University of Virginia Library

III. The Pattern of Annotation

Pollard's detailed study of the variants in Q1-5 and F drew attention to the large number of instances in which F restored a correct Q1 reading where Q3 had acquired an error, the latter originating either in the Q2 printer's copy for Q3, or in Q3 itself. In conjunction with the sweeping changes from typically authorial stage directions to typically theatrical directions made in F, this led Pollard to the conclusion that the printer's copy for F had been annotated with reference to the promptbook. F's accuracy in correcting Q3 could not be attributed to chance. Pollard also saw that such annotation could explain a number of Q/F variants which are unlikely to have arisen in the printing-house: F had acquired a promptbook reading which differed from that in Q. But this insight was not put forward as having much significance for the editor: Pollard explained most of the assumed textual variants between the promptbook and Q as actors' unauthorized departures from the true text, departures which had at some point been incorporated into the promptbook. The actual corrections of error in Q were therefore deemed to be few in number. Further, the process of annotation was declared to be 'very hastily and inadequately performed', words of caution that have sometimes been interpreted as words of denial.

In his new Arden edition, Peter Ure provided a detailed application of Pollard's analysis of variants, examining their implications for the authority of F. Ure suggested that Q1 was set from a non-authorial transcript which contained memorial errors; this unlikely hypothesis has since been questioned by both Charlton Hinman and Alan Craven.[13] Ure tends to discuss Q/F variants in terms of categories of F error—thereby assuming that the Folio variants are errors, which only need to be properly categorized. One such category is 'unauthorized editorial replacements'. These account for 'perhaps the majority of words of nearly equivalent meaning in the Folio', though 'some at least' can be ascribed to the actors (pp. xxi-xxii). This conjecture is not offered as an explanation for metrical sophistication (for instance), but simply as a reason for dismissing a group of variants that will prove of some interest, 'nearly equivalent words'.

'Unauthorized editorial interference' is, of course, a real possibility, with Richard II as with other Folio texts. Since someone had to go through Q3 altering it by reference to the manuscript, that someone might well have made on his own initiative other changes not warranted by the manuscript. But though Ure may be right in believing that editorial interference took place, he is almost certainly wrong about the nature of such interference. An editor might add stage directions which he believed to be necessary, or modernize obsolete or obscure spellings, or provide act and scene divisions, or remove profanities; he might even, according to some textual scholars, regularize speech prefixes, or substitute commonplace words for rare or arcane ones. But why should he deliberately and randomly insert words of 'nearly equivalent meaning' to substitute for those in his printed and manuscript sources? Such a procedure would be particularly inane if, as Ure contends, these adjustments were made as part of the process of consulting the promptbook in order to modify Q3: in other words, the 'editor' is looking at a manuscript which might contain such variants, but none of them actually derive from that source, being instead unnecessarily and uncharacteristically intruded by the editor himself. Ure does not properly distinguish between the sort of change an actor would make and that which would be introduced in the printing house, or by a literary 'editor'; he does not explain how actors' verbal substitutions got into the promptbook;[14] he explains neither the necessity for supposing this additional stage of contamination nor the motives behind such a strange endeavour. The effect of his suggestion is to move a vital group of variants away from the potential authority of the manuscript.

Ure's attitude towards the Folio text is most apparent in his comment on the F readings he admits as potential emendations. Ure claims that, whether he accepts them in his edition or not, 'I do not think we need entertain the possibility of prompt-book provenance for any of them' (p. xxii). Even though his own theory of memorial transmission means that Q1 is imperfect, even though F can occasionally have correct readings, F is only admitted to preserve promptbook readings when they are wrong. This is an especially striking conclusion in light of Pollard's original rejection of chance as an explanation for the considerable Folio restoration of Q1 readings: F should be as effective in correcting Q1 error as it is in correcting Q3, especially as neither the Q3 copy for F nor the promptbook would distinguish between these classes of variant. In

what amounts to a rejection of one of Pollard's most important and persuasive conclusions, Ure explains some of the F restorations of Q1 readings as chance, others as sophistication; 'a majority of other restorations may be ascribed to the unaided editorial acumen of the Folio' (p. xxiii).

Yet Ure is no conservative extremist: other editors have followed similar editorial procedures. Paradoxically, G. B. Evans in the Riverside edition admits that 'substantive readings which appear for the first time in F1 deserve some consideration as possible Shakespearean second thoughts' (p. 838), but in practice adopts fewer F readings than Ure. Evans recognizes not only that the promptbook is the most likely source for a number of F readings, but also that such readings are difficult to dismiss as corruptions of the text. The New Penguin text (1969) does not differ radically from the new Arden, but Stanley Wells similarly—if somewhat cryptically—remarks after discussing the variant stage directions, 'Other alterations were made, some of which may be considered improvements on Q1' (p. 270).

There are therefore two problems. First, the most respected commentators have cast doubts on F as a source of potentially authoritative readings. This scepticism begins with Pollard's criticism of the inefficiency of the collation of Q3 against the promptbook—cautionary remarks which have sometimes been more enthusiastically taken up than the implications of his more general observation, that such a collation must indeed have taken place. W. W. Greg described the collation as 'superficial'; this may be a fair summary of some of Pollard's words, but sounds more dismissive than Pollard intended.[15] Greg's view was endorsed by Walton (p. 243); Ure elaborates upon Greg. On the other hand, but still beginning with Pollard, some editors have recognized that F is likely to have rather more than the bare minimum of valid readings, and have noted the actual superiority of some F readings—without feeling they are in a position to adopt them. The view that F should be followed more often is a reasonable one, but it has been expressed only tentatively. Sympathetic editors have probably been deterred partly by lack of evidence and partly to avoid the accusation of opening the floodgates to personal whim.

