University of Virginia Library

The Shrinking Compositor A of the Shakespeare First Folio
by
Gary Taylor

This paper will attempt to establish that Jaggard's Compositor A did not set any of the First Folio pages of 1 Henry IV, 2 Henry IV, Henry VIII, Hamlet, and Troilus and Cressida. This in itself may not seem too surprising: of these plays Charlton Hinman himself originally attributed securely to A only the x-case pages in Troilus, and a little more than one page in 2 Henry IV. What may seem more surprising is that the peculiarities of these five plays apparently can only be satisfactorily explained by the presence of three (or possibly even four) compositors, none of whom is seen else-where in the Folio.

Compositor A was first identified by E. E. Willoughby in 1932,[1] and it was long assumed that he and Compositor B were alone jointly responsible for setting the entire volume. Charlton Hinman's identification of C and D significantly reduced A's share of the volume,[2] and this share has since been further diminished by T. H. Howard-Hill's identification of F.[3] Our interpretation of what is left of A will be significantly affected by our answers to two remaining questions: whether A was responsible for the x-case pages in 1 Henry IV, 2 Henry IV, Henry VIII, and Hamlet, which Hinman attributed to C* (= C or A) and A* (= A or C); and whether he was responsible for the x-case pages of Folio Troilus and Cressida. It will immediately be seen that, if A was not involved in setting these plays, he never worked on a simple reprint, while

both plays which he set from printed copy (Richard II and Richard III) involve disputes about which quarto was used. Moreover, as part of the evidence of printed copy for Folio Hamlet and 2 Henry IV is based on Alic Walker's identification of A in those texts, re-attribution of those stints to another compositor might partially undermine that hypothesis.[4] In any case, doubts about the Folio copy for those two plays—and it is possible to have doubts, even if A's presence is accepted[5] —in themselves seriously complicate any identification of the compositors involved. Adding to these difficulties is the fact that, although the problems of attribution in these five plays obviously interrelate and overlap, the solutions (apparently) do not: there is no simple and single piece of new evidence which neatly isolates or excludes A in all five. Consequently, we must first approach each problem separately, only gradually relating these tentative individual solutions to each other and to the larger pattern of Folio composition.

Hinman himself later retracted his doubts about A's presence in 1H4,[6] but he gave no reason for his retraction, and the arguments on which those doubts were founded have never been refuted. However, it is now possible to rule out entirely C's participation in any of Hinman's C* and A* pages, or in Troilus, for none of those pages contain the frequent terminal spaced commas in short lines which Howard-Hill (pp. 68-69) has since identified as a near-infallible indicator of C's presence. But the elimination of C does not guarantee the presence of A, particularly since Howard-Hill's discovery of F in the Comedies has opened up the possibility, as yet unexplored, that another compositor, in some ways indistinguishable from A, may have set portions of the Folio histories and tragedies.

Hinman himself had no reservations about assigning A the non-B pages in Troilus. Robert Lawson doubted the attribution to A, but this denial was part of a larger thesis which, being founded on methodological quicksand, has found few supporters (though some of Lawson's individual observations have been scavenged or expanded).[7] Similarly, the doubts of A. S. Cairncross about A's presence have been generally dismissed as part of his tendency to see Compositor E everywhere.[8] Alice Walker did notice that 'the work of the normally-reliable Compositor A went to pieces' in Troilus,[9] but she simply attributed this to haste[10] —an explanation which is of course circular, in that virtually the only evidence for haste is the deterioration in A's performance. Fluctuations in a compositor's standard of accuracy are by no means presumptive proof that his identification should be revised (as Paul Werstine's study of B has shown[11]), but in this case the really remarkable contrast between 1 Henry IV and Troilus coincides with changes in a number of important spelling habits.[12] Moreover, it must be remembered that the orthodox identification of A in Troilus is based upon a few spelling variants which, we now realize, could be the work of D or F as easily as A. Therefore, although there is a substantial case for the presence of an A-type compositor, there is no real evidence which specifically points to A himself rather than D or F.

The problem pages in Troilus must immediately be divided in half, by a count of spaces after medial commas in short (unjustified) lines.[13] Two distinct patterns (one indicated by a dagger) seem discernible, as is shown by a list of case x pages, in order of composition: ¶4 10 spaced/ 8 not spaced; ¶4v 32/9; ¶5 29/7; ¶5v τ3/21; ¶6 22/10; ¶6v 38/15; ¶¶3v 36/34; ¶¶3 26/23; ¶¶4v (lower b) 3/3; ¶¶2v 36/18; ¶¶5 22/22; ¶¶2 17/21;

¶¶5v 22/20; ¶¶1v 12/11; ¶¶6 24/32; ¶¶1a 10/14; ¶¶1b τ2/9; ¶¶6va τ4/34; ¶¶6vb 13/11; ¶¶¶1 †8/28; χ1v †1/18

