University of Virginia Library

The White Devil in Nicholas Okes's Shop
by
Antony Hammond

Preamble

Although The White Devil [1] is one of Webster's most frequently-edited plays, very little original bibliographical work has been done on it since the late Philip Williams's article in the first issue of this publication.[2] In this, Williams, whose primary interest was in the composition of King Lear, attempted to compare the work done on the Shakespeare play with that on two other plays also set in Nicholas Okes's shop, one of which was The White Devil. He identified two compositors who were working for Okes at that time; his study of their habits in The White Devil indicated that Compositor A, with certain exceptions, used terminal -y rather than -ie, generally used the apostrophe in the form 'I'le' and was "predisposed" to the form 'do'. B, on the other hand, frequently used -ie rather than -y, never used the apostrophe in 'Ile', and had a "marked preference" for 'doe'. On the basis of this evidence, Williams assigned B1r-v, C1r-F2v, G1r-2v, H24-v, H4r-v, I3r-v, I4v, K1r, K3v-4v, L3r-M2v to Compositor A, and B2r-4v, F3r-4v, G3r-4v,

H1r-v, H3r-v, I1r-2v, I4r, K1v-3r, L1r-2v to Compositor B. John Russell Brown, in his Revels Plays edition, comments that the spellings 'here/heere', terminal -e or -ee and initial -en or -in, confirm Williams's division. But he does not tabulate the evidence,[3] and makes no mention of it in his articles for SB.[4]

Some of Williams's assumptions have been queried,[5] and some of his determinations will be challenged in the present article. However, it should be acknowledged that when Williams did his work, the art of compositor-determination was in its infancy, and any defects of his survey may be attributed to this fact. However, it is high time that the work was done afresh. Williams surveyed only the three spellings indicated, and failed to distinguish between prose and verse.[6] Nor, naturally, could he make any use of the more sophisticated discriminators which subsequent investigators have employed.

The results of such a new survey are embodied in this article, the research for which was undertaken as part of the textual editing of the forthcoming Cambridge edition of the Works of John Webster.[7] I propose to deal with three questions crucial to the composition of The White Devil. First, what is the effect upon compositor-determination of the physical setting of the type? Secondly, do spelling-preferences and other discriminators allow us to identify the compositors with any greater certainty than Williams and Brown achieved? Thirdly, what information can be derived from a detailed examination of the way type was set from case to forme and distributed thence again? The other possible aspects of such study, namely what light the investigation of printing may shed upon the nature of the manuscript used by the compositors, and the proof-reading and press-correction of the play, were examined by Brown in detail; his investigations in this area (in his edition and in the articles mentioned) hold up satisfactorily, and although they will be addressed in the Textual Introduction to the forthcoming edition, will not be resumed here.

Basically, Williams's identification of A and B is shown to be sound in gatherings F-L inclusive (with the relatively minor changes that I2v must be assigned to A, and that L2v turns out to be shared between the two compositors). However, the evidence described below weakens Williams's assignment of the greater part of gatherings B-E inclusive to A, and disproves his assignation of B2r-4v to B. It has therefore become necessary to postulate another workman as the compositor of gatherings B-E. In order to retain

the initials of "A" and "B" for the majority of the book (and thereby, I hope, cause as little confusion as possible to readers of Brown's edition), I have designated this workman "Compositor N". Neither A nor B seem to have been the workmen for much of gathering M; and evidence to assign gathering A simply does not exist. The compositor-distribution then which I have arrived at, and which this article gives the evidence for, is as follows:

Part One: The Type-Line

Perhaps the single most important distinction to be drawn in attempting a new compositor-analysis is between passages set in verse, and those in prose. Unfortunately, this distinction is not as easy to draw as might appear at first glance. D. F. McKenzie has demonstrated that "indenting the stick" was a common way of aiding justification of verse lines in plays of this period.[8] Examination of The White Devil shows clearly that this method was used extensively. The evidence is a little tedious to resume, but as it has a bearing on the compositor-analysis, it will be necessary to do so. The stick was set at 23 ems, giving an approximate line-length of 96 mm. The usual short line, or indented line, was 19 ems, about 77-78 mm. Naturally these metric figures vary depending on the amount of shrinkage in the sheet of paper as it dried, but the use of a block 4 ems wide was clearly the handiest way of achieving an indentation and thus saving quantities of quads.

An examination of The White Devil reveals that in far more pages than McKenzie found in Lear, some such aid to justification was employed. In the following tabulation, "prose" is to be taken to mean that there is some prose on the page, not that the entire page is prose; there are some pages in the play wholly in prose—B3v and B4r are instances—but usually the page will contain a mixture. The following pages show no abnormalities: B1v, C1r, C1v, C4r, D1v, D2r, D4r, G4r, M1v; in addition, the following pages show no abnormalities other than that they contain prose: B1r, B4r, B4v, E2r, F2v, G3v, I2v, K2r, K2v, K4v, L1r, M2v. These pages have therefore been excluded from the Table. Thus, of the 84 pages of text in the play, only 21 reveal no

sign whatever of adjustment of the width of the type-line, or of some greater or lesser anomaly in connexion with line-widths.

Table 1 Line Widths

• If "prose" is not indicated, the page contains verse exclusively. The signs '+' and '-' after the number of ems indicate a line slightly longer or shorter than the norm, the difference almost certainly owing to variation in shrinkage of paper rather than to a re-setting of the stick. Where lines have to be turned up or down at the end, this is noted; "optional" means that the compositor had the choice whether to go up or down; "forced" means he lacked such a choice as the line above or below as the case may be was already full. "s-d" means stage-direction; "cw" means catchword.
• B2r 23+ ems, prose.
• B2v 23+ ems, prose.
• B3r 23 ems, prose. The top of the page is some 3 mm. wider than the foot.
• B3v 23 ems, prose. The top of the page is some 3 mm. wider than the foot.
• C2r S-d line 7 set flush to 19 ems, also line 9.
• C2v Up to line 24 only one line longer than 19; 2 s-ds also flush to 19.
• C3r Line 8 s-d flush to 19; only 2 lines on page longer (line 34, almost full-measure, and tightly-spaced, and line 36). At lines 16-17 there is an inked mark at the 19 em point which looks like the edge of a block. Immediately above there is a damaged period.
• C3v Line 13 only long line in first two-thirds of page; last five letters are disranged (as is final "r" of "thunder", line 17).
• C4v Block of verse, lines 16-23, justified to 19 ems; lines 28-31 justified to 18 ems—an obligatory question-mark omitted in line 30.
• D1r Cf. C3v; last two words of line 9 disranged, break at c. 19 ems. In lines 15-24 five lines are justified to 19 ems, and only three lines on the page are longer than 19 ems. However, one of them (line 17) breaks the 19-em sequence in lines 15-24, so that quads rather than a block must have been used for justification.
• D2v 23+ ems, prose (but s-d in line 11 set flush to 19 ems).
• D3r S-d in line 11 set flush to 19 ems. A mark of something inking is visible opposite lines 29-33; there are two ampersands used in line 30.
• D3v Something left an ink mark straggling over five lines or so in the top right corner, perhaps the irregular face of a large block improperly seated. At line 25 an exit-direction is set to 19 ems, and the line above, and lines 26, 28, 29 following are justified to 19 ems, as is line 36.
• D4v There is an ink-mark at the 23 em line opposite lines 20-23; lines 20-21 are justified to 19 ems, but the surrounding lines are shorter. There are inking spaces in the bottom left corner.
• E1r There is a turn-up at line 34 (an 18-syllable line!) which could easily have been avoided if the compositor had re-arranged the s-d at lines 24-25 to fit on a single line, for which there was ample room. The s-d at line 20, however, is set flush to 19 ems though lines in the vicinity are longer.
• E1v 23+ ems, prose.
• E2v 23 ems, but there is serious misalignment in line 5, the break occurring at the 19 em point; the s-d above is similarly affected, from which it is possible to infer the use of a block, imperfectly squared, to the right of lines 6-10. The second line of the s-d at lines 1-2 ends 3 mm short of flush-right, in order to give it a roughly "centred" appearance.
• E3r 23 ems. As on previous page, the s-d ends on line 10, 3 mm short.
• E3v No line longer than 19 ems until line 24; lines 6, 8, 10 justified to 19 ems.
• E4r Line 6 turned up (optional).

---

139




• E4v Only three lines longer than 19 ems; lines 3, 6, 8, 18, 21, 36(s-d) are justified to that width, none exceptionally tightly.
• F1r First half of page 19 ems; lines 2, 11, 13, 19 justified to that width; last sixteen lines are longer, including two (34, 37) full-measure; line 34 is turned down. There is an ink-mark at 23 ems opposite line 5.
• F1v Here the upper half of the page has longer lines (none full-measure), and the lower shorter. There is some evidence that an imperfect 5 em block was used, since the last letter of line 23, and the period at the end of line 25, which both slightly exceed 18 ems, seem to have slipped. Lines 34 and 37 are also set to 18 ems.
• F2r Lines 2, 10, 11, 18 set to 19 ems.
• F3r S-ds at line 23 and line 34 indented 2.5 and 3.5 ems respectively; 28 is on the same line as dialogue; 34 has its own line but is not centred and is incomplete anyway; it reads Enter the French; the word Ambassador is missing.
• F3v Line 28 exit s-d set to 20 ems, as is line 29. A 4 X 3 em block would fill the gap nicely.
• F4r S-d at line 34 is set an en-space short of flush-right, in a line where the dialogue approaches the direction closely.
• F4v Much seems to be set to 19 ems: lines 1-3, 11, 16, 17, 22, 24-25, 28 are justified to this width, including s-ds in lines 25 and 28; line 11 is especially tightly punctuated, and lines 16-17 tightly squeezed. Only lines 4, 5 and 31 are longer than 19 ems: 4 is set to 22; 5 to 20, and 31 to 21 ems, indicating that these lines were also justified.
• G1r Three lines full-measure, all showing inking shoulders (3, 5, 18). Three lines (20, 22, 28) are justified to 19+ ems; in 28 the spacing is extremely tight, a tilde is used, and the usual en-space preceding a speech-prefix at the beginning of a line is reduced to a mid-space. Something is inking against the outer edge of the lower half of the page, perhaps a really big, narrow block turned sideways—which would account for the reluctance to re-justify line 28.
• G1v All set to a maximum of 19 ems, including two s-ds, except the cw.
• G2r Several lines justified to 19 ems, including the s-d at line 36. Except the cw, no line is justified to 23 ems; even line 10, which contains an exit-direction, is not set to that measure. A small piece of ?inverted type (not pica) has left an impression at the 23-em point in line 28.
• G2v All 19 ems, save lines 30, 37 (and cw); lines 22, 24, 28 have s-ds justified at 19 ems, but no lines are particularly tight. There is a mark opposite lines 11-12 in the outer margin.
• G3r 23 ems. Two s-ds, lines 7 and 12, set respectively 13.5 and 15 mm short of flush-right.
• G4v Ink-marks in lower right corner, no sign of any special justification.
• H1r 23 ems; no real sign of 19 ems; line 20 however is very tight, abbreviated by use of a tilde and ampersand (though a VV for W occurs at the beginning—an upper-case W occurs eleven lines below), yet there seems to be ample space for a turn-under. So perhaps a block was used to fill out lines 21-27.
• H1v 23 ems; there is a forced turn-up at line 22. No sign of 19 ems.
• H2r 23 ems, prose; a rule or reglet is inking at top right.
• H2v A s-d located to the right of lines 9-12 has its left margin justified at 19 ems, the right margin is irregular.
• H3r The compositors liked to squeeze a whole verse line into a single line of type wherever possible. Note, in consequence, the problems encountered in lines 1-4 on this page:
  Ser.
  Vittoria my Lord.
  