Yet Pollard's comments are not the last possible word on the process of collation with the promptbook. Its linear progression can be investigated in some detail. The only such attempt has been Walton's study, which in a disappointing three-page discussion does not go beyond splitting

up the variants according to the five acts. Walton at least shows that (as might be expected) the annotation was not uniform: Q1 readings are restored less effectively in the second and third acts. The simple expedient of plotting F's restoration of Q1 readings as they occur against its perpetuation of Q3 error gives a more localized picture. This is shown in Table 1.

Anyone who suspects that the alternation between correction and perpetuation of error is random should try tossing a coin and plotting the

sequence of results. The average length of an unbroken sequence of heads or tails will be found to be about 1.5. This is far from the case in Table 1. This information alone allows us to establish that some parts of the text were more effectively corrected than others. Indeed some areas were completely corrected; others were untouched. If nothing else, this pattern points away from Ure's multiplicity of causes for Q1-F1 agreement against Q3. We see a single process, spasmodically careful and efficient. This can only make promptbook annotation more plausible.

Given this framework, the remaining indications of possible promptbook annotation can be set alongside it, as can contrary evidence of a failure to consult the promptbook. Each part of the promptbook manuscript would have been arranged in three vertical sections: speech-prefixes, text, and stage directions. All received attention from the annotator. Although he could have worked on the sections simultaneously, it is possible that he split his work into defined stages, handling the text separately from its adjuncts. For this reason, the evidence of speech-prefixes and stage directions is considered separately later, as is the evidence of another completely separable category of annotation, the marking of act and scene divisions. Many altered or added stage directions in F clearly derive from the promptbook, as do major Folio cuts. Editors agree on the need to adopt a number of F readings (though they disagree on whether those readings derive from the promptbook). The provenance of the alterations of God to heauen is uncertain, and needs investigating to establish whether some or all of the profanity had been eliminated in the manuscript. Some F variants appear to be manuscript misreadings; if so, they too derive from annotation. On the negative side, F occasionally attempts to emend Q3 without bringing it into line with Q1; elsewhere it perpetuates error in Q1-3 which cannot be expected to have stood in the promptbook. One further class of variants might be positive, negative or neutral in individual instances: F's relineation of Q3. Each of these categories of evidence will need to be discussed separately.[16]

Compounded Error

Whereas Table 1 expresses F's success as against its inaction in correcting Q3, it takes no account of actual failures. On nine occasions F compounds an error in Q3 which it apparently undertakes to correct. These are as follows:[17]

                         

	TLN	Q1	Q2	Q3	F1
1.1.71	82	spoke, or thou	spoke, or thou	spoke, or what	spoken, or
		canst worse	canst	thou canst	thou canst
2.1.18	659	whose taste the	whose state the		of his state:
		wise are found	wise are found		then there are
					sound
2.2.3	955	life-harming		half-harming	selfe-harming
2.2.103	1058	there no	there two		there
3.3.17	1601	our heads		your heads	your head
3.3.119	1706	princesse		a Prince	a Prince, is
3.4.24	1834	come	commeth		comes
4.1.41	1964	to see that	I to see the		to see the
5.3.63	2562	held		hald	had

This category may contain readings where the promptbook has been consulted but inadequately represented or even readings where the promptbook differed from Q1, though one would not expect either of these to occur regularly. In other respects the group of nine should indicate lines which escaped collation. If the annotator overlooked a quarto error, which was then noticed by the compositor setting up the line, the compositor could either consult the manuscript himself, or correct on his own initiative. Compound errors could, of course, only occur if the compositor chose the second of these two options, and the presence of such errors in Richard II (and other Folio texts) suggests that re-consultation of the manuscript in such circumstances was rare, or non-existent. Indeed it is possible that the manuscript could not be consulted, because it was not in the printing house.

The small number of compounded errors is itself significant when compared with the number of correct restorations of the Q1 reading. But to compare figures directly would be misleading, if the hypothesis of relatively well-corrected areas of text has any validity: one would need to examine in turn the situation inside and outside those areas.

Shared Error

Errors common to all the quartos and F will be a reliable guide to lines not checked against the promptbook—if such errors can be securely identified, and if they do not originate with Shakespeare himself. No such list can be definitive. Ours includes editorial emendations which are usually accepted, together with one convincing emendation (932) introduced by Kenneth Muir in his 1963 Signet edition,[18] and 2 proposed

readings (964, 2556) and a proposed alternative reading (1090) first put forward here.[19]                      

	TLN	Q1-F1	Emendation
1.1.163	168-9	obedience bids, | Obedience bids	Obedience bids
2.1.283	932	Iohn	Thomas
2.2.12	964	With . . . at	At . . . with
2.2.110	1065	Thus disorderly thrust	Thus thrust disorderly or Disorderly thus thrust
2.2.138	1090	Will the hatefull commons	Will the hatefull commoners or The hatefull commons will
3.4.11	1817	griefe	ioy
3.4.67	1879	you the	you then the
4.1.148	2068	Preuent it	Preuent
5.3.31	2528	my	the
5.3.57	2556	lest thy	lest

One odd variant does not quite fit into this category. Q1 has a self-evident dittography: F reads: F might have independent authority from the promptbook, though it surely gives the weaker reading. On the other hand, the simplest unauthoritative correction would have been to remove the first (nonsensical and extrametrical) 'not'. Perhaps an annotation was misunderstood by the compositor. In any case, this variant cannot confidently be put forward as a miscorrection resulting from failure to consult the promptbook.