Alongside one compositor who mixes spaced and unspaced commas freely, with the spaced clearly predominating on some pages, there seems to be another who very strongly preferred not to space, by a ratio of about 4-to-1 or better. Compositor C prefers to space after commas in short lines; A, D, and F do not. Since such psycho-mechanical evidence is (generally) of the most reliable kind, we are justified in tentatively conjecturing that two different compositors worked from case x on Troilus; and moreover, since A, D, and F prefer not to space after medial commas in short lines, while C (who does) has already been ruled out by the absence of terminal spaced commas, we are also justified in conjecturing that one of these two compositors is an as-yet-unidentified workman who set sigs. ¶5v, ¶¶1b, ¶¶6va, ¶¶¶, and χ1v. In what follows I will refer to the spaced pages as 'H1's' and the non-spaced pages as 'H2's'.[14] First I will consider the evidence in these two stints separately, beginning with H1; and then I will try to relate them.

Compositor A, in the stints now universally attributed to him, set the noun devil, plural or singular, 20 times. All 20 times he capitalized it, once against copy (R3 696). Moreover, if A was B's partner in 1H4, he set the word another 15 times, capitalizing it each time, 6 of these times against copy. In Troilus H1 set the word 9 times, 8 of those times without a capital; on all 9 occasions, he accepted the form of his Q copy. Disregarding 1H4 and Troilus, Compositor A on 19 occasions spelled devil (singular, plural, or adjective devilish) de-; only 4 times did he spell it di-, 3 of these involving the adjective and 1 following copy (R3 2031). In 1H4, B's partner used de- all 15 times, all 15 against copy. In Troilus, H1 used di- all 9 times, in each case simply reproducing the spelling of his copy. A similar pattern emerges with the word heaven(s). Compositor A consistently capitalizes it: 80 out of 82 times, 13 of those 80 against copy. In 1H4, B's partner capitalized the word 3 times against copy, only once following his copy's minuscule. In Troilus, H1 capitalized it on only 2 out of 16 occasions: once at the beginning of a verse line (where it had to be capitalized; this instance must therefore be discounted), and once following copy.

Two conclusions seem to follow from this evidence. First, whoever was B's partner in 1H4 did not set the H1 pages in Troilus; second, A

did not set the H1 pages in Troilus. The deduction that A was B's partner in 1H4, though suggested, is not of the same order of certainty, because it depends upon a mere agreement of those patterns in 1H4 with patterns elsewhere in A's work, and it is possible that another compositor besides A might share those traits.

Further support for re-assigning the H1 pages comes from the word young. A had a very strong preference for young (48 times, 9 against copy, to 2 for yong, once following copy); H1 in Troilus spelled it yong on all 9 occasions, 3 of these representing departures from copy young, and 1 of these departures occurring in an unjustified line. A also had a strong preference for deare (46, to 4 deere); in the H1 stints deere occurs 13 times, and although all but one of these follows copy, compared to 4 for all A's other stints the number is still remarkable.

                     

		1H4 (C-or-A)	A	TRO (H1)	TRO (H2)
Devil	{capital	15	20	1	0
	{minuscule	0	0	8	0
	{de-	15	19	0	0
	{di-	0	4	9	0
Heavens	{capital	3	80	1	0
	{minuscule	1	2	14	2
	{young		48	0	0
	{yong		2	9	2
	{deare		46	4	0
	{deere		4	13	0

It thus seems clear that A was not responsible for the H1 stints, and it can now be demonstrated that neither F nor D nor C was responsible for any of the x-case pages. C can be quickly ruled out, since the absence of terminal spaced commas is reinforced by the virtually complete suppression of his preferred form heere, which is altered to here 8 times, and retained only 3 times (once justified). The non-B pages of Troilus also show a clear predominance of the traits used to distinguish D from F: consistent use of the long form of the catchword (F is inconsistent); consistent centring of the second line of stage directions (F is inconsistent); 15 retentions (5 unjustified) of copy sweete and 16 alterations (6 unjustified) of copy sweet to sweete (F has an absolute preference for the short form); 24 retentions of copy -le in heele, sheele, weele, youle (F has an absolute preference for -'ll). The spellings of beauty, heavy, pitty, and pretty also argue strongly against F.[15]

Alongside this evidence against F or C is (taken in conjunction with the spacing pattern) similarly convincing evidence against D. There are no occurrences of anie, verie, euerie, ladie(s), or eie(s), though D elsewhere prefers these forms, often altering copy to impose them. The form 'ith, which D elsewhere does not use, occurs three times (though all derive from copy). D prefers mistris; in Troilus the word is spelled mistresse all 3 times, once against copy. D elsewhere has an absolute preference for maide and meete; at Troilus 1824, an unjustified line, the Folio has Maids (Q1 Mayds), and twice it prints meet against copy meete (though on both occasions in justified lines). The number of indented flow-overs is also far fewer than we would expect: 2 in 14 pages, as against an average of 2 or more per page in the Comedies. These suspicions are compounded when D's consistent accuracy in the Comedies is contrasted with the volume of unanimously-rejected readings in the x-case stints of Troilus: 171, by my count.[16] Even granting à la Honigmann that some of these may be authorial variants and first thoughts faithfully reproduced in the Folio but rejected by the editorial tradition,[17] a very large proportion of these unanimously-rejected readings are unmistakably errors, and the number and kind of such errors is markedly uncharacteristic of D's work elsewhere.[18]