  Fran.
  Wel: what of her? Ser. Is fled the Citty,Fra.Ha?
  
  Ser.
  With Duke Brachiano. Fra.Fled?Where's the Prince
  
  Ser.
  Gone with his father.
  Giouanni
  
  

  ---

  140

  
  
  
  Only line 3 would not have fit 23 ems, a thing the compositor might not have noticed at line 1. The consequent mislineation reveals clearly the workman's reluctance to re-set botched work. There is also an optional turn-down at line 15, and a s-d justified to 19 ems at line 25.
• H3v 23 ems, no sign of 19 ems. However, the register of the first few lines (1 and 2 are filled out with a s-d) is disturbed at about the 20 em point, disturbance which can also be seen in lines 6 and 9. This suggests movement of a block below line 9.
• H4r 23 ems, several lines set to 19 ems; turn-up (optional) at line 20.
• H4v 23 ems, prose; the last 8 lines do not exceed 19 ems.
• I1r 23— ems. Lines 6-21 are all less than 19 ems except for the last letter of line 20, which is either damaged or out of register. The s-d at lines 23-24 ends 8 and 13.5 mm short of flush-right.
• I1v 23— ems, prose. S-d lines 9-10 ends on line 10, 8 mm short of flush; s-d lines 11-12 ends on line 11 10 mm short; the direction would have fit on line 11.
• I2r 23— ems, prose. S-d lines 28-29 begins on 28, 2 mm short, ends on a line of its own, 9 mm short (and obviously, since 29 has nothing else in it, could have been entirely accommodated in 29).
• I3r 23— ems, prose. In line 24, the second line of a s-d is 2 mm short.
• I3v 23— ems, prose. S-d at line 19 is 1 mm short of full-measure.
• I4r 23— ems, prose. The s-d at lines 11-13 ends 8.5, 11.5 and 15 mm respectively short of flush. The direction would have fit on two lines if it had been flushed. Similarly, a direction beginning in line 18, five lines long, has one line (21) 4 mm short of flush.
• I4v 23— ems. S-d lines 15-16: line 16 ends 8 mm short of flush.
• K1r 23 ems, prose. Something is inking in the top-right corner, where the longest line is 20 ems. There is a fairly clear impression of an en-space at the end of line 9.
• K1v 23 ems, prose. S-d, line 7, set 2 mm short of flush; line 23 9.5 mm short (but this direction has been "centred", cf. E2v).
• K3r 23 ems; an optional turn-up, line 4; a s-d set to about 20 ems, line 14.
• K3v 23 ems; only one full-measure line (11, including a s-d); ten lines justified to 19 ems, including a s-d, line 13. There is an ink-mark in the lower right corner, opposite line 31.
• K4r 23 ems; four lines justified to 19 ems, nothing longer than that measure between line 9 and line 26. An em-quad inks in the direction-line.
• L1v 23 ems; s-ds set flush right; one turn-down (forced), line 15 (this occurs in a passage alternating between verse and prose, which even John Russell Brown mislines: prose begins at line 13 and continues until the beginning of the dirge).
• L2r 23 ems, but only one full-measure line; s-d line 2 set to 21 ems, line 36 to 20 ems. One line (28) set to 19 ems.
• L2v No line full-measure; 8 lines set to 19 ems, including 5 lines between 27 and 37; line 37 is abbreviated, employing a tilde, to make it fit. There is also a misprint in line 33 ("the" for "they") which may have been the compositor's unconscious response to his realizing that the line was going to be tight.
• L3r 23 ems, but four lines justified to 19 ems.
• L3v 23— ems; much of page is short, but only lines 14 and 36 set to 19 ems.
• L4r 23 ems, prose; lines 4-5 set to 19 ems.
• L4v 23 ems, prose; line 1 is spread out, to 21 ems, perhaps to match line 2, the same measure, but tightly set; there is a disarrangement in the near-full line 9 also at about the 20 em mark; lines 6, 11-12 are set to 19 ems; line 36 turned down (optional).
• M1r 23 ems but only two lines, and direction-line, full width. No sign of 19 ems.
• M2r 23 ems, prose; one s-d justified to 22 ems.

One of the most curious features to emerge from this survey is the practice of ending a stage-direction short of the flush-right margin. The practice is to be found on the following pages: F3r, F4r, G3r, H2v?, I1r, I1v, I2r, I3r, I3v, I4r, I4v, K1v, L1v, L4v?, M2r. The instances on H2v and L4v should perhaps be excluded, as the left margins of the directions on these pages are flush, so a different aesthetic principle is at work. For the same reasons, the directions on E2v and E3r are excluded, as the idea there seems to have been to centre the second line below the first. The other instances listed are clearly idiosyncratic, but they occur in both Compositor A's and Compositor B's work. Prior to F3r, entry directions were centred in their own line, or set flush-right. Exits are flush-right, as are descriptive directions—always allowing that "flush-right" means to the extent of the available space—i.e. 23 ems, or 19 ems, or sometimes some other figure. The emergence of a new way of treating directions in the middle of the play must be considered significant, especially as the commonest indentations (such as 8 mm) are not multiples of em- or en-spaces. The likeliest explanation is that it is a sign that the composition passed to other workmen than the compositor who had set the first four gatherings.

Some other typographical features of the quarto might be mentioned here. One such is the extremely heavy impression made by the headline and direction-line. It is relatively easy to see how the bottom of the pages could be over-impressed by a press with a warped platen (incidentally, there is no sign of this effect in the Pide Bull quarto of King Lear), but it is hard to see why the headline should be thus affected. For the record, spaces ink at the bottom of the page on B2v, 3r, 4r; C1r, 2r, 3r, 4v (i.e. the outer forme of C); D1v, 3r, 4r; E2v, 3r, 4r; F1v, 2v, 3r, 4r; G1r, 2v, 3v, 4r, 4v; H1v, 2v, 3r, 4r; I1r, 2v, 3r, 4r; K1r, 3r, 4v; L1v, 2r, 3v, 4r; M1r. The shoulder of a type also inks on line 35 on C4v. Naturally there is variation between copies, but this list represents the consensus of the three in the British Library. The effect this over-impression has on the determination of the skeletons will be mentioned below, in Part III.

The conclusion to be drawn from this investigation is that the use of blocks to fill out the stick, thereby saving quads, was common in the setting of this play. To the several instances of concrete evidence of this practice—the places where a block has left inked traces—we may add several very plausible inferences, where a line is set very tightly to 19 ems. Consequently, it is probably safe to assume that this method was used wherever convenient. In places blocks of other sizes seem to have been used, but it is not always possible to tell where a line was set to a predetermined length thanks to the presence of a block in the stick, and where upon imposition it was found possible to remove a large number of em-quads by inserting a block in their place.

The difference between the two methods is crucial. It has long been accepted, in compositor-study, that prose and verse are apples and oranges that cannot be directly compared nor mixed together to produce composite statistics. When a compositor sets a verse line, he will (so it is assumed) set a

standard space between words, follow his habitual preferences in spelling and punctuation and the use of spaces, and then at the end of the line simply fill up the stick with the necessary spaces to justify the line.[9] In prose, by contrast, while the compositor may indeed follow his spelling preferences, he may also be induced to sacrifice some or all of them to the process of justification. His "verse-habits" of spacing will necessarily be abandoned, since spacing between words is the principal method of justification. Hence data drawn from prose setting is suspect, and most investigators either eliminate it from their analyses, or classify it separately.

However, it is clear from this study of the indentation of the stick in The White Devil that a substantial number of verse lines were justified as if they were prose: that is, the final character in the line comes hard against the right margin. This is true not only for full-measure (23-em) verse lines, which are easy to see, but for a number of lines set to 18, 19 (chiefly), 20, sometimes 21 ems. Indeed there seems to be something of a preference for justifying a line of verse to a specified number of ems if it exceeded 19 ems. But this is merely an impression, impossible to substantiate. What, then, has happened to our verse-line, supposedly the norm for compositor-determination? It has become a much less certain guide, one which is no longer easy to spot. The text must be gone through carefully (bearing the imponderables of paper-shrinkage in mind) to determine which lines are actually justified to a short measure; these lines must be removed from the "verse" count, and placed with the prose. The conclusion that one must come to is that far more of The White Devil is "prose" than a glance would suggest. This has been carefully borne in mind in the subsequent section of this article. But even in such "prose" lines one can frequently find the compositor retaining some of his more cherished habits. And the bibliographer's life has become harder again.

Part Two: The Compositors

1. Punctuation.

Attempts have been made to show that Webster deliberately punctuated his manuscript in order to supply directions to the actors.[10] This notion, pretty unlikely anyway (the prompt-book and parts would have been made up by the book-keeper or his scribe, neither of them particularly likely to cherish Webster's foul-papers punctuation), is proved false by a detailed examination of the Quarto's punctuation. Altogether, not counting punctuation in the headline, the catchword, or that following a speech-prefix, there are used in the text of The White Devil 1476 periods (including 205 black-letter periods),

2043 commas, 245 semicolons, 237 colons (including italic), 92 exclamation marks (including italic) and 291 question marks (including italic). This is light punctuation for a play of over 3000 lines; it works out to fewer than three punctuation marks per two lines. Even so, as experience of contemporary dramatic manuscripts tells us, it is probably a good deal more than was in the holograph. However, more detailed investigation turns up a most surprising fact. Between B1r and E4v, the total quantity of punctuation is 1496 items for the 1182 lines, 1.27 punctuation marks per line. From F1r to the end of the play, there are 2879 punctuation marks in 1923 lines, or 1.5 per line. This is a statistically significant change.

More important, the character of the punctuation changes, as the following short tabulation of the number of specific punctuation marks per page will illustrate. The first three columns list respectively the minimum number of each punctuation mark occurring on each page, the maximum, and the average for each page of the text of the play. Columns 4-6 give the same breakdown of figures for the 32 pages B1r-E4v; the numbers in brackets in column 4 indicate the total number of pages in this part of the book entirely lacking the punctuation mark in question. Columns 7-9 break down the figures for F1r-M2v. These figures do not distinguish between italic and roman ; : ! ? or between roman and black-letter periods, since there is nothing to suggest that the compositors were doing anything other than picking up the first sort that came to hand, from cases where the punctuation sorts had besome somewhat disorderly.

