University of Virginia Library

New Approaches to Textual Problems in Shakespeare
by
Philip Williams, Jr. [*]

One of the most promising approaches to the textual problems in Shakespeare is the study of those compositors who from manuscript copy first set the plays in type. In the transmission of the texts, these type-setters played a role the importance of which can hardly be overemphasized: if they were literal and painstaking, they reproduced relatively accurately the manuscript from which they worked; if they were inept, careless, or intent on improving their copy, the printed texts differed considerably from their manuscript copy. Thus in a given text we may find what a printing house employee thought Shakespeare wrote, or what he thought Shakespeare might have written, or—more disastrous still—what a printing house employee thought Shakespeare should have written, rather than what he actually wrote. In what follows I shall not discuss the methods by which compositors can be identified, but assuming that identification is possible, I should like to examine some of the implications of compositor analysis and related bibliographical techniques for the texts of Shakespeare. I shall here be concerned with those plays in the 1623 folio that were set from manuscript rather than printed copy. For purposes of illustration, I shall rely mainly on the histories, although the comedies or tragedies would serve equally well: the same or related textual problems are encountered in all three sections of the first folio.

One of the weaknesses in editing Shakespeare has been a tendency to specialize: a single editor does one play. While such a procedure has much in its favor, it sometimes seems to prevent the editor's bringing to bear on the problems of his text a detailed knowledge of other related textual problems. Those editors who, like Dover Wilson and Peter Alexander,

have worked their way through the texts of many plays have each been able to make important contributions that might have been beyond the scope of the editor of a single play. The texts of the twenty-four plays that, according to Sir Walter Greg, were set from manuscript copy in the First Folio should be studied collectively as well as singly, just as those plays set from annotated quartos must be studied as a group as well as individually. Only when studied in this way can we obtain reliable information about the hypothetical editor of the folio, for example, and the character of his work. To make deductions about his editorial practices on the basis of the study of a single play is almost useless. The value of this broader approach is amply demonstrated by Dr. Alice Walker's Textual Problems of the First Folio.

For our purposes, it is unfortunate that William Jaggard printed so few plays. Unlike Okes, Simmes, and Eld, for example, the Jaggards did not specialize in drama. In the ten years from 1617 to 1627, no plays (except those in the First Folio) were printed from manuscript copy in their shop. In 1619, however, William Jaggard had printed ten plays for Thomas Pavier, but the so-called Pavier collection consists entirely of reprints of quartos that had appeared earlier. If instead of turning out rather pompous heraldries, histories, and books on theology, Jaggard had printed first editions of plays, these texts would serve as useful checks and controls for conclusions about how his compositors handled manuscript dramatic copy.

The study of these compositors in Jaggard's shop has had a brief yet significant history. In 1920, Mr. Thomas Satchell published in The Times Literary Supplement some observations on the spellings in the folio text of Macbeth, noting that some words consistently spelled in one way in the first part of the play were consistently spelled differently in the latter part. Satchell's deductions from these observations were modest: he simply raised the question whether the variant spellings were to be explained by a change in the manuscript copy or by a change in the compositors who set the type. Dr. Willoughby, in 1928, supplied the answer: the different spellings in Macbeth were to be explained by the habits of two compositors whom he distinguished as A and B, and some (but not all) of the spelling differences in Macbeth could be used to identify their work in other parts of the folio. Dr. Willoughby found these two compositors at work in The Tempest, with which the folio opens, and in Troilus and Cressida, the last of the plays to be printed. Although the importance of Willoughby's discovery should have been immediately recognized, compositor determination and analysis was almost completely neglected until an important article by Professor Hinman in 1940 stimulated interest in the subject. Since then, the study of the folio compositors has received considerable attention both here and in England.

