University of Virginia Library

Shakespeare's Text and the Bibliographical Method
by
Fredson Bowers [*]

Though definition of terms is dull work, any paper which attempts to discuss bibliographical method had better state what it means by that phrase, for of all modern terms for scholarly disciplines, 'bibliography' is perhaps the loosest used and the least understood. From among the various subdivisions of bibliography, two have the greatest bearing on the problems of Shakespeare's text. The first is analytical, or, as it is often called, critical bibliography. This I take to be the technical investigation of the printing of specific books, or of general printing practise, based exclusively on the physical evidence of the books themselves. Such a line of investigation may detect and explain cancels and reprinted substituted sheets of all kinds. It is especially concerned with the presswork of a book, and it may explain by the use of mechanical evidence the number of presses engaged on a book, the order of the formes through the press, the printing together as one sheet of widely separated leaves, or the assignment of parts in a book broken up and printed simultaneously in sections. In Sir Walter Greg's phrase, it treats a book not as a literary document but simply as a tangible object produced by a physical process. The evidence the method employs is correspondingly physical, and the ideal is to build up a case which could be taken to a court of law for a favorable verdict. This is analytical bibliography in its purest sense.

Two illustrations may be given of its operations on Shakespeare problems. The first is the famous case of the 1619 Shakespeare editions which were falsely dated reprints, the so-called Pavier quartos. By a magnificent piece of bibliographical detective work in analysis, Greg and Neidig first showed, on the mechanical evidence of the similarity in watermarked

paper and the fact that there was standing type in the title-pages, that these editions—which purported to have been printed at widely varying dates—were in truth manufactured continuously at one time, some years after the dates alleged on various titles. The importance of this discovery for Shakespearian textual criticism cannot be over-estimated. Since the dates on some of these quartos were the same as those for authentic first editions, scholars, who had previously been able to operate only by what critical acumen they possessed, had in some cases confused the correct order and had set up one or more of these late and unauthoritative reprints as the true first edition of the very highest textual authority. Although there had not been complete unanimity of opinion in this matter, there was no method by which critical judgment could positively demonstrate the correct order. Yet Greg's mechanical evidence could have been taken into a court of law. His case—being bibliographical in an analytical sense —could and did positively demonstrate the correct order by physical evidence admitting no further controversy. At one stroke questions of the relative order and authority of a number of texts were settled once and for all.

My second illustration has no such far-reaching consequences, but it is an especially neat example of how strict analytical bibliography can clarify muddied waters. The quarto of Shakespeare's Troilus and Cressida exists in two forms. In one, the title-page is conjugate in its gathering, and on it the play is stated to have been acted at the Globe. In the other this title-page is cancelled and is replaced by two leaves in a fold. On the first leaf, a new title-page omits reference to this public acting. An anonymous preface appears on the second leaf, which somewhat offensively, perhaps, denies that the play was ever publicly acted and hints that it is now being printed against the wishes of the 'grand possessors.' In the first investigations, the order of these two title-pages was confused and the true first was thought to be a correction of the second. However, the physical facts of cancellation and substitution were soon discovered to set that right—a piece of analytical bibliography in itself. Thereafter, it was automatically assumed that a considerable time-lag intervened between the printing of the first form and of its cancel substitute; and on the basis of this assumption a number of rather fantastic theories were evolved to explain the reasons for the printing of the new title and the preface.

What exact circumstances lie behind the situation we still do not know as clearly as we should like, perhaps, but of one thing we may now be certain: the substitute title and preface were not added to the play some time after publication as a result of any set of hypothetical circumstances. Dr. Philip Williams has been able to show by purely mechanical evidence,

analyzing the running-titles for what they demonstrate must have been the method of printing, that the two cancel leaves were printed as an integral part of the last sheet of the text, the full sheet then being cut apart and its two halves designed one to go at the beginning and the other at the end of the quarto. This physical demonstration does not, it is true, explain the curious statements in the preface—since these are not susceptible to attack by bibliographical evidence—but it has the virtue of severely limiting the scope of future theories by establishing that the play had not been published when the change was decided upon, and that less than a fortnight could have elapsed between the physical printing of the first title and the decision to cancel it in favor of the substitute halfsheet. Whatever explanation for the anomaly that is to be offered, therefore, must not overstep the limits set by these two pieces of evidence.