             

minimum	maximum	average	B1r-E4v min.	max.	average	F1r-M2v min.	max.	average
. 7	28	17.5	7	24	14.44	11	28	19.33
, 8	56	24.32	9	56	27.34	8	39	22.46
; 0	14	2.92	0 (14)	4	0.78	0 (3)	14	4.23
: 0	8	2.82	0 (4)	5	1.78	0 (4)	8	3.46
! 0	5	1.095	0 (24)	3	0.31	0 (12)	5	1.58
? 0	17	3.46	0 (9)	10	2.09	0 (5)	17	4.31

The semi-colon is used only 25 times up to E4v, and 14 pages lack it altogether; from F1r on it is used 5.4 times as often: 220 occurrences altogether, and only three pages are without one. There is significant variation in frequency of use of other marks: the exclamation mark is used 5.1 times as often; the question mark 2.1 times, the colon 1.9, and the period 1.34 times as often (this last being more significant than it seems, in view of the numbers involved). However, the use of the comma drops by nearly a fifth, to 0.82 times as frequently. Both compositors A and B use this "new" punctuation for the remainder of the play without much statistically significant difference. Table 2 assigns 22 pages to Compositor B and 25 to Compositor A (some of them doubtful); L2v is not included in the count since it was almost certainly shared between the two compositors, and gathering M cannot be reliably assigned to either. In B's pages, the semi-colon is used altogether 106 times, or 4.8 times per page; in A's, it is used 97 times, or 3.9 times per page. The

other punctuation marks follow a similar pattern: B uses 61 colons, 2.77 per page; A 82, 3.24 per page. B uses 43 exclamation marks, an average of 1.95; A uses 36, an average of 1.44 per page. The question-mark is used 110 times by B, or 5.0 per page; A's figures are 112, 4.48 per page. These averages are not identical (nor in view of the variety of the copy could they be expected to be), but they inhabit a different statistical world from the averages in column 6 of the table above. Various hypotheses may be advanced to account for these observations, but none is as persuasive as the likeliest and simplest: that two new compositors took over at the beginning of gathering F from the workman who had set B-E.

2. Terminal punctuation of speeches.

There are 43 speeches in the Quarto of The White Devil which end with no punctuation at all, (or where the terminal punctuation is in the associated stage-direction only). Although it is possible to argue that absence of punctuation within a speech is merely a sign of rhetorical rather than grammatical punctuation, it is hard to think that there should be nothing at all at the end of a speech. There are also 75 speeches in Q1 of The White Devil which end with a comma, a colon, or a semi-colon. Some of these may, perhaps, employ the mark in order to indicate that the speaker is interrupted, but by no means all. A few times a speech lacking punctuation is followed by a stage-direction which ends in a stop, but this is only occasionally the case, in B4v, C1v, and I1r for instance.

25 of the unpunctuated sentences and 50 of those terminating with a irregular punctuation occur up to E4v; the figures for the remainder of the play are 17 unpunctuated, and 25 with another mark: clearly, a much lower proportion than in the earlier part of the play. Of these 42 irregular sentences in F1r-M2v, we find, in the pages assigned to A, 6 which end with nothing and 15 with a comma or other irregular punctuation; in those assigned to B, 10 with nothing and 7 irregular, leaving 4 in unassigned pages. This would rather suggest that B was more prone to omit the punctuation mark altogether than A, who in turn was more likely to use a comma to end a speech than B. B-E4v has an average of 0.81 unpunctuated and 1.56 irregularly punctuated speeches per page; Compositor A's figures from F1r on are 0.24 and 0.6. The proportion of unpunctuated to irregularly punctuated sentences in Compositor A's pages is 1:2.5; that in B-E4v is 1:1.92.

Collectively, these studies of punctuation go far to establishing that the compositor of B-E4v was a different workman from the two compositors Williams identified as A and B. I think it fair to conclude that Webster's MS. must have had very defective or irregular punctuation, with which each compositor did his best, with variable success.

3. Compositor-analysis by spacing.

Presence or absence of spacing in conjunction with punctuation has become a new tool in compositor-identification in the last few years. It has been recognized that, especially in a Quarto, the number of words revealing idiosyncratic

spellings on any given page may be far too few to permit a reliable identification, especially of compositors whose habits are quite close. It seemed that the investigation of a wholly mechanical parameter, such as whether the compositor used a space after a comma, or before a colon, would be a much more reliable way of amassing the kind of objective evidence in quantities that would be statistically significant. However, the malleus bibliographici of our time, Professor D. F. McKenzie, has recently contrived to cast grave doubt on the validity of such an assumption.[11] With evil and puckish delight, McKenzie sets up a case where it looks utterly clear that analysis of spaced commas will determine the compositors, only to demonstrate from irrefutable external evidence in the Cambridge University Press's archives that the spacing evidence tells us nothing whatever about who set what, and that a compositor would space commas on one page, and not space them on another, with a fine abandon. He insists that the burden of proof is upon those who would wish to claim that practices in a commercial London printing house a hundred years earlier were different from those of the Cambridge compositors. Where, then, does this leave the bibliographer with no such external evidence?

In a word, in a situation of healthy scepticism and caution. What I have assumed in this article is that if discriminators agree (for instance if we find on one page that a compositor spells -ie and also usually spaces before a colon in verse, where on another page the usual spelling is -y, and the colon does not have a space before it) then the likelihood increases that these are the practices of two distinct workmen. The more discriminators that coincide, the stronger the supposition, but one must be clear that such suppositions are never more than that, and that virtually any pattern of apparently habitual composition practice can be susceptible to some alternative explanation. Nor am I in a postion to argue that Okes's men were different in their habits from those at the Cambridge University Press a century later. It seems to me that it is not unlikely that they were, but the evidence to prove it does not exist. I must reiterate that I speak only of spacing in association with punctuation. It is not yet within the scope of bibliographical knowledge to ascertain the number and kind of spaces a compositor used routinely between words to space them and to justify his lines. Nor do I believe that such information will ever become available. With these reasonable cautions in mind, let us see what an examination of spacing with punctuation in The White Devil can yield.

The results are found in Table 2 below, but may be summarized here. The first column lists the number of cases in which the compositor put, or failed to put, a space following a comma. Obviously, commas at ends of lines and those used as terminal punctuation are omitted from the count. Sometimes there is a space before as well as after, or a space before, not after. Taken as a whole, the text of The White Devil contains 899 commas which

are spaced (327 of them up to E4v, 572 thereafter) and 265 which are not spaced (165 up to E4v, 100 thereafter). The general statistic, then, is that the compositors were 3.4 times more likely to put a space after a comma than not (1.98 times up to E4v; 5.72 times afterwards). In B's 22 pages there are only 8 unspaced commas, compared to 257 spaced: he was 32.1 times more likely to space a comma than not, and he averaged 12 commas a page. In the 25 pages more or less firmly assigned to A, there are 256 spaced, and 85 unspaced commas; he was 3.0 times more likely to space a comma than not. This is a significantly lower proportion of unspaced commas than in the B-E4v section of the play.

The possibility that Compositor B could have worked in gatherings B-E is eliminated by the second class of spacing evidence, tabulated in column two of Table 2. In this I list the number of punctuation marks (the ! ? : and ;) which have a space before and after them, compared with those that only have a space after. Up to E4v, a mere 24 marks have spaces before and after; 107 have the space after, a proportion of 1:4.46. In Compositor B's pages from F3r on, the numbers are 267 spaced before and after, against 39 spaced after, a proportion of 6.8:1; in the pages assigned to Compositor A the numbers are 31 before and after, against 275 spaced after,[12] a proportion of 1:8.87, or rather more than the opposite of Compositor B. The discrepancy between the proportion for B-E4v and A's stint after F1r is statistically significant, though obviously not of the same order of difference as that between A and B.

In conclusion, and with due regard for Professor McKenzie's caveat, it seems clear to me that Compositor B almost invariably set a space before as well as after the four marks under discussion, and that by a substantial proportion he preferred to put a space after a comma; Compositor A very seldom put a space before any of the four marks, and was much more variable not only in his treatment of commas, but in the number of them he used on a page. Curiously, it seemed not to make a great deal of difference to these practices whether the compositor was setting "verse" or "prose", though in certain individual lines of course it is possible to detect use or absence of spaces as aids to justification. But it hardly affects the overall pattern, so in Table 2 I have been content to indicate if there is prose on the page in question rather than go into elaborate detail about it.

4. Compositor-analysis by spellings.

Compositor-analysis has traditionally been done by searching a number of words for evidence of habitual spellings. The net of potential words has grown larger over the years, but at the same time the method has grown more sophisticated. It has been recognized that the most telling kinds of habitual spelling are often not whole words but the treatment of suffixes and

verbal endings such as -d, -ed, -'d or -t to form the preterite. Even so, it must be admitted that less new evidence was yielded by this investigation than by those previously described.

I searched the following in Q1 of The White Devil, in this instance using the Malone copy for reference, and the results are tabulated below:

• Group 1. Endings -ie/ -y; -'d/ (unvoiced) -ed /-d /-t; -our / -or; -'s / -s in possessives; -l / -ll (and also, separately, the words all / al); -es / -ess / -esse.
• Group 2. The common doublet spellings: do /doe; go / goe; here / heere / [other forms of the word]; Ile or ile / I'll or I'le; were / weare; heart / hart; been / bin / byn / bene; blood / bloud; nere and ere / ne're and e're; mischeefe / mischiefe; whore / whoore; cossin / cousin; lie / lye; truely / truly; pitty / pity; poyson / poison; ith / i'th; tis / 'tis.
• Group 3. Single or double terminal e: be/bee; me/mee; he/hee; she/shee; we/wee.
• Group 4. Words which begin in -in or -en.
• Group 5. In addition I searched and recorded the spelling of 294 other words whose spelling appeared to be idiosyncratic.

Of the endings in Group 1, the most useful was the first. There is throughout the play a general tendency to form the preterite in 'd in verse, but the other terminations turn up irregularly and in insufficient numbers to help. Endings in -or are rather less frequent than endings in -our (1:1.3 up to the end of gathering E, 1:2.3 thereafter); both forms may be found however on the same page, and in view of the relatively small size of the sample (about 65 items) and the effect on the figures of the influence of "prose" setting, it does not prove possible to make much use of these endings as discriminators. The contrast between four -our forms on I1r and four -or forms on I4v may suggest a tendency on B's part to prefer -our, and A to prefer -or, but I would not put it more strongly than that. Possessives are formed with or without an apostrophe indiscriminately (though E3v is unusual, with no fewer than seven possessives without the apostrophe; by contrast, I4v uses apostrophes six times —most pages have far fewer possessives). Terminal -ll is greatly preferred to -l, and most of the times the latter does occur, it is in prose or a tightly-justified line; the form "al" does not occur. The -esse termination is preferred; -ess does not occur, and -es is rare.

Philip Williams, in his article, lit upon the two spellings from Group 2 that prove to be most helpful to compositor-determination in this play. The use or omission of the apostrophe in 'Ile' / 'I'le' is the most valuable discriminator, and the use or omission of the apostrophe in 'ne're' etc. follows sufficiently closely on its use or omission in 'I'le', for it to be tabulated in the same column (see below, Table 2). Rather unusually, both forms occur on pages C2r, D3r, E4v, I4v and K3v; elsewhere the pages conform to one or the other form. The compositors' practices in spelling 'do' or 'doe' are helpful, and are also tabulated below. There are only two occurrences of 'goe', both on I4r; 'heere' occurs relatively seldom also: on D1r, G3r(p), I1v (but so does 'here'), K1v(p), K3r(p), L1r, twice on L1v(p), and L2v; 'heare' is found

on C1v and L1r, 'heer' on I4v and 'heer's' on K2r. Elsewhere the spelling is 'here'. Few of the other listed variants occur frequently enough to be of any serious help. 'Weare' (=were), for instance, occurs occasionally (five times on B3r) up to D3v, but not thereafter. 'Heart' is the usual form, but 'hart' shares a page with it on B4v and H1v. The various unusual ways of spelling 'been' may reflect oddity in the copy more than anything else, with the exception of 'bin', which compositor B seems to have quite favoured: apart from two occurrences in prose on B2v, it is found on H1v, I1r, I1v, K2r and M1v (twice); it also occurs in Compositor A's page K4v, but there it is a rhyme-word. 'Bloud' is the usual form; 'blood' is found on E3r, L2r, and no fewer than four times on M1v! These words, together with those of Group 5, occur so seldom that while they may help to confirm the assignation of a page to a compositor, of themselves they cannot determine one.