The unfortunate neglect of the subject following the publication of Dr. Willoughby's monograph resulted in serious defects in several important editions of Shakespeare. Professor Shaaber, for example, in his almost model Variorum edition of 2 Henry IV, did not make use of compositor determination and analysis. Had he done so, the vexing problem of the exact relationship of the quarto and folio texts might have been solved then and there. Since the publication of the Variorum edition in 1940, Dr. Walker has argued that the folio text was set from an annotated quarto; Professor Shaaber has re-affirmed his earlier position that the folio text was set from a manuscript having independent authority. The implications of this controversy for the editor are of course considerable. In view of the basic disagreement between two outstanding authorities, the relationship of the quarto and folio texts will have to be re-examined. There are indications that neither Dr. Walker nor Professor Shaaber is wholly correct. As Professor Bowers has observed, the quarto and folio texts are clearly not independent as Professor Shaaber believed, and yet an annotated quarto can hardly have served directly as copy for the folio text. The abnormal spellings of both compositors A and B are not consistent with such a theory. Printed copy exerted a greater influence on the compositors than did manuscript and caused them to deviate from their normal practices, but the abnormal spellings in 2 Henry IV cannot be accounted for by the quarto text. The most likely explanation now seems to be that the copy for the folio text of 2 Henry IV was neither an independent manuscript nor an annotated quarto but rather a manuscript transcript of a quarto annotated for use in the theater. But until the work of the two compositors has been competently analyzed, the answer to this puzzling question must remain in doubt. Whatever the answer may be, I am confident that it will not be achieved without the use of compositor analysis.

But deductions based on even the most detailed study of a single text are at best hazardous and at worst dangerously misleading and wrong. After his remarkably thorough study of the Hamlet texts, Professor Dover Wilson concluded that the good second quarto had been set from Shakespeare's foul papers by a hurried compositor—perhaps a dishonest Welshman—who could not work quickly because he had not mastered his craft. On this theory, Wilson constructed his complex textual and critical structure. Professor Wilson's single harried compositor now turns out to have been two men, both regular compositors in Roberts's shop, and both competent workmen at that. I once made an outrageous generalization about a minor point in the folio text of Troilus and Cressida. In the Troilus quarto the spelling of the exclamation oh varies, apparently indiscriminately, between o and oh. The folio text reproduces this variation with remarkable fidelity. I used this evidence, validly I think, to support my

theory that the folio text had been set directly from a corrected copy of the quarto, but I added that the usual spelling of both folio compositors seemed to be o on the basis of my analysis of their work in Julius Caesar. It is true that the spelling in Julius Caesar is o, but it is likewise true that in The Comedy of Errors, another play set by the same two compositors, the spelling is oh. For reasons that I now do not pretend to understand, the folio compositors generally reproduced the copy spellings of this form. I can not seriously believe that two compositors mutually decided to use one spelling for one play, another for a second, and to mix the two spellings in a third. These o and oh spellings must then represent the spellings found in the copy, and because they do, they shed some light on the underlying manuscripts. It would be useful to know what Shakespeare's own usage was. Unfortunately the Hand D of Sir Thomas More (which I accept as Shakespeare's) is not much help, for the form appears only once, spelled o. But I would suggest that the answer is not beyond the reach of the analytical bibliographer. The good second quarto of Romeo and Juliet was almost certainly set from Shakespeare's foul papers. A study of Thomas Creede's compositors (not in this text alone but of their total output) might reveal that they generally reproduced the copy spelling without imposing personal idiosyncracies. If so, we might establish that Shakespeare, in his early career at least, used one spelling or the other, or both indiscriminately. We could then apply this evidence to the folio text of King John, which Sir Walter Greg believes was printed partly from Shakespeare's foul papers and partly from prompt copy. In the first three acts (which may have been set from foul papers) the spelling is o; in the last two acts, the spelling is predominantly oh. Knowing how conservatively Compositors A and B treated this form, we should have a neat, and I think rather convincing, bit of evidence to support Greg's theory about the copy for King John. For the later period, a like study of the second quarto of Hamlet might be applied to Timon of Athens to support Greg's belief that Timon was set from unfinished foul papers, or my belief that it was set from a fair transcript made by the same scribe who prepared the manuscript from which the folio text of Coriolanus was set. The trivia of analytical bibliography can sometimes be put to good use.