It is of importance to notice that analytical bibliography in treating a book as a tangible object deliberately operates without regard for the meaning of the text contained in the book. This is not purblindness but instead a logical and necessary restriction of the discipline to its proper mechanical evidence. I may illustrate with an example of a reading from Richard II. If one is investigating the history of its text the first problem that arises when one comes to the third edition is whether this Q3 was printed from Q1 or from Q2. There happens to be sufficient textual evidence, but in fact A. W. Pollard pointed out one piece of bibliographical proof which needs no further corroboration. In a particular line the word for appears clearly printed in Q1; in Q2 the word is again for, but the f is very faintly inked. The compositor of Q3 overlooked this faint f and printed the word as or. Trivial as the example is, it has important implications for method. The point is this. If we were to rely on meaning only, no evidence could be adduced. The word or instead of for is clearly wrong, but it could have occurred as a simple misprint if Q1 had been the copy. But without regard for rightness or wrongness, we can demonstrate, on this one piece of evidence alone, that Q2 was the copy because the faint inking of the letter f provides a mechanical reason for its omission in Q3. A more pertinent use of this sort of evidence comes in Dryden's Wild Gallant where it alone is sufficient to decide the difficult question of priority for the first and second editions printed in the same year. Here, in one edition, only the tip of a letter l inked, producing a mark that could easily be mistaken for an apostrophe. And as an apostrophe it appears in the other edition. A court of law would accept this evidence, for to reverse it one would be forced to argue that in a first edition the compositor set an apostrophe and in the second another compositor in the same place set an imperfect letter l which inked so as to resemble an apostrophe. I should not myself care to be briefed for such a defense.

I have spent time on these examples because they may assist somewhat when we come to the second branch of bibliography of which I spoke—this is textual bibliography. Since it may often occur that there is a difficulty in drawing a sharp line between textual and analytical bibliography in specific cases, I offer a rather over-simplified definition at the start. I take it that textual bibliography is the application of the evidence of analytical bibliography, or at least of its pertinent methods, to textual problems where meaning of some sort may be involved even though very slightly. Or if, in the strictest possible sense, meaning is not involved, textual bibliography deals with printing evidence which—in ordinary contrast to analytical bibliography—relies on cumulation for building up its case, and on some degree of interpretation. Because it is a form of bibliography, the textual method must still firmly base itself on the evidence of the printing process but it may associate palaeographical evidence with this, as well as evidence of any sort that seems pertinent. The nearer it approaches to analytical bibliography the more certain and mechanically demonstrable are its results; the farther it dissociates itself, the more its evidence becomes speculative, cumulative in its effect, and in need of interpretation. Very much in general, analytical bibliography has a tendency to find its normal material in larger questions of fact, whereas textual bibliography ultimately attempts to apply its findings to specific and even minute points. The field is so very broad that generalisations are difficult and I had better proceed to examples.

Before any logical approach could be made to the problem of the King Lear text, not only did the Pavier quarto need to be cleared out of the way as having any authority, but also the precise relationship of the first quarto and folio texts had to be established. By relationship I mean specifically their tangible relationship: was the Folio text printed from a manuscript or was it printed from an annotated copy of the first quarto. On the answer to this question hinges the entire logical approach one must take to the variants between the two texts in general, as well as in any specific reading. It happens that there is evidence which is at least on the fringes of the bibliographical method, that is, evidence of the Folio making a wrong rationalization in an attempt to ungarble a quarto corruption. In King Lear an example comes at II.iv.100 where in its unproofread state the first quarto read, "the dear father | Would with his daughter speak; come and tends service". The Folio, struggling with this unsatisfactory passage, changed "come and tends service" to the scarcely more satisfactory "commands, tends, service". "Commands her service" is of course the right reading as found in the proofread state of the quarto; the point here is simply that the Folio error can be explained only on the basis that the quarto error served as printer's copy. Another