Group 3 proved both helpful and frustrating, and the actual count of words spelled each way is listed in Table 2. It is clear that these discriminators can be misleading. For instance, while I have listed page F4r as revealing "B" preferences (on the grounds that there are four -ee endings against only one -e ending), at the same time I have shown page I3r as indicative of A's habits: even though it has five -ee endings, it also has eight -e endings. Clearly, Compositor A could set a few -ee endings per page if he felt like it. The range is from the extraordinarily definite page I4r, with 22 -ee, and no -e endings, to page G1v, with none of either!

In his edition (p. lxiii), John Russell Brown gives the impression that Group 4 is a significant discriminator, that A preferred -en and B preferred -in. However, I found very few -en forms, and very, very few words spelled both ways: there is 'entreat' on C3r and 'intreat' on C2v and G3r, 'endanger' on C4r and 'indanger' on D4r; 'enough' is the usual form, but 'inough' occurs on C3r (which rather counters the effect of the 'entreat' on the same page). It seems that this can scarcely be used as a test.

Of the Group 5 spellings, 136 had to be dismissed, as being invariant (for instance 'answere' is the only form of the word used in the play), or variable only where the results are uninformative ('claime' for instance is spelled thus in the play except on L1r, where 'clame' occurs—but in a prose passage), or where the word occurs only twice or thrice in the play.[13] There are for instance some particularly odd spellings in the italic description of the dumbshows on D4v—E1r, whch occur nowhere else. A few words, entertainingly, are spelled in different ways on the same page: 'bawdes' / 'baudes' on G1v; 'carracter' / 'character' on E3v; 'Citty' / 'Cittie' on H3r; 'dreampt' / 'drempt' on K3v, 'Physitions' / 'Physitians' on I4v. Such events are neither unusual nor surprising in dramatic texts of the period. Very few words recurred sufficiently in clearly different forms to be revealed as characteristic of either Compositor A or B: 11, to be precise, which I now list:

                         

Compositor A	Compositor B
already[#]	allready
battels	battailes
commander(ers)	commaund
coine[#]	coyne
conuei'd	conuayd
dogs	dogges[*]
dost	do'st[**]
fy	fie[***]
gaule	gall
money[#]	monie or mony (both on K2r)
oath, oathes	othe(s)
yeare[#]	yere[****]

These are slim pickings, which scarcely help at all in identifying uncertain pages. A number of words require some comment. In alphabetical order, they are:

The exclamation 'o' with a circumflex (which occurs 18 times in the play) looks at first like a possible authorial form, since it is found throughout the play, and in both A's and B's pages. Confirmation seems at hand when it turns up in A Monumental Column, C1r, line 6. That is, until one notices that the printer of this work was also Nicholas Okes. It may, therefore, be a "house" practice, if a rather intermittent one.

Both compositors use the form 'to th', but 'to'th' (a common form in The Devils Law-Case) occurs twice, in A pages.

'True' is the usual spelling, but 'trew' is found on B1v, B2v, B4v, E3v (twice), E4r and I4r (B). This is a rare case of a spelling preferred by B being found early in the play.

There are ample instances of compositorial inconsistency (that is, using more than one spelling for a word). For instance, both A and B spell 'breast' and 'brest', 'hower' and 'howre' (but never 'hour' or 'houre'); B spells 'flowers' and also 'flowres'; A spells both 'forty' and 'fourty'. 'Murder' and its derivatives are spelled with a d, except on I4v, where A allows himself a 'murther'; 'neere' and 'ore' are spelled thus by both compositors, but B also uses 'neare', and A usually prefers 'o're'. 'Polliticke' is the usual form in the play; B also uses 'politicke' (G4r) and 'pollitique' (G3v, in prose); A supplies one 'pollitick', but the line is tight (I4v). Generally, it must be concluded that wordspellings support, but do not materially aid, the distinction between Compositors A and B.

Two important hopes for definitive results from investigations of composition practices were frustrated. The first, and shorter, is the odd case of gathering M, which contains some very confusing contradictions in the evidence, as may be seen from Table 2. Specifically, while M1r—2r show clearly Compositor B's preferences for spacing, both of commas and other punctuation marks (columns 1 and 2), and for the spellings 'Ile' and 'doe', these pages show equally clearly Compositor A's preference for -y endings, while the evidence of -e/-ee terminations is ambiguous. In addition, gathering M contains some unique spellings: for instance 'ayre' which elsewhere in the play is 'aire'; 'behould' ('behold'); 'betraid' ('betray'); 'could' ('cold'); 'cought' ('caught') 'noe' (twice, as against 'no'—73 occurrences); 'Sirha' ('sirra' / 'sirrah'); and both 'woeman' and 'woemen' (twice), against 'woman' 22 times elsewhere, and the four occurrences of 'blood' on M1v already noted. These idiosyncratic spellings occur on M1r, M1v and M2r but not on M2v. In light of the conflicting evidence elsewhere in the gathering it is tempting indeed to think that another compositor might have taken over the final half gathering and preliminaries. The alternative, that both A and B worked on gathering M in a quite indiscriminate way, is difficult both to demonstrate and to imagine working in practice, especially in view of these unusual spellings. But I hardly think that they prove beyond question the presence of a Compositor C either.

The second question deals with the composition of gatherings B—E. The issues to be decided now are, first, whether any of the spelling evidence confirms or refutes the hypothesis already established that neither compositor A or B worked in these gatherings; and secondly whether there is anything to suggest that there was more than one compositor working in gatherings B—E. Evidence for the former is equivocal and scanty. Few of A's commonest word-spellings appear, and though the discriminators listed in Table 2 show a good deal of similarity between most of the pages in gatherings B—E and Compositor A's later in the book, there are anomalies and incomplete parallels in this evidence. The occasional signs of B's preferences surface so irregularly and randomly that we can safely rule out his participation.[14] One

line of investigation is to note spellings in B—E against which both A and B agree. There are some: 'eather' on B1v, against 'either'; 'fayre' against 'faire'; 'fier' on B3v and E3v, elsewhere 'fire'; 'only' on B1v, D3v, E3v (and 'onlie' on B3r) against 'onely' elsewhere; 'ritch' on E3v, 'rich' elsewhere; 'voyage' on B2v, B4r (prose) and E1r, against 'voiage' on L3v. It is not a rich harvest; the spellings of the earlier gatherings tend to have more unusual forms than those of the later, but this could be as readily accounted for by the compositor's increasing assurance with the MS. hand, and consequent decreasing tendency to reproduce copy-spellings, as by a change of workman.

In short, the study of spellings does not materially influence the conclusion reached above, that A and B did not work on the book prior to gathering F. It is true that this requires us to posit, in Compositors N and A, workmen whose practices were in many respects closely similar, and only susceptible of discrimination on the basis of statistical analysis rather than by gross variation in spelling preferences. However, this scenario is by no means an unlikely one in a shop like Nicholas Okes's, where much employment was made, as Blayney has shown, of casual journeymen. Also to be reckoned with is the likelihood that an apprentice will learn habits from his master that he will retain as a journeyman. In short, the differences between A and N are subtle but sufficiently clear, when all the evidence is weighed, to assert with some confidence that they were two separate workmen.

The speech-prefixes help to confirm this determination. The most suggestive is that for Francisco, the Duke of Florence, whose prefix in B—E is usually 'Fran.' (46 times), but occasionally 'Fra.' (9 times). Compositor A prefers 'Fra.' (38 times, to 8 'Fran.'), while Compositor B's preferences are opposite: he uses 'Fran.' 23 times, to 5 of 'Fra.' This alone would go a long way towards confirming that A and N are different workmen.[15] Flamineo's prefix is 'Fla.', except in gatherings C—E, where it is 'Flam.'; in the same way, Brachiano is usually 'Bra.', except in C—E, where he is (mostly) 'Brac.', though eleven 'Bra.' forms occur. In other words, it seems as if Compositor N agreed with both A and B in preferring 'Bra.' and 'Flam.' in gathering B only. Why he should have used 'Brac.' twelve times, and then preferred 'Bra.' in the proportion 55:12 is hard to guess, but it is worth noticing that 'Brac.' occurs only twice after gathering F begins. As for 'Fla.', it is possible that in casting off for gathering C, Compositor N might have realized that 'Fla.' for Flamineo and 'Fra.' was a potent source of error (Francisco does not appear in gathering B), and decided to distinguish the characters by prefixing Flamineo 'Flam.' If so, he was right: a mistaken prefix for Francisco occurs at least six times, and perhaps once for Flamineo. But why he should have altered the established prefix 'Fla.' instead of adding the letter to Francisco's prefix, I cannot venture to guess. No other prefix occurs regularly enough in variant form to be evidential: Vittoria is 'Vit.' except for three 'Vitto.' prefixes on B4r; Camillo is 'Cam.' until his untimely death

except for six 'Camil.' prefixes, which share B4r with four 'Cam.'s. To resume, the most significant speech-prefix is Francisco's, which confirms the distinction between A and N.

When we come to try to use occasional spellings in an attempt to determine if there was more than one workman in the B—E section, the evidence is even less promising, as the following list of variant spellings shows. (These variant spellings apply only to the first four gatherings; many of the words, let alone variant spellings of them, do not occur in later gatherings.)

It would perhaps be just possible to suggest, on the basis of this listing, that one compositor (N) set B1r-v, B3v—C1r, D3v, E1r—E3v, while another (O) set B2v—3r, C1v—D3r, D4r, E4r-v. There is no evidence for B2r, D1v, D4v. However, this "evidence" is so sparse that it cannot be regarded as more than suggestive (though the cluster of unusual spellings on E3v is worthy of remark). Besides, the principal discriminators listed in Table 2 fail to support any such assignment. Throughout the B—E section of the book the preferred spellings are 'do', 'go' and 'here', and words which could end in -e or -ee, or -ie or -y, tend to be found in the forms subsequently favoured by Compositor A, or at least in forms which do not give any support to the hypothesis that more than one workman was employed in the first four gatherings of The White Devil.

If we return to the matter of ending sentences without punctuation, or with a mark other than a period, we do find a suggestion of evidence that is not inconsistent with the hypothetical distinction between "N" and "O". In the pages suggested for compositor N, there are 10 speeches ending without any punctuation, and 35 with irregular pointing (28 with comma, 7 other marks); in those suggested for O, the figures are 16 without punctuation, and 13 with irregular (9 with commas, 4 other marks). (4 irregular sentences occur in unassigned pages.) That is, the "N" pages show some preference for irregular punctuation rather than none at all, a preference that is reversed in the "O" pages.