It thus seems clear that the H1 pages could not have been set by any known Folio compositor, while the H2 stints could not have been set by C, D, or F, the three A-type compositors. But there is equally good evidence that the H2 stints were not set by Compositor A, either. Compositor A prefers deed(s) to deede(s) by a margin of 38 to 3; but H2 in Troilus set deed(s) only once—in a justified line, and following copy (1688)—and deede(s) 5 times, 4 in unjustified lines, and 3 times against copy deed(s). Second, Compositor A capitalizes Heaven 80 out of 82 times, 13 against copy; but H2 in Troilus, encountering the word twice, twice let Q1's minuscule stand. Third, Compositor A prefers young to yong by a margin of 48 to 2; but H2 in Troilus once changed copy young to yong (2064), in an unjustified line, and in the word's only other occurrence retained copy yong (3444, justified). Fourth, Compositor A prefers traytor to traitor by a margin of 42 to 4; but H2 in Troilus, setting the word four times, twice accepted copy i, and once—in an unjustified line—altered copy y to i (3435).

At this point, enter the computer. After this investigation was completed

(manually), I was able to test its conclusions with the help of computer-prepared concordances to the stints of the relevant Folio compositors. These concordances were prepared by pooling tapes belonging to Oxford University Press (those prepared for the Oxford Shakespeare Concordances) with others privately prepared by Dr Trevor Howard-Hill. Mr Lou Burnard of the Oxford University Computing Service supplied the programming and Miss Christine Avern-Carr helped him prepare and check the text tapes. Concordances were made of the stints of all the Folio compositors, with separate concordances to the stints under examination in this essay.

The computer turned up a good deal more evidence casting doubt on whether A set the H2 pages in Troilus. H2 set sweete 13 times (5 against copy) and sweet 3 times; A preferred sweet by 58 to 1. H2 set duetie in the word's only occurrence, against copy (1628); A set dut(ie/y) 23 times, and duetie never. H2 set maister twice, once against copy (3535), and master once, following copy (39); A set master 79 times, and maister never. H2 set Voice once, following copy (88); A preferred voyce by 42 to 3. H2 set beate(s) both times he met the word, once against copy (1668); A preferred beat(s) by 23 to 1. H2 set months for copy monthes (3588); A spelled moneths 8 of 8 times. H2 set bloud 5 times, once against copy (1601), and blood never; A preferred the latter 120 to 7. H2 set goe 5 times, only once against copy, and accepted copy go 4 times; A preferred goe 189 to 9. H2 set meete, plucke, and indeede the only time he met each word, the last against copy; A strongly preferred all 3 shorter forms. Considering that there are only 3219 words in the H2 stints, the bulk and consistency of this evidence is remarkable, and the conclusion that H2 cannot have been Compositor A seems indisputable.

It therefore appears that neither the H1 nor the H2 stints in Troilus can have been set by any known Folio compositor. But the possibility remains that H1 and H2 were in fact the same man; so far the only evidence which distinguishes them is the use of spaced and non-spaced medial commas.

H1 seems to be more tolerant of copy do: in his stints it is accepted 9 times (2 justified), but it is never accepted in the H2 stints. (In A's work do occurs only 26 times, and doe 376.) But the force of this evidence is somewhat diminished by the fact that H2 only encounters copy do 9 times, while H1 encounters it 52 times. In the H1 stints copy suddenly is re-spelled sodainely (2419); in H2's stints copy sodaine becomes sudden (75). But H1 also accepts copy suddenly once (3272). H2 seems more tolerant of copy go; as I have already said, he accepts it four times, only altering it once, while H1 accepts it four times and alters it 29 times. But this turns out to be poor evidence too, because in both stints

most of the accepted go spellings occur in the interjection 'Go to', which is never spelled goe; this applies to 2 of the 4 occurrences of go in H1 (1542, 1831), and 3 of the 4 occurrences in H2 (77, 1685, 1685). H1 seems to prefer shorter speech prefix forms. H1 shortens copy Pan to Pa 11 times, copy Nest to Nes 6 times, and copy Agam or Aga to Aga or Ag 14 times. Compositor H2 on the other hand once expands copy Aga to Agam (its only appearance in his stints), once expands copy Pan to Pand (elsewhere adopting copy Pan), and once retains copy Nest. All of this evidence is meagre, and would never in itself have been regarded as sufficient to justify a distinction between two workmen.