example of what I mean comes from Richard II. Line I.i.77 reads in the first quarto, "What I have spoke or thou canst worse devise." Q2 corrupted the line by omitting 'worse' and reading, "What I have spoke or thou canst devise." The Q3 compositor mended the metre by adding 'what' to make the line read smoothly, "What I have spoke or what thou canst devise." This form for some reason was altered in the Folio by omitting the 'what' and changing 'spoke' to 'spoken': "What I have spoken or thou canst devise." Here the chain of rationalization is clear, as we see compositors or editors reconstructing a defective line according to their own tastes. Or one can illustrate further with the incorrect Folio "had his current" at V.iii.63, the result of the compositor's bewilderment at the spelling 'hald' in the "hald [or 'held'] his current" of his copy.

It is true that these Folio variants are readily explained by our ability to trace the printing of the text through its various forms. To this extent the method is bibliographical, but it is somewhat remotely so perhaps in contrast to the truly bibliographical form which the clinching evidence took. The first quarto of King Lear was very considerably proof-corrected during the course of its printing so that all but a few sheets exist in two states of the readings. That the Folio was definitely printed not from any original manuscript but at least basically from a copy of the quarto became clear when P. A. Daniel, and later Sir Walter Greg, pointed out that in some instances the Folio text follows the uncorrected readings of certain quarto sheets, as with "commands, tends" for "come and tends," an impossible occurrence for a text set from an author's manuscript since these readings must have been copies of the blunders and misreadings of the Q1 compositor. This is very close indeed to mechanical proof, and the example places textual bibliography as near to analytical as it can come.[*] Somewhat similar investigations in Shakespeare are those of Philip Williams with Troilus and Cressida, and Charlton Hinman with Othello. In the one case such printing evidence as the common variable use of italic or roman type for names, as well as the forms taken by the speech headings, settled the long-vexed question of the relation of the Folio Troilus to the quarto by showing beyond doubt that the Folio was indeed set from a copy of the quarto and not from a manuscript. In the

other, similar typographical evidence established that in Othello the Q2 passages of added text, first found in the Folio, were set from the Folio and not from another manuscript. This latter I believe offers a splendid example of the certainty of the bibliographical method when operating on its proper forms of evidence as against the uncertainty of the critical method. So long as the case was considered only from the critical point of view, doubt existed whether the variants in the Q2 text from the text in the Folio were not the result of printing from different copy, a manuscript. And in fact I believe I am correct in stating that every text available today actually takes over one or more of these Q2 variants, as representing an actual recovery of what Shakespeare really wrote. Yet Dr. Hinman, using mechanical evidence from the printing, has proved beyond any doubt that Q2 is a mere reprint of the Folio for the added passages and that no reading from it can have any independent authority whatsoever. The bibliographical method, here, shows that criticism—with all its positive virtues—can indeed be fallible.

But I am not interested today in listing the various horrid examples in Shakespeare in which bibliography has exposed the errors of what Sir Walter Greg calls 'metacritical' writers. I do want, however, to pursue a little more narrowly some differences between true bibliography and psuedo-bibliography, because this pseudo-bibliography has sometimes masqueraded as the real thing in Shakespearian textual criticism.[1]

Earlier I emphasized that to be called textual bibliography the method of investigation must be firmly rooted in the printing process; and I distinguished between analytical bibliography in which the evidence is not textual—at least in the purest state of analytical bibliography—and textual bibliography in which the chief concern is with the printing of the text as such, and the evidence is therefore drawn from the text and its characteristics although still in a largely mechanical manner. I can illustrate now, I believe, with a concrete example which may be of assistance. If a book were set by more than one compositor, it is of the highest importance to distinguish the exact pages composed by each workman, for each in all probability will have adopted somewhat different attitudes towards the text and, at the least, each will have imposed upon the author's manuscript somewhat different spelling and punctuational characteristics. Analytical bibliography has one certain method for making