However, in my opinion, it is best to dismiss all this speculation as the chasing of moonbeams, for the evidence is really no more substantial. The only responsible conclusion at this juncture is that we have no entirely compelling reason to see the hand of more than one workman before gathering F, a workman now adequately discriminated from the compositors who began to set in gathering F. The possibility that yet another workman may have been briefly involved during gathering E is discussed in Part III. As to the question of whether the compositor of B—E could also have set M and A, the answer cannot be firm either, but the likelihood is that he did not. There are spellings (already noted) which occur only in M, and few spellings or practices which resemble those of B—E.

I conclude this section by tabulating the various discriminators discussed above. The table reveals that the usual opinion (that compositors normally were responsible for setting an entire page) must be qualified for The White Devil. Note especially the peculiarities in F2v, H4v, I4v and L2v, all of which show some signs of shared setting, signs amounting to strong evidence in F2v and L2v at least. I wish to make it perfectly clear that the use of the letters "A" and "B" in all columns save the last does NOT mean that I am attributing the page to either compositor: on the contrary, as should by now be clear, I believe that gatherings B—E were set by another workman, Compositor N. The purpose of the letters is to show whether the discriminator in question corresponds with the practice of either A or B as described in detail above. The final column lists the compositor I believe set the page in question, on the basis of a comparison of all the discriminators. The very inadequacy of the evidence for either A or B in gatherings B—E is the chief reason why I believe them to have been set by a different workman altogether.

TABLE 2 Compositor Determinants

• Column 1 lists whether or not a comma is followed by a space; the figure to the left of the virgula indicates the number of commas that are spaced; that to the right indicates how many are unspaced.
• Column 2 indicates the kind of spacing associated with punctuation marks other than commas or periods: the number to the left of the virgula indicates punctuation marks which have spaces before and after them; the number to the right of the virgula indicates those that have spaces after them only.

---

154




• Column 3 indicates if the two kinds of spacing evidence together is characteristic of either Compositor A or Compositor B; a ? following the letter indicates a limited agreement; when a blank is left, this means that the discriminators are ambiguous.
• Column 4 tabulates optional -ie or -y endings (for example, 'away' is not tabulated because the form 'awaie' does not occur). In this column an entry such as 0/4|2/4 means that there are no -ie forms in verse, but four in prose on the page, while there are two -y forms in verse and four in prose.
• Column 5 attempts to indicate if the pages reveal either A's or B's habits, on the basis of Column 4's evidence, in the same manner as Column 3.
• Column 6 simply indicates whether the page as a whole shows signs of Compositor B's preference for 'Ile' and 'nere' or of Compositor A's preference for 'I'le' and 'ne're'.
• Column 7 tabulates 'do' and 'doe' in the same way. (In fact, 'doe' is not found before F2v.)
• Column 8 indicates the number of -ee and of -e endings found, and indicates which compositor's preference the page corresponds to.
• Column 9 represents a total assessment of the evidence, by assignation of pages to compositors N, A, and B, on the basis of evidence in this table, and elsewhere in this article.
• In columns 6-9, the presence of a letter indicates strong preference; and a ? indicates equivocal evidence; in all columns an = sign means that the discriminators are equal; a — means the discriminator does not appear on the page in question. Footnote references (marked *, #, ^) refer to notes at the end of the table.

                                                                 

	1	2	3	4	5	6	7	8		9
B1r	5/1	1/2		0|3	A	-	-	2/2	=	N
B1v	15/2	1/2		0|3	A	B	A	2/1	?	N
B2r	14/3^	3/6^	A	1/3|3/6	A	-	-	4/4	=	N
B2v	8/0	1/2		2|2/2		-	A	3/5	A?	N
B3r	11/3^	1/0		0/4|2/4		B?	A	2/7	A	N
B3v	19/4^	0/2		0/1|1/12	A	-	A?	2/9	A	N
B4r	12/7^	1/1		2/1|2/1	=	-	A	1/3	A	N
B4v	18/0	3/4		7|1	B	-	A	3/6	A	N
C1r	6/7	0/1		0|7	A	-	-	0/2	A	N
C1v	6/6	1/4		0|4	A	B	A	1/3	A	N
C2r	3/10	0/6	A	1|2		?	-	3/7	A	N
C2v	10/2	0/1		1|5	A	A	A	3/3	=	N
C3r	6/6	1/4		0|5	A	A	A	1/1	=	N
C3v	8/8	0/3		1|1	=	A	A	1/3	A	N
C4r	11/8	0/4	A	0|1		-	-	3/9	A	N
C4v	4/4	0/4		0|11	A	A	-	3/7	A	N
D1r	6/8 ^	0/5		0|8	A	A	A	2/5	A	N
D1v	13/9	1/6	A	0|4	A	A	A	2/2	=	N
D2r	15/2	0/6	A	0|4	A	?#.	A	0/5	A	N
D2v	15/9 ^	0/3	A	0|4/2	A	A	A	1/3	A	N
D3r	6/7 ^	1/2		0|3	A	=	A	2/0	B	N
D3v	5/7	0/2		0|3	A	=	A	3/5	?	N
D4r	8/3	0/3		0|3	A	A	A	1/3	A	N
D4v	27/14 ^	0/0		0|2	A	=	B	5/3	? N
E1r	19/5 ^	0/2		0/1|3	A	-	A	1/4	A	N
E1v	12/4	0/6	A	0|0/1	A?	-	A	2/4	A?	N
E2r	12/12 ^	0/1		0|1/4	A	A	A	1/2	?	N
E2v	8/3	3/2		0|3	A	B	A	3/2	?	N
E3r	4/3	0/3		0|5	A	-	-	5/0	B	N
E3v	6/6	3/8		3|6	A?	B	-	4/1	B	N
E4r	6/1	3/4		1*|6	A	-	-	10/0	B	N
E4v	9/1	0/8	A	0|12	A	?	A	0/2	A	N

                                                                                                       

F1r	7/3	0/4	A	1|7	A	B?	A	2/6	A	A
F1v	4/5	0/5	A	0|6	A	-	-	4/5	?	A?
F2r	9/2	0/6	A	1|2		-	A	1/3	A?	A?
F2v	20/2 ^	3/10	A	0/1|1/3		A?	B?	5/1	B	A?#
F3r	10/0	10/0	B	1/2|1/1		B	?	11/2	B	B
F3v	14/2	10/3	B	3/8|0/2	B	B	-	4/2	B?	B
F4r	6/0	7/1	B	4/1|2/0	B	B	B	4/1	B	B
F4v	5/3	1/14	A	0|11	A	B	A	0/4	A	A?
G1r	9/9	1/9*	A	2|3		A	-	2/1	?	A/B
G1v	6/0	2/3*		0|3	A	A	A	0/0	-	A
G2r	3/6	1/7*	A	0|10	A	A	A	0/6	A	A
G2v	5/7	0/13*	A	0|5	A	A?	A	0/5	A	A
G3r	13/0	9/2	B	5|0	B	B	-	5/0	B	B
G3v	14/0	20/2	B	3/3|2	B	B	B	13/1	B	B
G4r	14/0	10/2	B	10|2	B	B	-	5/0	B	B
G4v	12/0	24/3	B	1|2		B	B	8/0	B	B
H1r	13/0	14/1	B	3|1	B?	B	B	8/1	B	B
H1v	9/1	7/1	B	7/1|1	B	B	B	3/1	B?	B
H2r	18/5	4/13	A#	0|3/2	A	A	A	0/1	?	A
H2v	12/2	1/9	A#	1|3	A?	A	-	0/1	?	A
H3r	4/2	6/3	B?	4|0	B	-	B	6/2	B	B
H3v	14/0	14/2	B	4|4	=	B	B	10/1	B	B
H4r	11/2	0/12	A#	2|3		-	A	2/2	=	A?
H4v	11/3	1/9	A ^	0|5/3	A	A	-	6/1	B	A?#
I1r	18/1	7/1	B	8|5	B	B	-	6/2	B	B
I1v	5/0	14/2	B	2/1|3	=	B	B	5/0	B	B
I2r	12/0	14/3	B#	1/7|2/3	B	-	-	7/0	B	B
I2v	16/3	2/12	A	1|2/8	A	A	?	3/2	?	A#
I3r	8/4	1/8	A#	0|7/2	A	A?	A	5/8	A?	A.
I3v	8/0	2/11	A ^	0|7/4	A	A	A	5/4	?	A
I4r	14/0	17/0	B	2/1|1/1		B	A	22/0	B	B
I4v	8/0	3/14	A	0|3	A	=	-	2/2	=	A?#
K1r	10/1	6/10	A?#	0|4/2	A	A	=	1/2	?	A
K1v	15/0	8/3	B	4|1/3	B?	B	-	9/1	B	B
K2r	16/0	19/2	B	3/1|2	B?	B	B	8/2	B	B
K2v	5/0	7/1	B	2|1	B?	B?	-	1/2	?	B
K3r	7/0	14/0	B	9|2	B	B	-	12/0	B	B
K3v	6/2	0/20*	A	1|5	A	=	-	4/6	A?	A
K4r	14/2	0/9*	A	1|5	A	-	-	1/5	A	A
K4v	22/7	0/11	A	0|3/7	A	A	-	3/6	A	A
L1r	15/0	13/2	B	6|2	B	B	-	6/0	B	B
L1v	13/1	6/3	B?	4/1|4/1	=	B	B	10/1	B	B
L2r	14/1	17/2	B#	7|2	B	B	B	11/0	B	B
L2v	13/0	9/6	?#	3|8	A	B	A	1/1	=	B/A
L3r	6/2	1/15*	A	1|15	A	A	?	5/6	?	A
L3v	8/6	0/11*	A	0|9	A	A	A	0/4	A	A
L4r	13/6	4/3	?#	0|7/1	A	-	-	2/5	A	A
L4v	12/3	2/17	A	2|4/3	A	-	A?	4/3	?	A
M1r	9/1	12/3	B	0|5	A	B	B	5/3	B?	B?#
M1v	11/0	12/2	B	0|2	A	B	B	3/3	=	B?#
M2r	16/1	7/0	B	1|6	A	-	B	2/1	?	B?#
M2v	15/5	2/5	A#	0|2	A	-	B	2/2	=	A#

∧ Some prose.

• # D2r: two 'I'll', but three 'nere'.
• F2v: a number of B's preferred forms show up in the last half dozen lines on this page.
• H2r: first half of the page is prose.
• H2v: looks like an A page, despite the large number of spaced commas.
• H4r: looks like an A page, despite the large number of spaced commas.
• H4v: looks like an A page, but has 5 'hee' forms in lines 25-34.
• I2r: mostly prose.
• I2v: looks like an A page, despite the large number of spaced commas, and a number of B spellings: 'doe' (prose) 'mee' (prose) 'bee' (prose) (but also 'do' in prose and 'she' in verse), 'Gipsie' (verse). In fact almost the entire page is in prose, and the B forms mainly in the prose sections. Williams and Brown assigned this page to B, I think erroneously.
• I3r: about half prose.
• I4v: an odd-looking page, as if A did most of it, but B set a few lines.
• K1r: about half prose; most of the B forms are in the prose sections.
• K4v: mainly prose.
• L2r: not counting marginalia.
• L2v: the most difficult page in the book, the figures for L2v are confused by the large number of 19-em lines in the lower half, which produce some very odd spacing in lines 27, 33, and 37 etc. The upper half of the page has most of the signs of a B page, but I think it highly probable that A took over at about line 25.
• L4r: most of the before-and-after spaces are in prose.
• M1r-M2r: these three pages look quite like B's work, but they have virtually no -ie endings, and terminal -e is much more common than -ee.
• M2v: text at least is A; prose explicit may be B.
• * G1r: including three with no space before or after.
• G1v: including one with no space before or after.
• G2r: including one with no space before or after.
• G2v: including one with no space before or after.
• K3v: including seven with no space before or after.
• K4r: including two with no space before or after.
• L3r: including two with no space before or after.
• L3v: including one with no space before or after.
• L4v: including one with no space before or after.