Manuscript Misreadings

A number of errors in F are such that it is hard to imagine that they are misreadings of the printed quarto copy. They appear to be misread manuscript literals. These are easily explained as the annotator's misreadings of the promptbook or, if the book was a non-authorial transcript, as errors committed when the promptbook was prepared. In

either case they will indicate annotation (however mistaken). Variants of this kind are:              

1.1.57	62	doubled/doubly
1.2.42	259	complaine/complaint
1.3.28	325	plated/placed
1.3.71	368	vigour/rigour
1.4.28	602	smiles/soules
3.2.112	1470	beards/Beares
4.1.33	1956	simpathie/sympathize

Lineation

Changes to lineation can be made for various reasons. The commonest cause is the need to accommodate a long line of print to the narrow Folio column. Like other texts, Q1 sometimes preserves the manuscript feature of a run-on part line at the beginning of a speech. When F splits such a line, it proves nothing about promptbook consultation. It does not constitute negative evidence because the alteration could originate from the promptbook; alternatively, the promptbook may not itself have changed the lineation.

In one instance the Folio relineation is not only unsatisfactory, but also disturbs the following two lines. Q has:

The F compositor failed to realize that the abbreviated name constitutes a mid-speech part line, and set as follows: The relineation tells us nothing about promptbook influence. Even if F's expansion of the name and variant vpon derive from the promptbook, the compositor was himself almost certainly responsible for the relineation.[20] Once the name had been expanded, Q's first line would not fit F's measure, and the Folio compositor probably altered the line-breaks in order to fit his text into three type lines (the same number as in Q). He would hardly have done this if the annotator had indicated a line-break after Bullingbrooke (as in all modern editions). We can thus be fairly confident that the annotator did not 'correct' Q's lineation; but of course the annotator's manuscript may have treated 'Henry . . . hand' as one line—just as Q's manuscript had apparently done.

A similar relineation occurs at 2.2.92-94, TLN 1048-49:

The verbal variants in F derive from Q2 by way of Q3. Q's lineation would fit the F measure, but it seems that the combination of a part-line at the end of the preceding speech and the metrical irregularity of To day . . . there induced someone in the printing-house to interfere with the lineation. In other words, relineation is here a direct consequence of an earlier failure to consult the promptbook.

Almost all the discrepancies between the lineation common to Q and F and that followed in modern editions are due to the different licences observed in early texts. In just one passage both texts are recognized as being seriously mislined:

Line-breaks are required after too and vneuen in order both to regularize the metre and to establish the rhyming couplet. In Q1, 1074 appears at the end of sig. D4v. The book was set by formes, and the compositor had sufficient trouble with fitting the cast-off copy to his page to induce him actually to omit two full lines of text when setting the opposite (inner) forme of the same sheet.[21] Q1's lineation at 1074-75 is a printing-house expedient to save space. F thus perpetuates a layout which cannot have appeared in the promptbook. The same circumstances determine a less impressive example; in Q1 this occurs in the inner forme of sheet D, where the spacing problems were most acute. On D4 (1014) Q1 runs on a part-line at the end of a speech with the previous line; F does likewise.

The most obvious example of relineation deriving from the promptbook is in the adjustment to the text after the abdication episode. The changed wording saves a part-line, so the text is rearranged to take in the part-line at the end of the speech (2245-6; discussed below in Section V). Significantly, this is not the only case where verbal variation

accompanies relineation. Even conservatively-minded editors accept F in one such instance. The significance of F's relineation here is not that it must be a promptbook feature itself. But the relineation shows how far the F compositor needed to alter the text in order to accommodate the addition of the two words. Such a change needs a definite external cause, especially as the text in Q1 is not noticeably corrupt.

Another case is structurally somewhat similar:

Here editors sometimes approvingly repeat Pollard's view that F's Offence is 'pittifully weak'. To show that there is no offence metrically in Q1, Pollard explained: 'The words "make peace" are a cry of rage which can only be adequately rendered by giving to each the time of a full foot. The next two words are pronounced slowly, and after "hel" there is a slight pause marked by the dramatic comma, and then the next line follows with a swift rush' (A New Shakespeare Quarto, pp. 85-86). Despite this extraordinarily subjective apologia, Pollard may be right to accept Q1 as the wording in the printer's copy. If F gives the Q1 printer's copy reading, it is difficult to see how the Q1 compositor came to set the text he did. But Pollard and his followers require us to believe that an actor or someone preparing F both relined the passage and added to the wording, simply in order to eliminate a tetrameter in the preceding line.[22] Parallels for such an intervention are not easy to find, and those who were involved in the preparation of this and other Folio texts do not seem to have been unduly traumatized by the occasional tetrameter. If the Folio changes in these lines were deliberately introduced in the printing-house, then consultation of the promptbook provides a single, simple explanation for the change, an explanation which traces it to a source known to have been consulted elsewhere in this text; just as important, consultation of the manuscript provides a credible motive for the change. And if the F reading stood in the promptbook, it is perverse to attribute that variant to an actor when it could originate with the author himself. There might perhaps be some excuse for denying it to Shakespeare if F's reading was indeed 'pitifully weak'; but this is surely a case where Pollard's predisposition to blame the actors, or other agents of corruption, got the better of his critical sense. Offence gives the line a more positive ending; it brings out an otherwise latent irony, that the usual Christian virtue of making peace is here seen as a sin, an 'Offence'; it brings warre into more powerful juxtaposition with peace.