this distinction on purely mechanical grounds, although the method, unfortunately, works better for Restoration than for Elizabethan times, reflecting the difference between the fixed and the adjustable composing sticks. If one compositor takes over from another and does not have a stick of exactly the same length in which he sets the lines of type, the close measurement of the variant width of the type-page will infallibly detect the presence of these two workmen using slightly different setting sticks. This is mechanical evidence at its purest, and it is certainly analytical bibliography. Textual bibliography, using evidence based on the physical features of the text as such, may also distinguish the presence of more than one compositor by its own methods. Differences in the typographical characteristics of act and scene headings, of stage directions, of turned-over versus split lines of verse, of the setting of proper names, and of the forms of speech-prefixes have all been used with success in the dramatic texts. Of a different nature is a spelling analysis which classifies the variable habits of compositors and assigns different favorite methods of spelling the same word or similar groups of words as a distinction. Each bibliographical method employs the printing process though in different ways.

Now what I wish to point out specifically is that while the results may in some degree be inferential and cumulative, for textual bibliography, since the evidence is based on the variable and sometimes unpredictable characteristics of human beings as they perform their tasks, as against the result from the printer's stick which bases its evidence on a piece of equipment in the printer's shop; nevertheless, actual and physical evidence is employed, and inferences are drawn from a very large body of cumulative evidence, not—as with pseudo-bibliography—from mere speculations as to what might have happened in the course of printing. In other words, true bibliography even in its inferences building up to an hypothesis—I dare say even in its occasional speculations—must govern itself by laws of evidence of a nature very similar to those which govern a Queen's Counsel in court. These laws of evidence, I take it, concern the logical process by which a chain of ascertained facts is linked together to show a certain relation between presumed cause and indubitable effect. Circumstantial evidence welds these links together. In some cases time may show that because of certain missing evidence or of some logical fault in the series of interpretations, the ultimate case was wrong; but it is perhaps not unknown in courts for a judgment to miscarry because all evidence was not procurable, or because counsel made a particularly dazzling leap over what for a less ingenious man would have been a yawning chasm in the evidence. If miscarriages of justice occur, whether in law or in bibliography, the fault is not in the method but in the application. Parenthetically,

I should remark that the most dangerous fault in bibliography is for a scholar, self-convinced, to attempt to carry his evidence further than in fact it will reach without specifying the point at which he is forsaking irreversible evidence for inference, and inference for speculation.

Professor Dover Wilson's researches are frequently concerned with the bibliographical matters of the printing process and of compositorial and proofreaders' habits; but in his monograph on the manuscript of Hamlet and its printed transmissions he is prone to use personal generalizations and speculations about printing and to apply these assumptions to the text in question as if they were facts which had been previously ascertained as true by the corelation of large quantities of physical evidence derived from the actual texts under examination. This is a very dangerous method for unwary students to read, since they are often dazzled by the seemingly bibliographical display of information without recognizing that in truth the process of building up to an hypothesis through a chain of circumstantial evidence governed by strict laws has not been observed. The language of bibliography may have been used, but not the method.

It is incumbent upon me to illustrate, and I choose an example each from his consideration of the Quarto and the Folio texts of Hamlet. Professor Wilson makes two assumptions about the compositor of the good second quarto of Hamlet, our basic text, which are very important for his theory. (1) Only one compositor set the whole text. (2) This compositor was an untutored dolt who was setting difficult copy at a rate far beyond his normal speed. As proof for the first, that only one compositor set the text, he asserts only that the imperfections are so uniformly distributed throughout that it is inconceivable they belong to two different workmen. This is not bibliographical evidence, or proof, and it would be thrown out in any court of law, for if the imperfections of the copy were in some part responsible for the printed results there might well be a relatively even distribution of error. But convinced that the errors are compositorial, Wilson automatically accepts as authoritative various extra words supplied by the Folio, and then uses—with some circularity of argument—the appearance of these words in the Folio as demonstration that they were missing in the quarto. This may be textual criticism of a sort, but it is not bibliography. Bibliography would use demonstrable evidence at the very start of the argument to ascertain whether in fact one or two compositors set the text. The only scientific method of demonstration, the evidence of the printer's stick failing as it does, would be to determine by a spelling test whether variable or uniform compositorial habits are found in the spelling, leading by classification into the assignment by pages of composition by different workmen or to the hypothesis

that only one workman was concerned with the typesetting. This basic examination is only now being worked on.