Part Three: Type-Study

I should have liked to report that I had established the order of printing of the formes of The White Devil, but in some embarrassment I must confess that despite diligent attempts, the Povey lamp failed to yield me much illumination on this question.[16] Nor, unfortunately, was an examination of the watermarks very much more revealing. In fact I have not been able to examine the watermarks of every copy of Q1 of The White Devil; apart from the fact that some libraries take an understandably dim view of the kind of treatment the binding gets in such investigations, and that few institutions

are equipped with mechanical devices to make such examinations less damaging to the book, I have not been able to visit every library holding a copy of the Quarto. Of the large number that I have seen, I examined the paper of the three copies in the British Library, and that in the Bute Collection of the National Library of Scotland. These were sufficient to make it immediately evident that Okes did not buy a special heap of paper to print Webster's play. In these four copies, a great variety of pot paper appears, with at least 20 and probably more different watermarks. In addition, numerous gatherings are printed on paper which seems entirely devoid of watermark: half-gatherings A and M in all three BL copies, for instance (Bute has a watermark in M), but elsewhere in the book: BL copy 840 c.37, for instance, has no watermarks in A, C, I, K, M; Bute lacks them in A—D, but has them elsewhere. In view of the evidently random nature of the papers being used, I have not pursued this line of enquiry.

Another route to the determination of printing order lies in examination of the skeleton-formes. Although John Russell Brown complained, "The running titles in this quarto are so similar to each other that it is very difficult to trace the press-work in detail," ("Printing II," 117) and although it is true, as was noted above, that for some reason the headline impressed very heavily, thus obscuring the kinds of typographical damage on which identification is based, with care all the settings of the headline can be identified. Altogether, there are ten settings of the running-title, "Vittoria Corombona". Seven settings are used in gathering B (B1r of course has no running title); that is, two skeleton formes were made up. However, the skeleton employed for inner B was used to print both inner and outer formes of gathering C; the headlines were re-arranged for inner C and the skeleton rotated for the outer forme. Rotated again, this skeleton was used for the inner forme of D, and then for the outer forme of gatherings E—L, in the course of which it was rotated and re-arranged twice more. The headlines used in outer B, with one new one for the fourth page, are used in outer D, and then become the regular skeleton for the inner formes from E to K. When inner K was stripped two of the headlines apparently were distributed, since two new ones were set for inner L. One of these does not recur, but that on L3v is used again in half-gathering M, together with one headline from the inner skeleton and two from the outer. What this means, in summary, is that most of the book was regularly printed with two skeletons, though an apparent error was made in the early distribution of part of the inner forme skeletons after gathering K was machined. Rather more interestingly, since all of gatherings B and C was in type at the same time (as evidence described below will confirm), Compositor N must have imposed his pages for inner and outer C in separate chases, and transferred the skeleton from the inner forme, once machining was complete, to the chase containing the pages of the outer forme. Yet the skeleton of outer B was not broken up, since its headlines survive to reappear in outer D, and inner E—L. There is thus little in the skeleton-evidence to aid in determining order of printing.

These lines of enquiry not having proved particularly fruitful, let us turn

to the crucial matter of the analysis of damaged types. The three principal purposes of such an investigation are as follows: first, that such types can be used to determine beyond question whether the book was set by formes. That is, if the same identical type turns up in both inner and outer forme of the same gathering, the inescapable inference is that it must have been distributed back into the case after the forme was worked, and re-set into the next forme. A quarto set seriatim would need to have seven of the eight pages of a gathering in type before the inner forme could be machined. The likelihood, in such a case, that the inner forme could be distributed and its types used in setting the eighth page of the gathering is vanishingly small, unless by some chance there was some kind of a hiatus in printing. Q1 of The White Devil reveals no reliable instance of a type occurring in both inner and outer formes of the same gathering. In view of the number of damaged types recorded, the chance that none of them would have so recurred is distinctly slight, if setting by formes had indeed been used. I did in fact record two highly questionable possible cases: i number 1 on L3v and L4v, and half-cap T 3 on L1r and L3v.[17] In each instance the identification is classed only as "possible", and doubtless the letters are different, and look similar by mere coincidence. Certainly neither identification is reliable enough for any suggestion that setting by formes was employed to be based upon them. In any event, the pattern of composition already established by other evidence rules out the idea that the book was set by formes. Indeed, it would have been surprising, in view of the battered state of the fount, if no such instances of doubtful identification had turned up.

Secondly, some indications of the order of printing can be derived from the re-appearance of damaged types in new setting. The thing to be established is the pages at which previously-used types re-appear, which will indicate when the forme in which the type first occurred was ready for distribution. The assumption has sometimes been made that an entire forme was distributed at a time. The evidence suggests that this was not done in The White Devil; consequently, the unit adopted here is the page, and even that can be of doubtful validity. In practice, the page to be distributed must be taken from the stone or chase a few lines at a time, and there is no physical reason why the distribution could not be shared by different compositors. The irreducible minimum unit of distribution, in practical fact (as any printer will acknowledge), is the single line, nothing more. However, it is likely enough that, with certain exceptions to be mentioned later, each compositor in the main distributed an entire page.

Thirdly, the career of the types through the book can be traced, and from this study it is possible to determine with some accuracy the order of distribution and setting of the pages. It is also possible to indentify the actual cases used. The assumption that each compositor worked from his "own" case is not one that it is safe to make for a shop like Okes's, but where type-evidence supports spelling and punctuation-evidence, it can be valuable additional evidence in the matter of compositor-identification.

However, before any of these purposes can be achieved, the types must be identified. The first major work to attempt type-analysis was Charlton Hinman's The Printing and Proof-Reading of the First Folio of Shakespeare (2 vols., 1963), and Hinman's methods have been followed by most subsequent investigators. However, as Peter Blayney (op. cit., p. 91) very rightly observed, the quantity of types that proved sufficient for Hinman's purposes, working on a Folio forme, is quite inadequate for the much smaller dimensions of a quarto. For instance, in King Lear, Hinman's methods enabled him to identify an average of 23 types per forme, but in a quarto set seriatim, a comparable density to Hinman's would result in only about 5 types per page. Blayney's attempt to gather a quantity of evidence of a different order of magnitude can fairly be described as heroic. "Bodleian Library copy 1 of Lear was examined type by type under magnification. Every type which differed at all noticeably from the normal appearance of that character was noted by ringing the corresponding type on a photocopy of a facsimile. Over 7000 types were thus marked." He then repeated the procedure with the Capell copy, compared the results, eliminated inking flaws etc. and ended up with 6000. In the next phase, "many of the types originally marked were rejected . . . either because they were insufficiently distinctive or because there were far too many examples of nearly-identical appearance"; his final haul was "a list of over 2000 appearances made by 571 types", which gave him an average density of about 25 per page (op. cit., pp. 93-94). If his procedure had been as reliable as it sounds this would be excellent, but in practice the results he achieved do not bear out his claims.[18]

I suspect the density of evidence Blayney sought is simply not available. My working method was different: I kept the three British Library copies of Q1 before me as I worked, in order to avoid (so far as was possible) the process of assembling and then rejecting evidence that Blayney went through. I drew a detailed illustration of each letter that appeared irregular in all three copies, assigned it a number, and referred all subsequent and similarly damaged letters to the illustration, making direct comparison back and forth between the three copies as I went along. This procedure was not alone sufficient to secure an adequate standard of accuracy; in addition to the three British Library copies, I was obliged to seek confirmation also from the three copies in the Dyce collection in the Victoria and Albert Museum, and from libraries as far apart as the Bodleian, the Folger, and the Huntington. My list of queries includes types that were confirmed only after the consultation of the tenth copy, and some that resolutely refused to yield certainty. It takes an enormous amount of time, but yields more accurate results than Blayney's (which was partly imposed on him by the fact that no library owns more than two copies of the Pide Bull Lear). What it also does is demonstrate that the kind of density of evidence Blayney hoped to achieve is probably not possible. One must of course make all due allowance for the difference between the type used in Lear and that employed for The White Devil: the fount for the Webster play was larger, and the impression made by the type is less crisp

(and therefore subject to far more ambiguity) than that of Lear. Even with these allowances made, the result was at first sight disappointingly inferior to Blayney's: I ended up altogether with 734 appearances of 265 types, or 8.7 per page—but a number of these are of doubtful reliability. This has an important bearing on Blayney's observation (p. 91), "Unless the evidence-density is great enough to reveal at least two prior distributions in every quarto page, the order in which those pages were set cannot be proved beyond doubt". I believe he is right, but I do not believe that all it takes to provide such density of evidence is, like Boxer, to work harder. On the contrary, the more exacting the investigation, the more careful one becomes with unreliable evidence.

There are three different classes of damaged letter, which need to be carefully discriminated. First, there are those letters so individually and characteristically damaged that they are evident and unmistakable to the naked eye, and can even be made out on a xerox. Unfortunately, such types are likely to be noticed by the compositor or proof-reader, and discarded rather than distributed. In the present survey, for instance, of the 81 letters of this class which I found to recur, 41 recur once only. Secondly, there are types which have characteristic smaller flaws which can however be positively identified under magnification, if careful comparison of a number of original impressions is made. This is eye-breaking work, if it is to be done reliably; however, it ultimately yields firm results if persisted with.

The third class is where frustration sets in on a major scale. Types appear which are at first glance characteristically damaged, but which turn out to share the damage in a way that makes them useless: three lower case h on three successive pages, for instance, all so identically damaged that one cannot be distinguished from another. Letters which have a vertical joining a curve are most prone to damage at the join: b, d, h, n, m, p, q, r, u. Letters with a closed section will accumulate a plug of ink in the eye: a, b, d, e, g, o, p, q; this plug is sometimes permanent, and sometimes removed by washing: one can find a letter which looks identical in half a dozen copies, only to find in the seventh that the ink-plug has been cleaned. Letters with ascenders and descenders will sooner or later get bent: b, d, h, l, k, p, q, y. Anything with a kern will eventually be bent or broken: f, long s, and their associated ligatures, and j. Lower-case o is unusually prone to damage, with cracks at any point in its circumference; and unidentifiable damaged lower-case e are so frustratingly common that sometimes one feels one would do better counting the undamaged ones. Faults in impression must be firmly distinguished from damage to the type, which is sometimes difficult: if one copy shows a damaged letter where the others do not, is it a bad impression, or was the type damaged, and the copy with the flawed letter the last to be printed? Only if the type recurs can one be sure. Another common problem is the letter which, while undoubtedly damaged, prints differently in each copy.