But in fact this absence of evidence is even more disturbing than it looks. All previous investigations of compositorial spelling habits have been based on manually-collected tests of certain sample spellings; in these circumstances, a failure to turn up spelling evidence may result simply from a failure to check the right spellings. But in this case I have been able, with the help of the computer concordances, to check every spelling in both stints; and there are no differences between them. (I have disregarded cases where a contradictory spelling appears in a justified line, or simply follows copy.) This absence of spelling distinctions would be difficult to swallow in any circumstances; it is absolutely incredible when, as here, we are able to test both quantitative and qualitative evidence, comprehensively. I am therefore forced to conclude that the evidence of medial spaced commas, usually so reliable, is not reliable here. This may be partly due to the inherent difficulty in distinguishing, in some cases, whether or not a space has been used; the test is sometimes more subjective than it appears. Moreover, even by my original counts, the proportions fluctuate considerably: from 24/32 to 32/9 for H1, and from 1/18 to 8/28 for H2. The x-case compositor in Troilus may have been affected by a sparsity of space-types, or influenced by his printed copy, or been inconsistent about this particular feature of his setting. Whatever explanations we advance (and however these might in turn affect attributions elsewhere in the Folio), I think that the spaced-comma distinction must here bow to the complete absence of supporting evidence. H1 and H2 thus seem to me to be a single new workman, H.

The problem raised by the four remaining plays, 1 Henry IV, 2 Henry IV, Henry VIII, and Hamlet, and the pages in them which Hinman attributed to C* (C-or-A), is in fact two separate problems, one of which can be solved much more easily than the other. The x-case pages in Henry VIII and Hamlet were clearly not set by Compositor A, and just as clearly not set by the compositor who worked with B on 1 and 2

Henry IV. Hinman's own very strong evidence against A's presence here is confirmed by speech prefixes as well as a number of spellings.[19] (Tabulations of A preferences do not include the H pages in Troilus.)                        

	1 & 2H4 (C*)	H8 & HAM (C*)	A
{long catchwords	9	0	43
{short catchwords	1	7	0
{King.	32	3	(over 100)
{Kin.	0	59	1j
centre-rule distribution	normal	anomalous	normal
{Devil(l)	17	0	19
{Divel(l)	0	5	4
{heavie	4	0	24
{heavy	0	1	1
{soldier(s)	0	1	0
{souldi(e/ou)rs	12	0	90

In addition, there is the high frequency of heere spellings (27, to 12 of here), which Hinman remarked on in the H8 and Hamlet stints. But, as I said initially, C himself is ruled out by the absence of terminal spaced commas in these pages. C, D, and H consistently use long catch-words. F is ruled out by the clear preference for -le instead of -ll. D, F, and H, like A, strongly prefer here. The only Folio compositors yet identified who prefer heere are B, C, and E, and they cannot have set the C* pages. Moreover, what is needed is a compositor who, though he prefers heere, has a high tolerance for here. There thus seems no alternative but to posit yet another Folio compositor, I. About this man's preferences more no doubt could and should be discovered; but there is already enough consistent and significant evidence over a sizable portion of text to make it quite clear that the 'C*' pages in Hamlet and Henry VIII could not have been set by any known Folio compositor.

A concordance of suffixes turned up another valuable distinguishing trait for I, one which helps distinguish him not only from A but from every other known Folio compositor. A concordance was first run of the letter combinations nesse, ness, and nes at the end of a word; then these were sorted by hand to remove any words where one spelling or the other would have been misleading or impossible (e.g. 'lionesse', where 'liones' would be ambiguous). The figures for this suffix for all Folio compositors were:

                   

	-nesse	-nes
A	243	1j
B	815	20 (+34j)
C	132	10 (+10j)
D	59	2 (+4j)
E	163	13 (+2j)
F	35	1 (+1j)
H	44	1j
I	54	20 (+3j)
C* (1 & 2H4)	30	0

I is also unique on a similar test for the suffix variant ie/y. After the computer had run a concordance of the letter combinations -ie and -y at the end of a word, these were sorted by hand to remove all monosyllables, words ending in -ay or -oy, and a few other words like hereby and thereby (where the suffix is equivalent to the monosyllable by) and obey, grey, and convey (often spelled -ay). The results were:                    

	-ie	-y
A	1014	1920
B	2178	8057
C	503	1056
D	506	498
E	429	1531
F	60	438
H	104	314
I	23	502
C* (1 & 2H4)	217	404

This test clearly distinguishes between A and B, between D and all other Folio compositors (particularly F), between A and my proposed compositor H in Troilus (A =1/<2, H = 1/+3), and between I and every other Folio compositor. The closest any other compositor comes to I in this respect is F, whose 1-to-7 ratio is not really comparable to I's of 1-to-22.