The text of King John will serve to illustrate another use which compositor analysis can serve. Although he found it "difficult to envisage a composite manuscript actually made up partly of foul papers and partly of a section of the prompt-book," Sir Walter discovered what seemed to be confirmation of sorts when he noted that "in the first three acts the directions tend to appear a line or two late, in the others a line or two early." The fact that some pages of King John were set by Compositor A and others by Compositor B raises some question as to the significance of

Greg's evidence. There are, at the very most, twelve stage directions (of which two are very doubtful) that can be described as being either early or late. The ten certain ones occur on pages set by Compositor A; the two doubtful ones on pages set by Compositor B. Judging from his work throughout the folio, Compositor A tended to reproduce the copy position of directions, whereas B tended to normalize and was reluctant to interrupt a speech to insert even a needed stage-direction. Greg's theory therefore receives very little support from the positions of these stage directions, for the variation is better explained by the typographical practices of the two compositors who set the folio text.

If indeed there was a change in the character of the manuscript copy from which the folio text of King John was set, the evidence to establish the fact will be found by an analysis of the eight pages in Acts I, II, and III set by Compositor A as compared with the four pages in Acts IV and V set by the same compositor. Because Compositor B reproduced the accidentals of his copy less faithfully, the pages set by him are less likely to produce helpful evidence, but compositor analysis can certainly help to elucidate the textual problems of this play.

My discussion so far has been based on the assumption that in the transmission of the text from manuscript copy to type, some characteristics of the manuscript were preserved, perhaps quite inadvertently. This palimpsest of the underlying manuscript is fainter in the work of Compositor B, stronger in the work of Compositor A. The folio text of Titus Andronicus furnishes a useful illustration to support the validity of the basic assumption. Titus Andronicus was not, of course, set from manuscript in the First Folio: the folio text was set from a copy of the third quarto of 1611. But one scene in the folio text (III.2) does not appear in any of the preceding quartos and must therefore have been set from manuscript copy. Had no copy of the three earlier quartos survived, it would nevertheless have been possible to show that the copy for Act III, scene 2, differed from the copy for the rest of the play. In this scene the name Tamora is spelled with an i; in a speech-heading Andronicus is used rather than Titus, and—more interesting—the word upon which appears four times in this short scene is each time spelled with a double p—a spelling that is extremely rare elsewhere in the First Folio. Because of the weight that I would have the textual scholar attach to such apparently trivial evidence, it is reassuring to have this one example that has the unquestioned check of external evidence. One would like to see what compositor analysis can reveal about the troublesome Hecate scenes in Macbeth or the Fool's distasteful prophecy in Lear. Unfortunately, both were set by Compositor B who may well have obscured any traces of a change in copy.

The three parts of Henry VI present problems that may be insoluble;

certainly at present they are far from solution. All three parts were set from manuscript in the First Folio, but for parts 2 and 3, some use was made of the earlier corrupt quartos which the folio texts supersede. No version of Part 1—whether memorial reconstruction or early draft—had previously appeared in print. It is Part 1 that I wish briefly to consider for what compositor analysis can and what it can not contribute to the study of this text.

Of the folio text, Sir Walter Greg writes: "Many directions are elaborate and descriptive; most noticeably so in the first two acts, but 'Enter in skirmish with bloody pates' is a nice example from the third, and the closely following 'Begin again' also suggests the author. Several . . . [directions] begin with 'Here': this has an archaic appearance and may be a personal peculiarity. . . . Instances are confined to the first act, except for one in the third . . . . As in other early plays we seem to have an authorial manuscript used with little modification as prompt copy. Either composite authorship or revision would explain certain contradictions and inconsistencies in the text. Whether the copy itself was 'heterogeneous' is uncertain. . . ."

We now know that the text of 1 Henry VI was set by Compositors A and B. The former set all of Acts I, II, and III save for the first 46 lines of Act II; Compositor B set all of Acts IV and V save for page 113 which is the work of A. The unusual features of the text (including the 'Here' directions) that Greg comments on are all found on pages set by Compositor A. Such directions may, as Greg suggests, be a personal peculiarity of one of the disputed hands in the play, and one wonders if the absence of the unusual directions in Acts IV and V is not to be explained by Compositor B's revamping the directions of his copy. He was, as we know from his work elsewhere, quite capable of doing so.

But the most striking textual anomaly of 1 Henry VI can not be attributed to the two compositors and their differing practices. The act division is very imperfect and irregular: only Acts III and IV are divided into scenes, and Act V consists of a single 108-line scene. Credit for the introduction of scene division can not be given Compositor B, nor can the absence of division be blamed on Compositor A, for having set Acts I and II without division, Compositor A set all of Act III, which is divided into four scenes. It is incredible that having set two acts without scene division, Compositor A should have introduced the convention with the third act. We are therefore left with the puzzle of manuscript copy with scene division in Acts III and IV only.