But for the sake of argument let us hypothecate (what is actually untrue) that only one compositor set the second quarto of Hamlet. If he were the illiterate numbskull that Wilson makes him out to be, it is difficult to see how he was hired except as an apprentice in a competitive and rigorously controlled trade. Wilson's confident statements that this compositor could memorize no more than one or two words at a time before turning to his cases to set the required type are, of course, absurd since a disability of this nature would certainly have prevented any printer from hiring such a workman. Moreover, the very examples of anticipation errors which Wilson provides and assigns to this compositor disprove automatically his theory. But normal or abnormal, it is of some importance to learn whether this compositor was in fact immoderately rushed in his typesetting, for such a theory has some bearing on the fidelity of his text. Wilson assumes that he was, on the basis of his errors. But this is not bibliographical evidence, since the errors may have been caused by other factors. Indeed, the evidence of analytical bibliography suggests quite the contrary, and that the compositor was under no especial pressure. The details are too technical to offer here. Suffice it to say that certain assumptions can be made about the speed of presswork in relation to compositorial speed of typesetting, at this time, on the evidence of the specific manner in which the type-formes are imposed for the press, specifically on the evidence for the number of skeleton-formes constructed to hold the type-pages. That is, for this date it seems established that the apparent inefficiency of the one-skeleton method is not actually time-wasting if the press for various reasons is ahead of the compositor. On the other hand, for the two-skeleton method to be worth attempting the compositor must be comfortably ahead of the press. Hence it is more than ordinarily significant that Q2 Hamlet is printed by the two-skeleton method, an unlikely situation if the compositor were as hurried as Wilson asserts.

But the evidence may be even more specific. Sheets B, C, and D are imposed with a pair of skeleton-formes; but with sheet E two more skeletons are constructed and thereafter the two sets of pairs roughly alternate, the B-C-D pair imposing sheets F and I whereas the new pair imposes sheets G, H, K, and L, until with sheets M, N, and O the two pairs mix together according to a pattern which at the moment I have not worked at enough to comment on. This general alternation involving the use of four skeleton-formes is inexplicable for printing with one press; yet if we hypothecate two presses it follows that there must have been more than one compositor. And recently Dr. Alice Walker has communicated

to me a spelling analysis she has made of Q2 in which the results in an almost uncanny fashion coincide with the sheets imposed by different skeletons to indicate with considerable certainty the presence of two compositors. This dramatic new evidence may possibly be buttressed, her preliminary further analysis shows, by evidence of these same compositors in other Roberts' quartos. If all this proves true, and there would seem to be little question of the validity of the evidence, then so much for Wilson's non-bibliographical assumptions about the uniformity of compositorial error throughout Q2 and his highly colored picture of the harried and driven single workman. If so much for these assumptions, then perhaps so much for a number of far-reaching conclusions he has drawn about the text on the basis of his false evidence.

The case is no better for his consideration of the printing of the Folio text. In examining what he believes to be the remarkably higher percentage of error in the Folio Hamlet in comparison to Antony and Cleopatra, the purpose being to assign the error to the playhouse scribe and not to the Folio compositors, he writes, "So wide indeed is the disparity that assuming as we legitimately may that the texts were set up by compositors who were, if not the same workmen at least of similar efficiency, it can only be assigned to a fundamental difference in the copies supplied them."