In these lines the relineation is a vital part of the evidence in support of promptbook annotation. Other instances of significant relineation do not involve verbal variants, and must be considered solely on their own merits. For example:

Line-breaks are required after liege and pleasd. Q1 is explicable: the copy probably ran on the part-line (writing 'Oh . . . pleasd' on one text line); the compositor began to do likewise, but had to improvise a relineation in order to adjust the lines to his measure. F makes a strange attempt to rectify this. But in Q3 the end of liege is just below the end of Vnckle. A carelessly-placed line-break mark, intended to split the line after liege, could easily be misinterpreted by the Folio compositor as an instruction to introduce a break after Vnckle instead. Once the compositor had run on the line after Liege, he would find himself unable to fit 'pardon . . . pleas'd' on the same line. His compromise guaranteed no repetition of the problem on the following line. Surprisingly, the failed relineation is best explained as the result of an authoritative annotation.

Where the Q1 copy evidently ran on part-lines, a correction in F can plausibly be attributed to promptbook annotation if it was not undertaken to allow for the narrower F measure. There are three such examples:

(It will be noticed that the second instance contains a transposition, apparently an error in F.)

Profanity

Each of the classes of variant considered so far as potential additions to the picture established by Table 1 is in isolation too small to be of much statistical significance. This is not so for the God(s)/heauen(s) variants. There are 58 lines with God(s) in Q1; 26 of these are emended to heauen(s) in F1.[23] Another revision of profanity is Would God / I would (4.1.117; TLN 2037).

Cuts

As there are no cuts introduced in F whose length is of 2-3 lines, the actual cuts may conveniently be divided into two groups: single-line cuts and those of four lines and more. The latter are almost certainly deliberate measures taken to bring the text into conformity with the promptbook. They are usually described as 'theatrical' cuts, as though this is all we know, and all we need to know. But to say that the cuts are theatrical in historical fact ( represent the text as performed) does not prove that they are theatrical in origin ( made by others, without Shakespeare's approval), or that they are 'theatrical' in nature ( cheap, melodramatic, coarse, or inartistic). Cuts made in the theatre by Shakespeare's company could have been made by Shakespeare himself, resident dramatist and sharer; even if he did not initiate them, it can be assumed

(in the absence of evidence to the contrary) that he approved of them and lent them his authority by so doing.[24] There are five of these longer cuts: 1.3.129-133, 1.3.239-242, 1.3.268-293, 3.2.29-32 and 4.1.52-59 (following TLN 426, 531, 557, 1388 and 1974). The single-line cuts may have similar authority, but as they could easily arise through compositorial error, they are less reliable as evidence of manuscript consultation.[25]

Corrections of Q1 Error

The most crucial readings of all from the point of view of confirming the pattern established by the F correction of Q3 error are those which correct error in Q1-3. For the sake of objectivity, the readings considered at this juncture are only those accepted in one, two or all of three representative editions: the new Arden (A), New Penguin (P) and Riverside (R). There are 20 such readings:[26]

                                             

	TLN	Q1	F	F accepted by
1.1.118	123	by	by my	A P R
1.1.139	144	Ah but (Q(a), Q2-3); But (Q(b))	But	A P R
1.1.152	157	gentleman	Gentlemen	A P R
1.1.192	201	parlee	parle	A P
1.3.33	330	comes	com'st	A P R
1.3.172	465	sentence	sentence then	A P R
1.3.180	473	y'owe	you owe	A P
1.3.222	515	nightes	night	A P R
1.4.20	594	Coosens Coosin	Cosin (Cosin)	A P
1.4.52	627	Enter Bushie with newes.	Enter Bushy.
			Bushy, what newes?	A P R
2.1.48	687	as	as a	A P R
2.1.102	747	inraged	incaged	A P R
2.1.177	824	a	the	A P R
2.1.284	933	Coines	Quoint	A P R
2.2.16	968	eyes	eye	A P
2.3.99	1210	now	now the	A
3.3.13	1597	briefe	briefe with you	A P R
3.3.31	1615	Lords	Lord	A P R
3.3.206	1843	two	too	A P R
5.6.8	2802	Oxford, Salisbury, Blunt	Salsbury, Spencer,
		(Oxford, Salisbury Q2-3)	Blunt	A P R

Preliminary Summary

The collation of all the above data needs some care. Table 1 is itself by no means definitive, for several reasons. It gives no indication of the spatial distribution of variants, which are sometimes tightly clustered and elsewhere completely absent. It has wide indeterminate areas between some of the areas of relative correction and uncorrection. And it contains several random elements: fortuitous unauthoritative correction, isolated correction, light annotation, and oversight. A mechanism needs to be introduced to handle the elements of randomness. To achieve this, quite simply, a run of two or more indications of correction may be regarded as an area of correction. Such an area may be allowed to extend over single indications of non-annotation, but ends when two such opposing items occur consecutively. The same rules therefore apply for the uncorrected areas. (The terms 'corrected' and 'uncorrected' remain relative, for there may actually be any number of interim degrees.) This procedure may sound, if anything, conducive to chaos, for the number that defines a group could literally not be smaller. In fact, the exercise confirms and only minimally redefines the situation in Table 1. When the new data is included, only 17 areas of correction are established. These are as follows:

• 29-65
• 144-260
• 320-398
• 414-557
• 585-730
• 747-836
• 858-942
• 968-1004
• 1053-1055
• 1210-1229
• 1388-1415
• 1470-1615
• 1734-1798
• 1974-2017
• 2307-2368
• 2446-2517
• 2574-2716

The areas so defined give a working basis for describing the fluctuation within the process of annotation. But the approach remains explorative, and its limitations should be obvious enough. Most crucially, the guideline areas of correction fail to distinguish between the various degrees of consistency of annotation. Table 2, which gives the aggregated data, also gives a ready indication of the variation and of those areas of the text which remain indeterminate. For example, the passage from 1800 to 2100 might best be described as lightly annotated rather than corrected or uncorrected; this area alone accounts for over half the Q1-F agreements against Q3 that occur in passages defined as relatively uncorrected. The corollary is that elsewhere the distinction is sharper than the averaged-out statistics will suggest.

Corrected areas account for 51% of the available text.[27] This nicely medial figure may be compared with the proportions of the various categories of data to fall within these corrected areas.

                     

	{Major cuts	5/5	100%
	{A, P and R readings from F	18/20	90%
	{Manuscript misreadings	6/7	86%
Evidence of consultation	{F restorations of Q1	64/76	84%
	{Significant relineation (positive)	5/6	83%
	{Emended profanity	22/27	81%
	{Compounded error	2/9	22%
	{Shared error	2/10	20%
Evidence of failure to consult	{Q1 God(s) retained in F	6/32	19%
	{Uncorrected Q3 error	11/75	14%
	{Significant relineation (negative)	0/3	0%

Individually the above figures demonstrate little. The bridging mechanism designed to compensate for randomness can enhance small groups whilst detracting from large groups where the individual items involved are of less statistical significance. But this criticism cannot apply to the comparison of corrected and uncorrected areas for the accumulated evidence for and against annotation, whereby each item is of equal value:    

Positive data in corrected areas	120/141	85%
Negative data in corrected areas	21/129	16%

Nor is the initial information from Table 1 unduly distorting the picture. Without it, the results are:    

Positive data in corrected areas	55/65	85%
Negative data in corrected areas	10/54	19%

Nevertheless Table 1 provided a remarkably useful starting point. The rest of the evidence confirms and refines what it suggested. The only way the demonstrated correlations make any sense is to suppose that the vast majority of the listed changes in F were made during a single process of annotation carried out with reference to the promptbook. Most important of all for the editor of Richard II is the irresistible association between (a) the most plausible emendations of Q1 in F and (b) other indications of annotation.

Stage Directions

The stage directions are more enigmatic. Not only are added and changed stage directions no commoner in corrected areas than elsewhere, they actually occur in corrected areas noticeably less frequently than we should expect if they were randomly distributed. Only 34% of added and changed directions are found in corrected areas, suggesting the inverse of the expected relationship. The number of directions involved is too large for this proportion to be dismissed as insignificant. And the situation is not meaningfully alleviated by assuming that the more trivial changes were introduced by compositors. Even if we exclude another possibly untrustworthy group, the scene-break directions, we are left with a similar situation for the remaining minority of mid-scene directions. It is in their interaction with Table 2 that the added and changed directions can usefully be divided into two groups. In the first 1000 lines, 63% of them occur in corrected areas; this distribution is almost certainly random, as the area has a greater than average amount of correction. But only 20% of the altered directions in the rest of the text occur in corrected areas. Even though corrected areas account for only 36% of this second half of the play, this dissociation—only 1 altered direction in 5 occurring in an area which shows signs of corrections to dialogue—is unlikely to be random.

Speech Prefixes

No explanation of the anomalous distribution of stage-direction annotation can be offered without first considering the speech-prefix variation. The most striking instance of the latter is the reorganization of prefixes for Richard and Bolingbroke. For Richard, Q1 invariably has 'King' except for (a) the first prefix in 5.1, and (b) throughout 5.5, the prison scene. For the latter, he is unkinged 'Rich.'. Still in Q1, Bolingbroke's prefixes change from 'Bull.' or 'Bul.' (once 'Bulling.') to 'King H.' or 'King' for 5.3 and 5.6. F leaves 1.1 unchanged, but thereafter systematically (without exception) changes Richard's 'King' prefixes to 'Rich.' (occasionally 'Ric.' or 'Ri.').[28] Bolingbroke's 'King (H.)' prefixes all become 'Bul.' in F. These changes go far beyond anything that could be expected by way of regularization. They clearly must echo the pattern found in the promptbook; there is no other reason why the trouble should have been taken to effect such a detailed change when the prefixes

in Q1 were perfectly logical according to their own system.[29]

Most of the 'King' / 'Ri(c)(h).' variants are in corrected areas, and many of those that are not do occur in indeterminate or bordering areas. This correspondence is probably not between the categories of variant themselves, but between areas of annotated text and scenes in which Richard is on stage. (There is no comparable patterning of 'King (H.)' / 'Bul.' variants.) At first sight this correspondence might appear to be explained by the coincidence of heavy annotation in the first 900 lines and the exceptional number of scenes in which Richard is present in the same part of the play. After 870 (2.1.222) Richard does not appear for almost 700 lines, and after 994 (2.1.295) there are only very brief snatches of correction until Richard returns to the stage. Such a situation may in itself be no coincidence. But even within the first 900 lines, the parts of the text with Richard present are more consistently annotated than the others. The same also holds true after 1000, when both Richard's appearances and the signs of annotation are scantier. The dramatic falling-off of annotation around 1000 may be a (belated) response to Richard's long absence. The presence of the play's principal character seems to have been one factor which influenced the annotator's fluctuations.