If this anomaly does indicate a change in the manuscript copy for 1 Henry VI, analysis of Compositor A's work in Acts I and II compared with his work in Act III should disclose it. The evidence may be slight, but because of the literalness with which he followed copy, a change, if

there was one, should be demonstrable. I cite the following as examples of the kind of evidence that should be brought forward to support such a theory. In Acts I and II, Burgundy is invariably spelled with a d, but when the name appears in Act III (coinciding with the introduction of scene division) Compositor A consistently uses the spelling Burgonie. Even more striking is the treatment of the name of Joan of Arc. In undivided Act I, Compositor A consistently spells the name Puzel for the eighteen times that it occurs in directions, speech-headings, and text. But in divided Act III, in which the name appears twenty-six times, he invariably spells it Pucell. In short, the name appears spelled with a z in the pages before scene division, and spelled with a c (by both compositors) in the pages after scene division is introduced. Because the change is not due to a shift in compositors, and because Compositor A was literal in such matters, it is safe to conclude that the change represents a spelling change in the manuscript. Coinciding as it does with the introduction of scene division, the obvious conclusion is that there was some change in the manuscript at the beginning of Act III. When corroborated by additional evidence of this kind, it will be possible to refine Greg's statement "Whether the copy itself was 'heterogeneous' is uncertain." The copy for 1 Henry VI was almost certainly heterogeneous.

But evidence of this kind can be very tricky and must be handled judiciously. In this same play, for example, two spellings for the name Joan are used: Joane and Jone. At first sight it might appear that this variation is as significant as the Puzel—Pucell variation, but such is not the case. The o spellings all appear on pages set by Compositor B; the oa spellings on pages set by A. Without the check of compositor analysis, a naive scholar might well fall into the error of attempting further to disintegrate the text on the basis of these variant spellings.

Henry VIII, the last of the histories, offers an even better illustration of the curb that compositor analysis can put on unbridled and injudicious speculation. If the bibliographer cannot solve the principal problem of this play, he can at least lay solid foundations upon which other disciplines can build. The problem here, of course, is divided authorship.

When Alfred Lord Tennyson remarked to his friend James Spedding that "many passages in Henry VIII were very much in the manner of Fletcher," he set up a puzzle that has fascinated Shakespearians ever since. Spedding—who was a scholar—took the poet's hint and made a 'stylistic' analysis, publishing his findings in The Gentleman's Magazine for August, 1850. To Shakespeare he assigned Act I, scenes 1 and 2; Act II, scenes 3 and 4; Act III, scene 2, lines 1-203; and Act V, scene 1. This division, corroborated by many subsequent studies, has been generally accepted. To distinguish the hands of Shakespeare and Fletcher, Professor Thorndike

later developed and applied the 'em vs. them test; Dr. Partridge analyzed other contracted forms; and Professor Oras, in a study published only last year, found confirmation for the traditional assignments to Shakespeare and Fletcher. Among recognized authorities, only Professor Alexander has held out for single and Shakespearian authorship. His argument, however, is far from convincing: the 'stylistic' differences that have so often been reduced to statistics do exist; they are in the printed text to be counted, analyzed, and interpreted as one sees fit.

In all the counting and tabulating that has been done, not one of the investigators has been aware that the folio text of Henry VIII was set in type by two compositors with very different working habits and very different ways of treating some of the data that have been subjected to statistical study.

Henry VIII occupies about twenty-seven pages in the folio. Of the twelve pages attributed to Shakespeare, Compositor B set seven and Compositor A set five; of the fifteen non-Shakespearian pages, B set six and Compositor A set nine. It is therefore significant, in the vocabulary of statistics, that of the sixty-six 'em's found in the text, fifty-five appear on pages set by Compositor A, and only eleven on pages set by Compositor B. The conclusion is, I think, inescapable: Compositor A reproduced his copy more literally than did B who, taking expected liberties, normalized many of his copy spellings to them. If the paucity of 'em's in the pages set by Compositor B represents the compositor's sophistication of copy rather than a Shakespearian characteristic in the copy from which he worked, I am afraid that in a text worked on by Compositor B any division between Shakespeare and Fletcher on the basis of this test is of questionable value and may be dangerously misleading.