It is time to cry "hold, enough": every investigator who works on damaged type will come to the realization that a lot of the "evidence" is worthless if really closely examined: I routinely eliminated from the investigation such

false fire as the ligatures long s-t, long s-h and ct in which the ligature is cracked at the top; lower case g in which the lower bowl is not closed; e and s with the ends of the letters worn; tilted or plugged e; plugged a; and a goodly number of other unusable letters. Even after all this elimination, I am left with a substantial number of identifications that must be classed as only "possible", which are listed with a question-mark in the Appendix of Damaged Types at the end of this article. These are letters that I think are the same, but which I will not rely on if other evidence seems to contradict them. Upper-case letters and numbers, italic sorts, punctuation and other typographical signs often offer clearer and more reliable kinds of damage than lower-case letters. But they recur so infrequently that they are seldom useful for the kind of analysis that must be undertaken. And they too can be susceptible to misidentification: for instance, capital I often develops a crushed serif when in conjunction with an apostrophe: compare the two 'I'le's on D1v, lines 1 and 2, which are virtually impossible to distinguish. With these caveats and reservations in mind, let us proceed to analyse what the evidence of damaged types can tell us, firstly about the order in which the pages of the book were distributed.

The results of this enquiry are found in Table 3 below, but some general observations are in order before seeing what the table reveals. First, it is apparent that the fount used to print the play was large enough to enable two entire gatherings (four formes) to be in type at the same time. This is proved by the fact that no distribution of gathering B was necessary in order to set gathering C.[19] I have not gone to the heroic extent of counting each type in these two gatherings, but at a rough estimate a full page of 37 lines in prose would contain almost 2300 characters (including spaces); the first two gatherings are about 75% verse overall, which would work out to around 27,500 pieces of type, or a good deal more than a "quarter bill".[20]

There are no special signs of type-shortage in the setting of gatherings B and C, or indeed, throughout the book. The use of italic or roman punctuation (or black-letter and roman periods) seems essentially motiveless (that is, the compositor(s) could not be bothered sorting them out, rather than that they are signs of type-shortages as such). There are, indeed, not a few wrong-fount punctuation marks apart from the black-letter periods (note for instance the comma in line 20, and the second comma in line 29, of B4r: the first clearly belongs to a fount of smaller size than pica, the second to a larger). The use of italic capital Y instead of roman is occasional throughout the play, but this seems to be the only sort in the fount in generally short supply; occasionally VV is set for W, in E2r, H1r (thrice), K3v (5 times) and K4r (twice); half-cap vv is set for half-cap w quite often in gathering E: E1r, E2r (thrice), E2v and E3r: the problem here is the speech-prefix 'Lavv.' for 'Lawyer'. Some of the long s-h ligatures may come from another fount: note

the instances in lines 1 and 12 of B4v, for example, which do not range with the other letters. Either a half-capital o or a zero (in the absence of numerals in the play it is impossible to be certain) is intermittently used instead of lower-case o; the first one is on B2v line 17, the second o in "consumption"; there are 32 altogether in the play, occurring in pages set from both cases. Elsewhere, a lower-case o may substitute for the half-capital: see the speech-prefix 'Lodo.', B1v line 10, for instance.

However, these isolated instances of petty carelessness hardly amount to evidence of general type-shortage; on the contrary, the absence of any more evidence than this implies that the fount must have been pretty large. One's initial reaction is, that since a single pair of cases could never have held enough type to set four formes, two cases must have been used, and two compositors therefore at work. (As an experiment, I counted the number of lower-case o in my case of 12-point Bembo. The box was about half full, and held 280 letters. Okes's sort-box—assuming it was about the same size—would probably have held between 500 and 600 letters before they started to slide over the rim into the box below. But there are 1508 lower-case o in gatherings B and C of The White Devil.) Of course the assumption is false: if most of the type was locked up in formes, the quantity in the case never need be more than the compositor immediately required. That is, the compositor would distribute into his case, as required, from the chase that held the type in its previous employment, and set B and C from that source. Obviously, by the time all four formes were in type, he must have been scraping the bottom of the barrel, an inference which is supported by the great cascade of types from B3v (which as the table shows was the first page to be distributed) which show up in D1r. But the method of working was entirely viable.

A second thing which the table makes evident is that as work proceeded, the types from many pages tend to reappear in clusters. Of the 84 pages in gatherings B-M, 23 pages yield types that reappear in some numbers on certain pages (those listed in column 4 in the table). Another 12 pages resurface in minor clusters; 31 pages do not cluster, and 18 pages seem not to have been distributed. The clustering effect is most obvious in the middle of the book, especially in gatherings E, F, G, H and I. Gatherings D and L have a number of pages with minor clusters. The inference is that a cluster occurs when a distribution has just taken place into a depleted case. One note of caution needs to be sounded. It is often assumed[21] that the types last distributed will normally be the first to be re-set. But this is contrary to the experience of those who set type themselves, and are trying to recover a misdistributed sort from a box. Because the cases are at an angle, and because type-metal is heavy, gravity ensures that mixing takes place. The likelihood of a cluster occurring, then, is increased if the case is getting low. Type appears to have been at its shortest in the first pages of E (in which types cluster on E1r, E1v, E3r and E3v) when C was distributed. The shortages at

the beginning of G, the middle of H, and the beginning of I are accounted for by simultaneous working of two compositors: the activity of the workmen ran the cases nearly dry in I. Subsequently, in K and L some mixing of the cases may have taken place (as is described below), possibly to ease shortages of sorts in one case.

Let us now turn to the table, and analyse the information to be obtained from it. It should be stressed that inferences concerning which case a page was distributed into can only be drawn from observing in which pages damaged types from the distributed page recur. Thus E1r was undoubtedly distributed into case x, since a large number of its types recur in G1v, set by Compositor A. Only one case was in use during the setting of gatherings B-D; this might be also called case x. But to do so would be misleading, since while much of case y was created by distribution, as described below, this practice would not ensure an adequate supply of all sorts. Types must have been imported from the original case to supply deficiencies during case y's establishment. I therefore believe it to be a more accurate reflection of the facts of the situation to regard the case in use for gatherings B-D as the ancestor of both cases x and y, whose identities do not become distinct until gathering C was distributed. And even then it is too early to say whether the distributor was Compositor N, or one or both of Compositors A, or B, or some other journeyman entirely. Not until compositors A and B begin regular work is it possible to assign the distribution of any page reliably.

TABLE 3 Type-Distribution

The first column is the distributed page; column two is the first page on which a type from that page definitely recurs; the third column lists any recurrence of a "possible" type earlier than that in column two; column four lists the page(s) in which a major influx of types from the distributed page is found; column five the pages with a minor concentration (i.e. at least two types). Column six lists the pages whose distribution can be assigned to either Compositor A or B; and column seven the case into which the page was distributed, where this can be determined.

A summary of the distribution of the book revealed by this table shows that the inner forme of B, beginning with B3v was the first to be distributed; the remainder of the inner forme followed promptly. B2v was probably the first page of the outer forme to be distributed (since a type turns up, doubtless as a proof-correction, in C2r), and was followed by the other pages of outer B. It seems likely enough, then, that inner B was machined first, followed by the outer forme; both formes of C seem to have been distributed at roughly the same time. This could mean (though it need not) that another workman was aiding in distribution. The first types from inner and outer D likewise show up simultaneously, at E4v. Then inner E is the first to appear, at F2v; outer E does not surface until G1v, but this gap is undoubtedly accounted for by the distribution now going into separate cases. It does, however, seem that E2r was distributed before the rest of the forme. F4r is the first inner F page to appear, as early as G3v; the outer forme was not ready for distribution until H3r; since both were distributed by Compositor B, it seems clear that a genuine delay was involved. By contrast, both formes of G were ready for distribution at the same time: types from both G1v and G4v appear on I1r, which, together with the rapid reappearance of other pages of G, confirms that type was growing short in this gathering.

At this point printing must have been moving ahead briskly; inner H was ready for distribution when I4v was still being set; the outer forme first appears on K1v, but thanks to the compositors' division of labour, the rest of the forme does not appear until a good deal later. I1r appears on K3v, in advance of any types from the inner forme, the first of which comes on K4r from I2r. Evidence begins to fade away at this point; there is a possible appearance of a type from K3v on L2r (if it is really the same type, it was probably a proofing-change), but the first reliable occurrence is inner forme K1v on L4v; outer K does not appear until M1v (K1r), and most of K does not turn up at all; nor does any reliable type from L. The conclusion is that the printing would appear to have been entirely regular, with the inner forme routinely being machined first. Outer I was perhaps ready for distribution before the inner forme, but they could also have both been available, and the compositor merely have chosen to unlock the outer forme first. In summary, there is no reason to think that The White Devil was printed in any way other than regularly, inner forme before outer, and that the pages were distributed as the compositor(s) came to them, in no particular order.

It is now time to return to columns six and seven of the preceding table. Let us take it that the normal method of operation was that a compositor would ordinarily distribute a page of type into his case, and then set his next page from the same case. Assuming this, it is possible to list in column six the workman in question, from gathering F onwards, and also in column seven to specify the case into which each page was distributed. Some anomalies surface as a result in the form of pages whose types turn up in more than one compositor's stint. These pages are G2r, G4v, and I2r. Readers will recall from Part Two of this article that page I2v was classed as doubtful, assigned by Williams to Compositor B but in this article re-assigned to A, despite the

presence in it of a number of B's preferred spellings. The types from G2r which occur in it are found in lines 4, 5, 11 and 30—well-scattered, in other words, and therefore not leading one to suspect that the distribution of G2r was shared, but rather that A distributed the entire page. The likely explanation is that the "possible" type from G2r found on I1r (a certain B page) is probably a false identification. In the same way, it is more likely that type m 2 from G4v is the type found in line 10 of I1r than that h 2 is the same type as in line 5 of H4r, which would entail the shared distribution of G4v. However, I2r presents a trickier problem. Even without wondering about the unassigned pages from gathering M, there is a conflict between K4v, an A page, and L1r, a B page. In this instance the plea of misidentification will not answer, since the type on K4v, long s 11, is quite reliably indentified, and the cluster of types from the same page on L1r is unmistakable. It will be necessary to return to this case subsequently. However, it is fairly clear that after Compositors A and B had settled into their stints of work, on the whole the expected pattern prevails: that a compositor distributes a page into his case, and sets his next work out of it.

With this evidence to hand, and a number of detailed analyses which I will refrain from reproducing here, it is possible to be pretty definite about the pattern of setting and distribution of type in the text of The White Devil. As has already been observed, gatherings B and C were set before any type was distributed from them. As we have also seen, there was no need for this to imply more than one case or more than one workman. It is difficult to be precise as to how his work progressed in gathering D, partly because by mere chance there happen to be very few identifiable types in outer B (19, including 8 doubtful, against 46, including 12 doubtful, in inner B). During the setting of gathering E, both formes of C were distributed at intervals; indeed there is a sudden increase in types from outer C at E3r and E3v. It has already been remarked in Part Two of this article that there is something slightly irregular about these pages: they share a number of idiosyncratic spellings against most of the rest of the play, and the figures for the compositor-discriminators in Table 1 show some signs of difference from those of the preceding gatherings.