Other significant F alterations of Q speech prefixes are limited and localised. Some, such as 'Herald' / 'I Har.' (401), 'Herald 2' / '2 Har.' (407) and 'H. Per.' / 'Percie' (1130, etc.) suggest but do not demonstrate annotation. More definite are two altered character assignations, 'King' / 'North.' (2346) and 'yorke' / 'Bul.' (2612); a moved prefix ('Bag.', 1100-1); and the distinctive changes 'Welch' / 'Capt.' (1285, 1291) and 'Man' / 'Ser.' (1850, 1865, 2657, 2660). Of the changes more definitely associated with promptbook annotation, only those at 2346, 2612, 2657 and 2660 fall within corrected areas. Stage directions whose rewording may be connected with the speech-prefix variants are 'a Welch captaine' / 'a Captaine' (1284), 'Gardeners' / 'a Gardiner, and two Seruants' (1833), and '&c.' / 'and Seruants' (2651). However, the reassignations at 2346 and 2612 are not particularly near changed stage directions. There are some signs that the annotated speech prefixes resulted from quite careful collation; there is no reason to associate it with annotation of the dialogue, and little evidence for an association with the annotated stage directions—though if (as appears to be the case) both speech prefixes and stage directions were collated thoroughly and systematically, such an association

could exist, without being detectable. If the collator caught all of the speech-prefix and stage-direction variants, then coincidence of annotation in the two categories would result only from coincidence of variation between Q3 and the manuscript. The lack of a discernible association between stage-direction and speech-prefix annotation therefore demonstrates nothing; but the lack of association between speech-prefix and dialogue annotation is significant.

Act and Scene Divisions

The Folio's introduction of act and scene divisions supplies the final piece in the jigsaw. They were almost certainly marked on to the quarto copy. The theatrical significance of act divisions remains in some doubt, but it is most probable that by 1610 the King's Men's promptbooks were regularly marked with act divisions.[30] There is no particular reason to doubt the validity of F's act division. (No one will doubt that the beginning of Act 4 at least is correctly positioned.) The scene divisions, on the other hand, include a generally-recognized error. Despite Q1's direction 'Exeunt. Manet sir Pierce Exton, &c.' (2651-52), almost all editors agree that the required staging is indicated in F: 'Exeunt. | Enter Exton and Seruants'. But F does not introduce a scene break, instead continuing its scene 5.3, and so misnumbering 5.5 and 5.6 as 5.4 and 5.5. This single anomaly can scarcely be unconnected with the form of the direction in Q.

The evidence of Richard II itself is here supported by an examination of other texts. The Folio contains eight plays set from lightly annotated quarto copy. One of these (Romeo and Juliet) has no act or scene divisions, and shows few if any signs of promptbook consultation;[31] five (Love's Labour's Lost, The Merchant of Venice, A Midsummer Night's Dream, Much Ado about Nothing, and Titus Andronicus), which were collated against theatrical manuscripts, provide act divisions only.[32] In Titus and Dream the text has been adjusted to accommodate an act division, which therefore seems likely to have come from the promptbook;

the authority of the division in Love's Labour's Lost has been doubted, but the act divisions in the two other comedies are at least acceptable and may derive from theatrical manuscripts. On the basis of these six other plays we might expect consultation of a promptbook to result in the provision of act divisions, but not scene-breaks. Only Richard II and I Henry IV provide act and scene divisions; as with the other texts, the act divisions are sensible, but in both plays an error in the scene divisions results from a misleading quarto direction. In I Henry IV, after TLN 2890 Q5 (following Q1) has the stage direction, Whether or not one accepts Rowe's interpolation of 'and exeunt' after 'embrace', the stage is evidently cleared at some point before the entrance of Douglas and Blunt; all editors mark a new scene (5.3). The Folio omits 'Here', substitutes 'entereth' for 'enters' and 'alarum vnto' for 'alarme to', but does not supply the missing scene division, and consequently mis-numbers the two following scenes. As with Richard II, it seems fairly evident that scene divisions were not marked in the annotator's manuscript, and that the annotator failed to mark a division here because he did not see the need for one. He could easily have been misled by the fact that Q5 does not contain the keyword 'Exeunt' or 'Exit'; either the manuscript also failed to supply the word, or he failed to transfer it from that source.