Another characteristic that has been used to distinguish the hands of Shakespeare and Fletcher is Fletcher's well established fondness for the pronoun ye. On the pages set by Compositor B, we find that the ratio is 208 you's to 25 ye's, or eight to one. On the pages set by Compositor A, the ratio is 191 to 48, or four to one. In other words, the ratio between you and ye is almost exactly twice as great on A pages as it is on B pages. John Fletcher may have had a preference for 'em and ye, but we can be sure that the anonymous Compositor B of Jaggard's shop was not William Shakespeare.

I do not mean to suggest that the differences between what are generally called the Shakespearian parts of Henry VIII and the non-Shake-spearian parts are all to be explained away by the practices of the two folio compositors. The differences in the play are most striking, and Tennyson's ear was a good one. But I would insist that no stylistic analysis of Henry VIII can be satisfactory that is not based on a knowledge of the two

compositors, the pages they set, and their practices, not only in this play but throughout the folio.

I have illustrated some of the many uses that compositor analysis can serve in solving problems in the text of Shakespeare. The most important use (from the editor's point of view) has not been mentioned: the number and kinds of errors that the different compositors were likely to make. The uses to which this relatively new approach can be put are so varied that there has been no need for me to poach on what has become Dr. Walker's particular territory. But almost as important for the editor is the information that compositor analysis can supply about the character of the manuscripts from which the folio texts were set.

When we find a compositor consistently tending to mix his spellings of words for which he has an established preferential spelling, it is safe to conclude that the aberrant forms represent copy spellings. If, in a single play, Compositor A consistently gives up his characteristic deare-deuel-young spellings while retaining other characteristic spellings of his own, he presumably did so because of the influence exerted by his copy. When Compositor B intersperses frequent deuell-young-Weele spellings, the same conclusion seems inescapable. After compositor spellings have been isolated and the copy spellings established, the manuscripts from which the folio texts were set can be grouped according to common spelling characteristics. We may find one group of manuscripts characterized by the chain of doe-heere-honour-Wee'll (and so on) spellings; another group by do-here-honor-Wee'l. A third group will, I suspect, be related by short spellings of wil, shal, tel and the like, a marked preference for final ie instead of y, and the long spelling of pronouns ending in e. When we find Compositor A spelling to the as toth' (one word) in Macbeth and Coriolanus, and nowhere else in the folio, we have strong evidence linking the manuscripts from which these two plays were set. And when we find the pages set by Compositor A in Henry VIII and Hamlet characterized by a phenomenally great number of semi-colons, quite in excess of his usual usage, we have a link between the manuscripts which is supported by his aberrant spelling Wee'l etc. in both texts.

When the varnish of the folio compositors has been removed, the grain of the underlying manuscripts will be revealed in its true color. Then, and not until then, shall we have exhausted the evidence possible about the kinds of copy from which these plays were set. I may be too optimistic, but I venture to hope that when the job is done, we shall be able to distinguish those plays which were set from Shakespeare's holograph from those set from scribal transcripts, and to assign with some certainty the transcripts to the scribes who made them. A good place to start will be with the five plays which are now thought to have been set from transcripts

made by Ralph Crane. The extant manuscripts in his hand will first have to be studied not only to isolate his personal idiosyncracies that may show through in the folio texts, but also to establish the latitude of variation that can be expected in his work. In spite of similarities, there are marked differences in his work as exemplified, for example, by the manuscript of Barnavelt and the Bodleian manuscript of A Game at Chess. When the spade work has been done, we may find absolutely convincing evidence to support Greg's hypothesis that The Tempest, The Two Gentlemen of Verona, The Merry Wives of Windsor, Measure for Measure, and The Winter's Tale were set from Crane manuscripts—or we may find something else. Until the possibilities of compositor analysis are fully exploited, we can not be sure.