A third suffix test also proved useful in distinguishing I from A, D, H, and B's partner in 1 and 2 Henry IV. This is the triple variant -ies/-yes/-ys.

                   

	-ies	-yes	-ys
A	205	323	9
B	945	761	36
C	283	127	19
D	141	38	21
E	248	127	1
F	52	31	1
H	29	32	2
I	51	33	1
C* (1 & 2H4)	40	66	0

Only A and C* actually prefer the -yes spelling; D is unique in his relatively high tolerance of -ys. I is clearly distinct from A (2-to-3 against 5-to-3), and H does not look like A either.

The distinction between A and I, and between I and the C* compositor in 1 and 2 Henry IV, is important not only in itself but also because it will affect our attributions of another four pages in 2 Henry IV and Hamlet, which Hinman assigned to A or A*. Hinman himself had grave doubts about the two pages in Hamlet, and though he expressed no such reservations about the 2 Henry IV pages, his identification there was based upon the same negative reasoning: the absence of any traits which would require an attribution to C. In other words, Hinman presumes A's presence, unless there is positive evidence to the contrary.

I will return to the 2 Henry IV pages in a moment, but the Hamlet pages, nn6 and nn6v, can be dealt with immediately. Staid occurs twice in the two pages, at 522 and 435, both in unjustified lines, both representing possible departures from copy (Q2 stayed, stay'd). A has an absolute preference for stayd; of the word's nine occurrences in his stints, only one is justified, and one represents a departure from copy (R2 1285). Compositor I set the word only once elsewhere, but there he spelled it staid, where Q2 has stayd (Hamlet 988). Similarly, at 538 F has ranck, a spelling which never occurs in A's stints; A spells ranke on each of the 8 occasions he encounters the word, once against copy (R2 1213). Compositor I met it twice elsewhere, once spelling ranck (H8 535) and once ranke (HAM 911). Since the attribution of these two pages to A rests wholly on negative evidence, staid and ranck seem to me sufficient to justify assigning them to I. Certainly, for anyone who believes that the Folio was set from Q2, the evidence for I is compelling; but even without such an assumption, it can rightly be maintained that we have no real reason to suspect A's presence in these quires at all, so that even a single piece of good evidence would be quite sufficient to confirm that these pages, like the rest, should be assigned to I.

With Troilus, Henry VIII, and Hamlet behind us, we can now turn to the most problematic of Hinman's C* attributions, those in 1 and 2 Henry IV. Hinman's case against A was based almost entirely on 3 words, or more specifically 3 spellings: Go-too, houre, and heere (I, 374-377; II, 84, 101-103). Go-too, with a hyphen, appears 5 times on χgg1v; A never uses the hyphen. As the hyphens do not appear in Q1, we must assume either that A did not set the page or that Q1 for 2H4 was not used as copy (or both). A also prefers goe (189) to go (9, of which 7 are justified). As for houre, on 2 occasions (1 in 1H4, 1 in 2H4) it is substituted for copy howre. Although A will tolerate houre, Hinman thinks it unlikely that A would prefer it to copy howre, which coincides with

his own preference. He may be right; but in fact A, as currently defined, does not (quantitatively) prefer howre. He spells howre 17 times (1 justified), 6 representing a departure from copy; he spells houre 34 times (5 justified). In itself the houre can tell us nothing.

The real difficulty is created by heere. In quire e (1H4), B's partner never altered copy heere to here; on 8 occasions (4 justified), he instead substituted heere for copy here. In quire f, he twice in unjustified lines altered copy heere to here, but 3 times (once unjustified) substituted heere for copy here. In quire g, he substituted here for copy heere only once, in a justified line, making the reverse change 7 times (6 justified). In χgg3r-4v, he set heere 10 times, 9 of these being against copy, and 6 of the 9 occurring in unjustified lines. In the two plays as a whole, B's partner set heere 33 times (only 17 justified), and here only 50 times.

A set here(s) 23 times in Richard II, 11 of these against copy. Only once did he spell heere, on which occasion he eye-rhymed it with neere. Throughout his stints he spelt heere only 27 times, 18 of these in justified lines. He spelled here 291 times (only 51 justified). His preference for the short form in unjustified lines is thus 240 to 9, or over 26 to 1. He never, to our knowledge, spelt heere when his copy had here except for the eye-rhyme with neere and once in a justified line in R3 (1675). Consequently the 8 heeres in quire e are extremely irregular, to say the least, particularly because for 1H4 there can be no doubt about the Folio copy. The evidence from quire f would be less disturbing in itself—since it only involves 1 change against preference in a short line, which could be dismissed as a mere fluke—but in the context of the quire e evidence, it is hard to ignore. Likewise, one could claim that 2H4 was not set from Q1, and thereby explain the Folio heeres as due to the influence of scribal copy, but the concentration of these spellings in χgg3r-4v—6 in unjustified lines, as compared to 9 in unjustified lines in all A's other stints combined—is extremely difficult to dismiss, particularly in the context of similar anomalies throughout quires e-f. Moreover, 2 further spellings, which were originally used to distinguish F and A, also contradict the assignment to A: mistris, appearing 8 times, as opposed to only 1 occurrence of A's strongly preferred mistresse (1H4 2099 (j); Q5 mistris), and indeede, 12 times in 1 and 2 Henry IV, as opposed to only 2 (justified) occurrences of A's indeed.