Hypothesis

The Folio's printed copy (Q3) for Richard II received at least three kinds of annotation: the systematic provision of act and scene divisions (the latter apparently editorial), the systematic collation of stage directions and speech prefixes, and the sporadic collation of dialogue variants. The fluctuations of accuracy in this third stage are documented in Table 2. Explanations for that pattern of fluctuation are necessarily conjectural. To some extent, no doubt, they reflect the universal rhythm of alternating alertness and inattentiveness.[33] We have already suggested that the annotator may have been more alert to dialogue variants in scenes where the protagonist is present. From the beginning the annotator left some short passages uncorrected, with no obvious motive for his alterations in procedure. But just as the annotator of Love's Labour's Lost stopped his thorough collation shortly after TLN 977, so the annotator of Richard II after 1003-4 must have decided that it was necessary

to lighten further the amount of attention he was giving the promptbook. The sudden slackening in his thoroughness is indeed quite dramatic (though not so total as in Love's Labour's Lost). The annotator apparently saved time partly by avoiding work on those parts of the play which had already received at least some annotation; that is, around the added and changed directions themselves. A few exceptions crept in between 2364 and 2704, but his basic approach remained the same until the end of the play. Such a procedure is not entirely illogical, and is psychologically plausible. It also fits neatly into the larger pattern of Folio annotation.

We have already pointed out that the layout of manuscript pages would have provided a physical encouragement and basis for separating the annotation of speech prefixes and stage directions from the annotation of dialogue. Such a separation also has an obvious basis in the nature of the two authorities the collator was comparing. His printed copy was a reprint of a good quarto, set from autograph or fair copy, presumably reliable in its verbal detail but deficient in the very theatrical features (stage directions and regularity of speech prefixes) supplied by a promptbook. Hemmings and Condell would have known very well the differences between these two kinds of documents, and even if they were not themselves responsible for collating quarto and manuscript, one would expect them to have discussed with the publishers, at least in general terms, the differences between the material in their possession and the printed texts already available. Finally, such a method of working would have been immediately attractive and intelligible to a printer or publisher: Moxon explains the normal two-step compositorial procedure for setting pages with 'Marginal Notes . . . down the Side or Sides of the Pages'.[34]

Before Richard II, four Folio texts had been set from annotated quarto copy: Much Ado about Nothing, Love's Labour's Lost, A Midsummer Night's Dream, and The Merchant of Venice. In each case, the annotation was sparse, chiefly affecting act divisions, stage directions and speech prefixes. But in the case of Richard II it seems to have been

decided, for whatever reason, at whatever point in time, to attempt a more extensive collation of Q3 against the manuscript. This collation of dialogue variants was, by modern standards, unevenly, even 'superficially' carried out, but it nevertheless represents a clear departure from previous Folio practice. And I Henry IV, the next Folio play to be set from quarto copy, shows similar signs of two different kinds of annotation: again, directions and prefixes have been altered throughout the text (often in trivial ways), and act and scene divisions provided, but alteration of the dialogue is sporadic.[35] I Henry IV and Richard II both, in different ways, offer evidence of a transition in methods of Folio annotation, from the consistent practice in the Comedies to the more extensive annotation first undertaken in earnest in Richard III (up to the end of quire s).[36] After this experiment in extensive annotation, laziness returned: the end of Richard III (t1-2v), Romeo, and Titus are given the absolute minimum of annotation (again concentrating on stage directions, prefixes, and act divisions), and the same was clearly intended at first for Troilus. This reversion to the earlier procedure may have been connected with the introduction of Compositor E, who by all accounts was an inexperienced compositor who may not have been able to cope with heavily-annotated copy; or Compositor E may have been called in because a decision had already been taken to make do with sparsely-annotated copy. In any case, after this lapse the Folio finally reverts to extensively prepared quarto copy for both King Lear and Troilus, the last two plays to have been set from printed copy.[37] In short, the Folio itself, like the individual plays within it, shows a pattern of varying thoroughness in the annotation of quarto copy; within this pattern, Richard II and I Henry IV are pivotal, and both reveal a discrepancy between their treatment of, on the one hand, directions, prefixes, and act and scene divisions, and, on the other, their annotation of dialogue.

This apparent pattern in the nature of the copy from which Jaggard's compositors set inevitably raises a question about who prepared that copy. In the case of the 1647 Beaumont and Fletcher Folio, a single publisher (Moseley) provided copy for six different printers; all of the texts seem to have been set from manuscript copy, and the collection and preparation of the copy was clearly the publisher's rather than the printer's responsibility.[38] The 1616 Jonson Folio was both published and printed by William Stansby; Jonson himself was clearly instrumental in supervising the preparation of this volume, and hence was probably personally responsible for the annotation of those plays set from annotated quartos.[39] The Shakespeare First Folio falls somewhere between these two eminent stools. Rather like Stansby, Jaggard—the sole printer—was also co-publisher; but the 1623 Folio, like that of 1647, was a posthumous publication, and the collection of texts to print was at least partly and perhaps wholly the work of Heminges and Condell. Unlike Jonson, Shakespeare was not available to annotate quartos himself; yet like the Jonson Folio, Shakespeare's contains plays undoubtedly set from annotated quarto copy. Who annotated those quartos? Heminges and Condell may have done so; or the publishers may have. In this case, however, the printer was himself one of those publishers—indeed the senior partner in the syndicate: as Greg concludes, 'There can then be little doubt that . . . the moving spirit of the scheme on the publishing side was Isaac Jaggard' (First Folio, p. 6).