It is evident that at some point in the composition of the play a second case was brought into use, and the likeliest place for this to have happened is when C2v was being distributed. This would account for the number of types from pages in gatherings B and C which are found in both compositors' work later in the play. The pages which seem to form the basis of case y are (besides C2v) D1v, D2r, D3r, D3v, and D4r (though there are some minor anomalies). As to whether Compositor B distributed all these pages, or whether another journeyman created case y and even perhaps set a few pages of E from it, the evidence is just not sufficiently clear to tell. As was made clear in Part Two of this article, there is no evidence of Compositor B's setting anywhere in the book before the last four or five lines of F2v.

On the basis of the evidence already described, it is reasonable to assume that divided setting began with gathering F. In Compositor A's first stint we

find types from C1v, C3v, C4r, C4v, D1r, D2v and D3v. Compositor B's case, by contrast, contained types from C2v, D2r, D3r and D3v. The only page whose distribution was shared, then, was D3v, and only one type from that page (M 1) appears in case x; the remaining types recur in case y. A's next stint begins with F4v and contains types from E1r (a large cluster on G1v), E1v, E2r and E4v, as well as some types from earlier gatherings. When B resumes at G3r, his next four pages contain types from D1v, D2r, D3r and D4r as well as more recent ones from E2v, E3v and E4r. Apart from the fact that Compositor B took over the last few lines of F2v, the division of labour has for these two gatherings been most regular, with A taking the first four pages and B the second. It will be appreciated from this description that (as was said above) the compositors were not each responsible for distributing a forme, but seem to have taken pages from the stone virtually at random.

Gathering H is divided differently, into pairs of pages: A set both rectos and versos of H2 and H4, and B of H1 and H3. As the manuscript must have been cast off and divided anyway, this method of working is no more difficult than the other. B must have distributed F3v before setting H1r, and A likewise distributed F1v and F2r before beginning H2r. B made another distribution, chiefly of F3r (from which five types appear on H3r) but also of F4r, and A in turn distributed F1r, F2v (which leave respectively four and two types on H4r) and F4v before the gathering was finished. So far the work has been a pattern of orderliness and regularity, and so it continues for the first three pages of gathering I, set by B from case y, into which he distributed G1v and G4r. Then Compositor A took over in I2v, which I re-assigned in Part 2 above from B to A. I2v contains numerous types from a distribution of G2r and G2v, but it also contains types from E3v, F4v and G3r, all of which were previously found in case y. The simplest inference would be that for some reason Compositor A was working from case y at this point: there is one type from case x in I2v: B 3 from D2v. But by this time, what with proofing and other possible causes of type-movement (such as accidental pying or stripping accidents), we ought not to take too seriously the anomalous appearance of a type from so much earlier in the book.

The problems increase in I3r, which shares types from case x (C1v, E1r, E2r) and case y (F4r, G2r). This page is undoubtedly of Compositor A's setting, and the likelihood, in view of the fact that G2r types appeared in number in the previous page, is that both were set from the same case. If this is correct, it means that while each compositor still set four pages of the gathering, some printing-shop exigency led A to forsake his usual case; in consequence six of the eight pages in the gathering were set from case y. This being so, it is not unlikely that some signs of pending shortage of sorts led Compositor A to transfer various types from case x to case y to replenish it, thereby causing some mixing of the identified types. This does not seem an unlikely scenario: when it happens in my printing shop, I usually raid the case alongside, to even up the numbers, and I cannot see any reason why Okes's workmen should not have done likewise. In such circumstances it is best to be guided by the more recent types, when trying to decide which case was being

used. For these reasons, I believe that I3r was also set by Compositor A out of case y. For I3v, Compositor A seems to have moved back to his usual case x, since the types in it do not conflict with any case y types. There are very few distinctive types in I4r and I4v, but what there are are not in any way inconsistent with their being set regularly, by B from y and by A from x.

K1r shares with I4v the first types from inner H, H4r, and was set by A from case x. K1v (B-y) has types from H1v, H3v and I1v on it, but only one (from E1r) from an earlier distribution. Although E1r was an x page, we may appeal to the sharing of types already postulated to account for this single anomaly, as we may for the E1v type in the next page but one, K2v, otherwise regularly from case y; there are no problems in K2r and K3r, regular B-y pages. K3v (A-x) however does contain an anomalous type from G4r (y) whose presence I am at a loss to explain. Shared distribution will not account for it, for another type from the same line is found on I1v, the page on which most of the G4r types recur. There are various possible explanations, all of which entail stretching the long arm of coincidence: such as that type l 5, though I believed it to be reliably identified, is actually two types so identically damaged that they cannot be distinguished; or that some minor chance in the shop led to letter l 5 being dropped by compositor B and picked up and distributed into case x by compositor A, or that it was used from case y by the proof-reader in correcting K3v. None of these hypotheses is impossible; none of them is susceptible of proof.

K4r is a normal A-x page, but an apparent problem arises on K4v, on which we find types from both cases: from G3r and G3v (y), and from H2v and H4r (x). Both pages were set by compositor A, and while we have seen how types from case x could have been used to fill up short boxes in case y, it is difficult to see why the reverse procedure should have been required at this point. However, the recurrence from G3v is a doubtful one, and we have already seen that Compositor A was setting from case y at I2v; it is therefore possible that the type from G3r was returned by him to case x as part of the process of making proof-corrections in I2v.

Another problem occurs on L1r, in all respects a normal B-y page, in which types from I2r appear in some number. However, type long s 11 (also from I2r) has just appeared on K4v (an A-x page) and another type from I2r subsequently appears on L2v. This last may be dismissed: the type occurs in the first part of the page, and was therefore set by B. But to account for long s 11, we will have to invoke some kind of coincidence similar to that required for type l 5 in page K3v. And something similar must be appealed to for the presence on K3v (A-x) of a type (n 20) from I1r, which otherwise seems to have been distributed by B and used to set L1v. These three somewhat intractable letters do not constitute an alarming amount of anomaly in view of the total numbers involved, and in any event effectively bring us to the end of the problems associated with the book. L2r is a B-y page; the composition of L2v was shared, and although it is likely that compositor A took over towards the end of the page, it is very difficult to say just where, or which case was in use at what time during this page. Compositor A's stint through the

rest of gathering L uses types mainly from I2v, I3r, I3v and I4v, pages which had not been used before; there are also some types from other case x pages, and one anomalous type from case y (from G2v, on L4r). But G is sufficiently far back that by the end of L some normal minor mixing of types would have occurred. The types that surface in the disputed gathering M are largely case y types, which strengthens the possibility (it cannot be called more) that despite the unusual spelling forms found therein, B might have set the first three pages of the gathering: there is a surprisingly large number of types from early gatherings, but also a few from I1v and I2r, both case y pages.

To conclude, it is reasonably safe to generalize that while the earlier gatherings shared between Compositors A and B were very regularly divided, there seems to have been increasing irregularity in their working from the middle of gathering I onwards. At this point A set two pages from B's usual case, and evidently "borrowed" some types from his own usual case to replenish case y. Later, at L2v, the compositors seem to have shared work on the page, and there are a few typographical anomalies in these later gatherings which elude simple explanation. Nonetheless, none of the difficulties with the evidence are so severe as to make it necessary to abandon the hypothesis that two compositors worked out of two cases for gatherings F-L of The White Devil.

Conclusion

I do not by any means claim this study to be exhaustive. I have certainly not been able to duplicate the scale of the unprecedented researches Peter Blayney undertook into the Pide Bull quarto of King Lear. Within my more modest limits, however, I believe I have gone much further into the physical evidence than has any previous work on Webster, and I have attempted to examine all the bibliographical questions which could shed useful light on the composition and printing of The White Devil. If, after turning over all the stones (some of them mightily massive and slow to be turned, at that) little has been found under some of them, that is in the nature of bibliographical investigation. To say this is not to say that I am disappointed with the results, nor that the time spent in the investigation was not well-spent. On the contrary, I am satisfied that the actual editing of Webster can now proceed with a better appreciation of just how much information concerning Okes and his workmen's share in the production of the text exists, and can be textually exploited.

What has been established, I hope beyond reasonable query, is first that compositor-determination in this quarto is unusually difficult because there are so many "concealed prose" lines—lines whose method of justification requires them to be treated as prose. Secondly, Philip Williams's division of the book's composition between two compositors in gatherings F-L holds up well (with only one page needing to be re-assigned, and another to be seen as divided between the workmen). However, his attribution of pages B-E and M is unreliable, and his belief that Compositor B set some of gathering B is surely mistaken. On balance, reviewing all the evidence contained

in this article, it seems to me highly probable that a single workman, whom I have nominated Compositor N, set B-E (though the possibility that another compositor, different from A, B, and N, set some of the later pages of gathering E, when case y was being established, cannot be absolutely ruled out). The evidence of punctuation, spacing, and (to a much lesser extent) spelling, suggests that Compositor N and Compositor A shared many practices and preferences, but it does seem possible to discriminate between them on statistical grounds. Some further confirmation comes from the treatment of speech-prefixes. Thanks to Blayney's researches, we know that Okes tended to employ a variety of journeyman compositors, and, in the circumstances, the probability that more than two men could have worked on The White Devil is quite high; and the prospects of reaching absolute certainty in discriminating between them are not as good as one would like. This obviously has a bearing on assumptions made about compositors and their fidelity to copy when actually editing the text.

Finally, thanks to an examination of damaged types perhaps more exacting than has hitherto been attempted, it has proved possible to trace the pattern of distribution and composition through gatherings F-L, thereby confirming the evidence of spelling and spacing. While this survey turned up some minor irregularities in the setting of later gatherings, by and large the quarto seems to have been printed very regularly. Perhaps one of the most valuable conclusions to be drawn from the exercise is that whereas the study of damaged types and their distribution and reappearance is a bibliographical tool of value, it will not in itself suffice to determine the compositors, or without some ambiguity, the precise order of setting. It is, in other words, a valuable tool, but not a panacea. It is also perhaps the most time-consuming and potentially frustrating bibliographical labor yet conceived, and should not be undertaken lightly in hope of easy and certain results.

APPENDIX RECURRENT TYPES

The following table lists all identified types in the first quarto. All lines on the page are counted (save the headline and blank lines), and the figure following the virgula indicates the precise letter within the line. Thus, to take the first example, B4r (33/2) indicates that the first identifiable letter a occurs on B4r, on the 33rd line, and it is the second a in that line. Where there is no virgula, there is only one such letter in the line. Corrected and uncorrected copies are indicated by "c" and "u". A query following the listing indicates a probable identification, but one which remains slightly uncertain even after examination of multiple copies; a double query indicates much the same degree of uncertainty, compounded by bad inking in all examined copies. An exclamation point indicates absolute certainty of identity. When no indication is given, it means that my investigations satisfy me that the types are the same, without being of that absolute certainty (usually a grossly damaged type) which would be beyond any question whatever (the group marked with a !). Altogether, this table lists the 734 appearances of 265 pieces of type in the quarto. I should add that while all reasonable care has been exerted to ensure that these listings are accurate, it has not been possible to re-check every item recorded here, and in the nature of such a job, it is likely that some errors of transcription and recording may have crept in. The listing is offered, then, as the lawyers says, E & OE.

Notes