Several speech prefixes corroborate the doubts about A's presence in 1 and 2 Henry IV. In 2H4 B's partner set Glo 8 times; only 3 of these could be attributed to Q1, which has Hum in the other 5. But Compositor A elsewhere set Glost or Glouc 88 times, as against 20 appearances of the shorter form—and of those 20, 18 are from copy (R3) and the other 2 are in very crowded lines (1H6). Likewise, B's partner in 2H4 set War

18 times. Unfortunately, all of these agree with Q1, so that the value of this evidence is partly dependent on our hypothesis about the Folio's copy; but if F was not set from Q1, it is certainly significant that A elsewhere preferred the longer form Warw by a margin of 63 to 18. Of those 18 appearances of War, 6 are in long lines, and 13 are concentrated on a single page (q2v). B's partner in 1 and 2H4 set West 20 times (19 in 2H4, all agreeing with Q1), and Westm only twice; A set Westm 8 times (7 in 3H6, set immediately before and after 1 and 2H4) and West 6 times (all in Henry V, set 5 quires earlier). B's partner in 2H4 set Clar 10 times, and Cla twice; only 1 of the shorter forms could be attributed to Q1 copy. Compositor A set Clarence 12 times, Clar 14 (many in crowded verse lines), and Cla never. B's partner set Hostesse 17 times, and Host 30; the longer form in each case expands his copy, but it is nevertheless significant that Compositor A in Henry V never used the shorter form. If these speech prefix preferences were our only evidence for a different compositor, they would not be compelling, but in conjunction with the spelling differences it is difficult to disregard the fact that B's partner in setting these two plays consistently preferred shorter abbreviations for these 5 names than Compositor A did elsewhere, particularly in the plays he set immediately before and after the composition of 1 and 2 Henry IV. If 2 Henry IV was not set from quarto copy, this evidence is even more significant.

                   

	A	1 & 2H4 (C*)	I
{heere	9(+18j)	16(+17j)	20(+7j)
{here	240(+51j)	37(+13j)	8(+4j)
common spp	longer	shorter
{mistris	1(+1j)	5(+2j)	2
{mistresse	17(+6j)	1j	0
{Indeed	23(+6j)	2j	17
{Indeede	7(+5j)	5(+7j)	1
{Deed(s)	38	1j	6
{Deede(s)	3	4(+1j)	0

Unfortunately, I have found no hard psycho-mechanical evidence which would differentiate A and the C* compositor here. In respect to unspaced internal commas, unspaced terminal commas, positioning of exit directions, and flow-over verse lines, the x-case pages in 1 and 2 Henry IV are indistinguishable from A's work elsewhere. But such evidence, though convincing when we can get it, is necessarily confined to a handful of binary options, and it is mathematically inevitable that some Jacobean compositors must have shared all the same psycho-mechanical preferences—especially as some of the alternatives, like terminal spaced commas, are rare. Moreover, in regard to one such psycho-mechanical

trait, catchwords, the agreement of A and C* is not so perfect as it might appear: if A did set these C* pages, then e4 of 1 Henry IV is the only time out of 53 opportunities when A set a short catchword. Flukes, of course, do happen; but concatenations of flukes should not. The speech-prefix evidence is midway in reliability between psychomechanical and spelling habits, and it shows a consistent difference between A and C*. Finally, the spelling evidence is more convincing than most such evidence, insofar as for 1 Henry IV we know the spellings of C*'s copy, and we are able to compare these not only with A's overall preferences but also with his reactions when setting from printed copy. Moreover, in the case of here/heere we are dealing with a very common word and a very strong preference, shown in A's setting before and after 1 Henry IV, from both manuscript and printed copy. And although the copy for 2 Henry IV remains disputed, we should not let this unresolved ambiguity blind us to the evidence from the Folio text of that play. In some respects the evidence against A is stronger if the Folio was set from printed copy; in others, it is stronger if printed copy was not used. This means that, however the issue of copy is eventually decided, some of the spelling and speech prefix evidence against A will remain intact. And in respect to heere/here, the simple quantitative count is unmistakably anomalous anyway, regardless of the copy.

After the evidence above had been collected manually, the computer turned up a great deal more. (This list is not exhaustive, by any means.)

