t boxes before the point on E2 where dialogue between Strato and Diphilus (speech prefixes, Stra. and Diph.) begins, probably in an effort to keep the type in these boxes at a convenient level. It is possible that small s's were added to the boxes at the same time.[18]

One other evidence of type shortage in sheet E is worthy of comment. The roman capital of the speech-prefixes was of the same font as the capitals used in the text; therefore, a severe demand was imposed on certain sorts in this font when a number of speech-prefixes beginning with the same letter happened to occur in conjunction with portions of the text that also required that letter to be capitalized. The speech-prefixes and text of D(i) and E1v had required a total of fifty-five roman A's, a larger number than had been used in any previous forme. The distribution of D(o) added twenty-nine A's to the box, but these pieces were quickly consumed, for, as luck would have it, the remainder of sheet E contained dialogue between Aspatia and Antiphila as well as lines belonging to Amintor. Setting the speech prefixes for these characters and capitalizing the requisite text letters seem to have exhausted the supply of roman A's, so that when the compositor reached the bottom of E3 he was forced to set three speech-prefixes as Amin rather than Amin. The substitution of italic for roman A's in speech-prefixes occurs only in this place; therefore, it appears that type must have been distributed, and the supply of A's augmented after the setting of E3. The forme distributed, as it has been shown, was probably D(i).

In sheet F the sizes of the small capitals are indiscriminately mixed in the speech-prefixes and, therefore, provide no evidence for the order of

the composition of the pages. However, one would expect the sheet to be set in the same manner as the preceding sheets:

That this arrangement is at least partially correct is shown by the fact that type from E(i) appears on F3v: the "o" of "no" (E3v, l. 4) is found in "know" (F3v, l.29), and the second "p" of "Suppose" (E1v, l. 15) is found in "friendship" (F3v, l. 30). E(i), then, was distributed at some time before the setting of F3v was completed. That this time was probably between the setting of F4v and F1v is indicated by the pattern of roman and italic A's in the sheet. In the following tabulation the number of roman A's is shown on the left of the slash and the number of italic A's substituted for roman on the right.

F

   

	1	2v	3	4v	1v	2	3v	4
A/A	8/0	3/0	6/4	3/5	9/0	10/0	7/5	9/2

The method of substituting the italic for the roman type was unusual. It will be recalled that on one previous occasion, on E3, the compositor substituted italic for roman in three speech-prefixes, but one judges that this use of the italic was not to his taste. In setting both formes of sheet F he seems to have attempted to save his decreasing store of roman A's for the speech-prefixes by substituting italic for roman in the text only, while there were still roman A's in the box. The resumption of the exclusive use of roman A's on F1v would indicate that the supply of this sort was probably augmented by the distribution of E(i) after the setting of F4v; the fact that italic for roman A's do not appear after F4 would indicate that E(o) was distributed after the setting of that page.

The reason for the shortage of roman A's in F3 and F4v, as well as F3v and F4, is not far to seek. The substitution of italic for roman A's in the speech-prefixes of sheet E took place when there were 116 roman A's in the standing type.[19] When the composition of sheet F was commenced, both E(o), containing thirty-seven roman A's, and E(i), containing forty-two,

were standing. F1 required eight pieces of this type and F2v three; thus the total standing was raised to ninety. It was at this point (on F3) that the compositor first began to stretch his supply of roman A's by the fairly consistent use of italic for roman in the text. Nevertheless, by the time the composition of F(o) was completed, ninety-nine roman A's had been used. The distribution of E(i) provided thirty-seven pieces, thus reducing the number of standing roman A's to sixty-two before the composition of F(i) was begun, but F1v and F2 consumed nineteen pieces. The number standing, then, was eighty-one when the second round of substitution began on F3v. On both occasions when this measure was introduced, composition could have continued with the roman type without completely exhausting the supply, although F(o) would have very nearly drained the box. It seems clear, however, that in neither instance was the compositor willing to risk running so low, probably because he could not have been absolutely sure of the time when the forme on the press would have been available for distribution.

In the next sheet a different problem presents itself. One characteristic which marked the composition of sheets D through F was the appearance of a wrong-font roman capital I of the text. It seems strange, therefore, that the wrong-font I's are found in sheet G on G3 only, in spite of the fact that sheet G required 125 I's, more than any other sheet in the first section of the book. If the compositor ran low enough in sheet D to introduce small I's when seventy-two large I's were standing (twenty-nine in C(o), thirteen in D1, sixteen in D2v, nine in D3, and five in the first few lines of D4v), he also should have run low during the setting of G2.[20]

If a sudden supply of new type is excluded from consideration, these explanations suggest themselves: (1) type from one forme of G was distributed before the setting of the other began, (2) some pages of G were set from a different case from that used for sheets D through F, or (3) some pages of G were set by a different compositor from the one who set D through F — a man who felt strongly enough about the use of wrong-font types to eschew them deliberately even when the stock was low in the box. The first explanation can be rejected without further consideration on the grounds that the press delay involved would have been intolerable; the second can be rejected because type which appears in earlier sheets of the book reappears in sheet G.[21] Only the third seems to have been possible under the circumstances.

Spelling evidence indicates that the third explanation is probably correct. In sheets B through F the word "honor" occurs eleven times, eight times as honour and three times as honour. Honor occurs once in sheet B, once in sheet C, and once in sheet D. In sheet G, however, the word is found eleven times, always as honor, in this pattern:

G

   

	1	1v	2	2v	3	3v	4	4v
honor	1	2	1	2		1	2	2

In addition, in sheets B through F the word "again" appears sixteen times, four times as agen and twelve times as againe. The distribution of the two forms appears to be random. But in sheet G there is this distribution:

G

     

	1	1v	2	2v	3	3v	4	4v
againe			1		1
agen		2	2		1			1

In view of the B—F workman's previous record of againe/agen spellings, the occurrence of both forms on G3 is not surprising; however, the predominance of agen in the other pages of sheet G argues against his presence there. To these words a third may be added, although its occurrence is too infrequent for it to bear much weight as evidence. The word "young" appears four times in sheets B through F as young but on G2 and G3v as yong.

In sheet G, then, at least two preferred forms, honour and againe, are discarded in favor of honor and agen, except on G3 where the word "honor" does not occur and where both spellings of "again" are found. On G2 the combination of the two spellings of "again" might suggest an assignment to the original workman, but the occurrence of honor and yong on the same page makes his hand unlikely. Moreover, it is only on G3 that the wrong-font I's appear in the text. From this evidence it may be inferred that, while G3 was set by the same workman who set sheets B through F, a different workman composed G1v, 2, 3v, 4, 1, 2v, and 4v. Apparently the B-F compositor (whom I designate Compositor S) was called off the job for some reason and was replaced by the G workman (Compositor T) for the composition of the first six pages of the sheet to be set. He then returned to the job to compose G3 and was then relieved once more by his colleague who completed the first section of the book by setting G4v. There is nothing to show why this rather curious method of alternation was adopted.