Having spoken at such length about the importance of compositor analysis, it is embarrassing to own up to some of its deficiencies and to point out difficulties that, in the case of the Jaggard compositors, have yet to be overcome. Neither Dr. Walker nor I have yet been able satisfactorily to identify the compositor who set the folio texts of Titus Andronicus and Romeo and Juliet and whose work we suspect in other plays.[*] While Compositor B remained remarkably stable and changed very little from 1619, when he worked on the Pavier collection, through 1623 and the last play printed in the folio, Compositor A, who was more sensitive to copy, is more difficult to isolate. There are long stretches in the folio that may have been set by Compositor A or that may have been set by a third compositor whose work, either because of copy influence or because of his own A-type characteristics, closely resembles Compositor A's work. The deficiencies in spelling tests based on a few words only have been well demonstrated. On some pages the key words simply do not appear, or appear so infrequently as to be of little use. What is wanted is a fuller list of preferential spellings, supplemented for Compositor A by a list of copy spellings that he would tolerate and that B would not. Neither compositor was willing to reproduce Crane's habitual spelling theis for these, but there are many spellings that A might reproduce but which B would balk at. Although a great deal of work remains to be done before we shall have perfected the techniques by which Jaggard's compositors can be pin-pointed, there is every reason to believe that the effort will be successful, especially so if supported by other investigations that the analytical bibliographer can make.[**]

An analysis of the press work of the First Folio is urgently needed. If Dr. Walker is correct in her assumption that Compositors A and B followed

their copy more closely when working rapidly, it follows that we should like to know when the compositors were hurried by the press or presses. I should like to see corroborative mechanical evidence to explain the almost literatim faithfulness with which the folio reproduces the quarto texts of Much Ado About Nothing and Titus Andronicus, if hurried composition is indeed the answer. Or, from many important examples that come to mind, to cite a minor one: the use of white-space in the folio should be investigated. It is surprising how often act or scene division coincides with the top of a column. In King John, six of the sixteen divisions are found in this position. In addition to other ways of conserving or wasting space, the compositor could leave white-space above and below stage directions. Sometimes he left space above and below, sometimes above only, and sometimes none at all. This practice explains the loose appearance of Compositor B's page 305 in Lear and the tighter appearance of the same compositor's page 135 in Macbeth. The subject has never been investigated, but a study might reveal that some of the copy for the folio was looked over and roughly counted off before it reached the compositors. Unless it was known by pre-arrangement that Compositor A was to take over in Hamlet at II.2.219, it is difficult to explain why Compositor B stretched out the prose speeches of Polonius and Hamlet into irregular lines of verse at the bottom of 003. If copy for the folio was first read over by the printer for this purpose, perhaps other indications to guide the compositors were made at the same time. Sir Walter Greg suggested long ago that it may have been the printer rather than the illusive folio editor who removed some of the profanity from certain of the 1623 texts.

The more closely compositor analysis is related to other approaches, the more productive it will be. It can support the paleographer who attempts to recover manuscript spellings by analysis of errors caused by misreading Elizabethan handwriting. The linguist who brings order out of the seeming chaos of Elizabethan spelling can be of great help. Is the quarto spelling arre (Troilus and Cressida, I.3.392) a compositor's error and therefore to be emended, or is it an acceptable spelling of a 17th-century variant form of the folio's tarre? Is there any real significance in the folio's alternation between does and do's, or is it just chance that in Hamlet the contracted form is used in unstressed syllables and the full form in stressed? Anyone interested in Shakespeare's prosody would like to know.

Much of the bibliographical work that remains to be done for the text of Shakespeare may seem terribly time-consuming and unrewarding. Much of it certainly is. The collation of the extant copies of the 1609 Troilus and Cressida quarto revealed only the most trivial and uninteresting variants. But until the job was done, the editor of this play—if he had a conscience—had qualms when he spoke of "the quarto reading," for he could

not be sure. All of this tedious collation and analysis must be done. The task is not more overwhelming (although it may be more complex) than that presented by the text of Chaucer. The Chicago Chaucer was undertaken, and the Chaucer scholar has a text for which he need not apologize—as we must today for the antiquated Old Cambridge edition of Shakespeare, reproduced substantially in recent years by Kittredge, Neilson, Harrison, Craig, and others. Inexpensive old-spelling editions of Milton are readily available, and few of us would care to read Paradise Lost in a modernized text. When an old-spelling edition of Shakespeare finally does appear, the study of the compositors in the dozen shops where substantive editions of Shakespeare were printed, will have contributed to the excellence of this badly needed edition.

Notes