                                     

	C* (1 & 2H4)	A
been	1 (+1j)	28 (+10j)
beene	21	57 (+11j)
behind	1	12
behinde	5	6
deed	1	28
deede	4 (+1j)	2 (+1j)
he	39 (+19j)	689
hee	43 (+87j)	144
Iack	9	1 (+1j)
Iacke	17	0
me	179	854
mee	18	17
meet	4 (+1j)	46
meete	8 (+1j)	5
need	3 (+1j)	33
neede	7 (+1j)	13
pluck	1	21
plucke	0	2

                                   

(al)ready	2j	21
(al)readie	6 (+4j)	14
speed	3	14
speede	3	4
sweet	4 (+6j)	58
sweete	1 (+4j)	0
ta'ne	4	0
tane	0	1
thin	0	1
thinne	1 (+2j)	0
war	14 (+4j)	14 (+7j)
warre	10 (+2j)	54 (+22j)
we	17 (+10j)	500
wee	41 (+22j)	37 (+36j)
wind	0	7 (+1j)
winde	7 (+1j)	12 (+1j)
ye	11	28
yee	4	3

Most of these involve a greater preference for redundant final -e, which is especially remarkable for the group hee/mee/wee/yee. (Compositor I, incidentally, agrees with A in preferring the shorter forms—further evidence of the need to split Hinman's C*.) In my opinion, this new evidence makes it impossible to believe that A could have set the x-case pages in either of these plays.

It thus seems to me that Hinman's original doubts about A's presence in these two plays were justified, and that another partner for B should be sought. But if so, neither C, D, F, H, nor I will do. The objections against H and I have already been stated. We can accept C's presence only if we postulate the complete abandoning of his distinctive terminal spaced commas, and while there is a considerable gap between C's last appearance (in King John) and the setting of 1 Henry IV, to presume the loss of such a major trait we would need overwhelming evidence of the appearance of all C's other characteristics; for all intents and purposes, if C changed this habit between King John and 1 Henry IV, then the later C might as well be another man. In fact one of the most outstanding traits of B's partner in 1 Henry IV is his accuracy, whereas C's work with reprints in the Comedies suggests a compositor as careless as Compositor B, and prone to the same types of (memorial) error (O'Connor, 'Qualitative Analysis'). Consequently, even if C and B's partner in 1 Henry IV prove to share every other preference, it seems to me impossible to have any confidence that they are the same man.[20]

Compositor D can also be ruled out. C* has an absolute preference for Devil (17); D, an absolute preference for divell (17). C* capitalizes Heaven 10 out of 11 times; D never capitalizes it in 29 occurrences. D has an absolute preference for meete; the C* stints of 2 Henry IV have meet 3 times (all 3 against copy, if Q was copy). The evidence against F is just as compelling. F has an absolute preference for sweet; the C* stints of 2 Henry IV have the long form 4 times, once unjustified. Even more convincing, the C* stints on 12 occasions use the contraction -le, 3 of them in unjustified lines; F has an absolute preference for -ll. And D and F, after all, both strongly prefer here—as does H, B's partner in Troilus.

To summarize: the C* stints in 1 and 2 Henry IV cannot be assigned to A, C, D, F, or H. Neither can the C* stints in Henry VIII and Hamlet. However, the C* stints in 1 and 2 Henry IV cannot be assigned to the same compositor as those in Henry VIII and Hamlet (I). I therefore propose to identify the C* compositor in quires e-χgg as another new compositor, J.

We can now reconsider the pages in this part of the Folio which Hinman assigned to A. A's complete absence from quires e-χgg, involving the whole of 1 and 2 Henry IV, makes his presence in χgg2 and the end of χgg2v relatively improbable—especially when Hinman's attribution to A rested almost solely on the absence of evidence for C, coupled with the assumption that the Folio was set from Q1, so that the 5 heres on page χgg2 represented departures from copy heere. On the other hand, the 5 here spellings, the fact that the pages involved were the last set from case x before A's return, and that B took over from J for the middle of χgg2v, do lend plausibility to Hinman's attribution. However, the only evidence for A is the recurrent here, and if Q1 was not the Folio copy, the value of that evidence is severely diminished, because J clearly had a fairly high tolerance for here. Moreover, 5 spellings in the page and part-page Hinman assigned to A argue against his presence: indeede (1954, 1971), deedes (2126), and bloudie (1902, 1908). Compositor A preferred bloody or bloudy (to bloodie or bloudie) by a margin

of 36 (2 justified) to 7, substituting it once for copy -ie; on χgg2—if Q1 was copy, as the attribution to A presumes—both -ie spellings represent alteration of copy, whereas the single -y follows copy. Three of the 5 anomalous spellings, moreover, agree with J's preferences elsewhere. In the only occurrences of these words in unjustified lines in his stints, he spells bloody twice, but both times following copy (1H4 1925, 1956); more significantly, he spells deedes once (1H4 1966) and indeede three times (1H4 2565, 2H4 1545, 2558). In 1 Henry IV, indeede occurs four times against copy; the two indeeds follow copy. Given that the use of Q1 2 Henry IV as copy for the Folio remains conjectural (and in my opinion unlikely), the case for J in the disputed pages seems stronger than that for A. And the presence of hee and wee (13 to 3) confirms J's presence.