Jaggard's shop was clearly in a position to provide the skills of collating and proofreading, and to ensure that the quartos were annotated in the manner most convenient for the compositors. Three more specific arguments suggest that even if the annotator was not 'Jaggard's man', he nevertheless worked in association with the printers. First, it does not seem likely that theatrical professionals would have introduced act divisions as inept as those in The Taming of the Shrew, Henry V, and I Henry VI; the divisions in those plays, at least, seem to originate outside the theatre, and if so a similar origin for other editorial annotations of copy must be suspected. Secondly, it is certainly suspicious that the evidence for consultation of a manuscript in Folio Love's Labour's Lost ceases so abruptly only nine lines from the bottom of a Folio page—nine lines which would have required no other annotation. This might

be a coincidence, but it suggests that annotation and casting-off of copy sometimes went hand in hand. Finally, a dramatic change in the character of the annotation at the end of Richard III coincides with an irregularity in the sequence of setting. After most of Richard III had been set up from heavily-annotated quarto copy, at the end of quire s work on the play was broken off, and setting of the Tragedies commenced. When eventually the compositors returned to Richard III, extreme variation from quarto copy continued for only half of the first column of t1; the remainder of the play was set from copy that had been only nominally edited, almost wholly in respect to speech prefixes and stage directions. This coincidence seems most plausibly explained in one of two ways, either of which presumes a close connection between the annotator of the quarto and Jaggard's compositors. The temporary abandonment of work on Richard III may have been due, as Hinman conjectured (II, 127), to difficulties in securing the copy for Henry VIII (which completes the Histories section of the volume). If this explanation is correct, then the drastic change in the level of annotation on column a of t1 may have occurred because the annotator had, in the interim, prepared copy for Titus and Romeo (already set or being set, when work on Richard III resumed), and as a result quickly decided to revert to the more economical degree of annotation employed for those two texts. This seems to us unlikely: why should the annotator stop work on Richard III just because copy for Henry VIII was not ready, why should there have been any problem about the copy for Henry VIII, and why when resuming Richard III should he extensively annotate half a column of Folio text before abruptly changing his methods? It seems likelier that the abrupt switch to the Tragedies occurred because the annotator had encountered some difficulty or deficiency in the manuscript he was consulting, halfway down the first page of a new quire, and that completion of Richard III was delayed while an attempt was made to rectify this difficulty. A full justification of this hypothesis would require an extensive investigation of the vexing problem of printer's copy for Folio Richard III; here we must limit ourselves to the statement that such an investigation has convinced us of the validity of this hypothesis.[40] Certainly, it would be most economical to assume that the change in the nature of Jaggard's copy at the end of Richard III bears some relation to the change in the order of Jaggard's setting. This does not prove that the annotator was employed by Jaggard, or worked in his shop, but it does suggest—like the other evidence—that he worked in close association with Jaggard's compositors, and that he may sometimes have cast off pages as he annotated them.

Two further facts, specific to Richard II, support the same conclusion. First, it seems most likely (as we have suggested) that the apparent influence of Q5 in a 53-line stretch near the end of the play results from the fact that someone in the theatre transcribed that passage from Q5, in order to rectify a gap in the promptbook. If someone in the theatre also presented Jaggard with annotated printed copy, why was Q3 (rather than Q5) used? The most obvious agent to have performed both functions would have been the book-keeper, who would have done both jobs between 1615 and 1622; we would expect the same quarto to be used for both, even if (as is possible) there had been a change of book-keeper.

Secondly, we have already remarked on the fact that Folio Richard II betrays evidence of annotation more extensive than that in the Comedies; I Henry IV has been similarly annotated; Richard III, the next play set from quarto copy, was even more extensively annotated. Matthew Law, who was not a part of the syndicate which published the Folio, held the copyright to all three plays (and to no others printed in the Folio). One does not like to attribute such coincidences to coincidence. In the case of Troilus and Cressida, it has been widely accepted that the difference between Jaggard's abandoned first setting of the play (after Romeo, from quarto copy as minimally annotated as that for Titus and Romeo) and the second (between the Histories and Tragedies, from heavily annotated quarto copy) probably resulted from difficulties of copyright: the heavy annotation was, in other words, a ruse to evade someone else's copyright. (See Greg, First Folio, pp. 447-449.) We find it difficult not to see the same motive at work in the change in the level of annotation which takes place between the Comedies and Richard II.[41] And if the annotators were working closely in conjunction with Jaggard's compositors, then the irregularity in setting which interrupted work on Richard II may be related to that which interrupted work on Richard III. Henry V, 1 and 2 Henry VI, and part of 3 Henry VI (all from manuscript copy) were set before Richard II was completed and 1 Henry IV begun. Could this interruption be due to the fact that the annotator was not far ahead of the compositors whose copy he prepared, and that in order for him to have time to finish the (more extensive than usual) annotation of Richard II and 1 Henry IV, the compositors had to leapfrog him and begin work on Henry V? We cannot confidently answer this question, but it is surely legitimate to ask it.

All of this evidence suggests that the man (or men) who annotated quarto copy for use by Jaggard's compositors worked closely with Jaggard's compositors. Jaggard's dual position as sole printer and co-publisher would have made this possible. But would the King's Men have turned any of their promptbooks over to Jaggard? Perhaps not. If Heminges and Condell would not let their promptbooks out of the theatre, then (obviously) the annotator must have performed the actual task of marking up a quarto in the theatre itself, rather than the printing house.[42] If the King's Men had no other manuscript of the play on hand, and if Jaggard for copyright reasons needed a source of variant readings, and if the King's Men wouldn't let go of their promptbook, then someone would have had to collate Jaggard's quarto against their promptbook.[43] Whoever that someone was, he must have worked, for a while, in close association with both the King's Men and Jaggard.