Though a more extended investigation would no doubt discover further distinguishing traits of H, I, and J, the basic identifications seem to me fairly secure. All three compositors make a single and fairly brief appearance in the Folio, working on a succession of bibliographically related pages from case x in tandem with Compositor B at case y. Each is, I suspect, a journeyman, only occasionally employed by Jaggard. Certainly, the work of all the minor Folio compositors now falls into a fairly coherent pattern, which fits the hypothesis of a succession of journeymen. A, C, D, F, H, I, and J all have their exits and their entrances; B alone abides, to be joined eventually by the apprentice E.

Our portrait of Compositor A has shrunk yet again; he is now restricted to work on ten plays. He played no part in the setting of the disputed-copy texts of 2 Henry IV and Hamlet, and none in Troilus and Cressida. None of the texts he set is a simple reprint; the 'normally reliable' compositor that Alice Walker described in Folio 1 Henry IV was not A but another man, who helped set only one other Folio text. Consequently, an editor's expectation of how much and what kind of errors to expect in A's work can only be constructed on an analysis of unanimously-rejected readings in the plays he set from annotated or manuscript copy. As this has not yet been done, A's reputation as a uniquely reliable and trustworthy compositor, whose work editors should be loath to emend, is almost wholly unfounded.[21]

APPENDIX I PAGES SET BY COMPOSITORS H, I, and J

                         

		Compositor H(1)
¶	TRO	1179-1307	¶¶2v	2350-2478
¶4v		1308-1438	¶¶3	2479-2607
¶5		1439-1567	¶¶3v	2608-2739
¶6		1698-1829	¶¶4v pt	2973-3003
¶6v		1830-1959	¶¶5	3004-3135
¶¶1 pt		1960-2025	¶¶5v	3136-3267
¶¶1v		2092-2221	¶¶6	3268-3395
¶¶2		2222-2349	¶¶6v pt	3461-3524
		Compositor H(2)
	χ1v	33- 126	¶¶6v pt	3396-3460
	¶5v	1568-1697	¶¶¶1	3525-3592
	¶¶1 pt	2026-2091

                                                     

		Compositor I
t4	H8	210- 336	x3v		3149-3274
t4v		337- 467	x4		3275-3398
t5		468- 593	x4v		3399-3463
t5v		594- 718	nn6	HAM	353- 479
t6		719- 843	nn6v		480- 610
t6v		844- 973	001 pt		611- 708
v1		974-1098	001v		743- 874
v1v		1099-1224	002		875-1000
v2		1225-1349	003v		1258-1388
v3		1482-1613	pp4 pt		3062-3081
v3v		1614-1739	pp4v		3082-3211
x3		3018-3148
		Compositor J
e4	1H4	1109-1240	g4v	2H4	989-1109
e4v		1241-1370	g5		1110-1239
e5		1372-1497	g5v		1240-1367
e5v		1498-1622	g6		1368-1488
e6		1623-1746	g6v		1489-1612
e6v		1747-1866	χgg1v		1743-1867
f1		1867-1996	χgg2		1868-1995
f1v		1997-2119	χgg2v pt		1996-2080
f2v		2244-2373			2116-2126
f3		2374-2492	χgg3		2127-2258
f3v		2493-2615	χgg3v		2259-2382
			χgg4		2383-2514
			χgg4v		2515-2642

APPENDIX II

The following charts are intended to provide a simplified graphic summary of some of the more important evidence for the distinctions proposed in this essay. Some kinds of evidence are much more important and reliable than others, and

some preferences much stronger than others, but the charts do convey at a glance the patterns discursively described in the necessarily-complex foregoing argument. Statements as to preference are based on computer concordances for all compositors.

The number of different preferences distinguishing each compositor from every other can thus be further summarized:    

H1 vs I : 6	H1 vs J : 7	H1 vs H2: 1
I vs J : 8	I vs H2: 5	J vs H2: 6

(Thus, the best-supported distinction is between J and the compositors in Troilus, Henry VIII, and Hamlet; the least-supported is that between H1 and H2 in Troilus itself, resting almost entirely on the psychomechanical evidence of spaced medial commas.)

(In addition, there is I's uniquely high tolerance for -nes, and uniquely high preference for -y over -ie.)

(J uses heere and here 33 and 50 times, respectively; C prefers heere 185 to 78)

Again, the best-supported identification, on a purely numerical count, is that of J. But in all five plays the evidence is incompatible with any previously-known compositors; even if, for some reason, the spaced medial comma test should prove unreliable in Troilus, so that H1 and H2 are indeed identical, there can be no doubt that the x-case pages in that play were set by someone other than A, C, D, or F.

Notes