Now without regard for whether the conclusions reached are correct or incorrect, the method used to reach them was only pseudo-bibliographical since it played about with speculations concerning printing but without a single piece of strictly bibliographical evidence to back up these speculations. "Assuming as we legitimately may that the texts were set up by compositors who were, if not the same workmen, at least of similar efficiency." This is not bibliographical evidence, at least not for a court of law. Why may we assume that if there was more than one workman the two were of equal efficiency? Presumably because they were both employed in Jaggard's shop. But on this logic we should be able to argue that the much maligned second-quarto compositors for Roberts' Hamlet were the equal of the workman who set up King Lear for Nicholas Okes since all were employed by reputable London printers. Moreover, it is noticeable that the evidence is begged whether there were one or two compositors for these plays under scrutiny and whether they were the same workmen. The spelling-evidence which enables us to distinguish the two Folio compositors was known at the time Wilson contrived this passage, but he did not avail himself of it to apply such tests to the plays about which he was writing. If he had, he would have found that whereas Hamlet was set up by the two alternating compositors customarily distinguished as A and B, Antony and Cleopatra was composed by B alone. Thus, at least at the start, no inferences can be made from a comparison

of the texts of the two plays in toto, but rather only by comparing that part of Hamlet set by B with his Antony and Cleopatra text. Thereupon it would be necessary to test the faithfulness to copy of both A and B. This work has recently been started independently at the University of Virginia and by Dr. Alice Walker in this country. And in her new book on Problems of the First Folio, which I have seen in proof, Dr. Walker shows that B was a more careless workman than A and was more prone to error. It could be argued, therefore, that the Wilson hypothesis for faulty Hamlet copy is correct, since Antony and Cleopatra (set in its entirety by B) is in his opinion a superior text to Hamlet (in which the more conservatively faithful A participated). But in fact, one must query the whole basis for Wilson's assignment of error in both plays. Since no previous text is available, only manifest error can be detected in Antony and Cleopatra; and we have no means whatsoever for knowing when the compositor departed from copy but yet made sufficient sense so that we cannot detect his divergence. For Hamlet divergence can be only partially known and seldom (given our present state of information) with great certainty. Nonetheless, we do have some sources for determining corruption in the Folio Hamlet which are denied us in the Folio Antony. Thus if we had the full facts about Antony, which we shall never have, the incidence of error might, Wilson to the contrary, be at least as high as in Hamlet, or even higher. After all, what should we feel if the cases were truly parallel and the only known Hamlet text were that in the Folio? To argue about the error in two manuscripts, both lost, when the materials by which the two printed results may be surveyed are so unequal is, I feel, dangerous 'bibliography.' And since Wilson's initial assumptions about the equality of the compositors have been proved wrong by Dr. Walker, one may question whether his ever was a true bibliographical argument in the first place.

When we consider the millions of words and the books piled on books that have been written about Shakespeare, it is astonishing to contemplate how little of a basic nature has been done to establish his text, and how much remains to do. One reason is that the bibliographical method for textual enquiry, which was developed first in this country by such great men as A. W. Pollard, R. B. McKerrow, and Sir Walter Greg, is still comparatively new. Another is that techniques are still being invented, or refined, to bring to bear this method, at its most powerful, to an analysis of the relation of the printing to Shakespeare's text. I turn now to a brief account of what I take to be some of the central problems and the methods that I anticipate will be applied to come to grips with them.

It is a commonplace of textual criticism that no editor is equipped

to tackle the details of his text until he has come to some general view about its nature. What we disapprove of most at the present day in 18th and 19th-century editions of Shakespeare is that the editors almost invariably picked and chose among variants with no fixed principle except their literary taste. This taste was often excellent but it did not save them from blunders, as in the example I have mentioned where up to half a dozen readings from the unauthoritative second quarto of Othello were commonly introduced as authentic recoveries of Shakespeare's own words, which they are not.

This general view of a text consists, perhaps, of two major parts. When there is more than one text available, it is important to estimate the exact relationship between them; and this leads inevitably into an enquiry about the exact process by which they descended—whether by direct derivation or collaterally—from Shakespeare's own manuscript. In this investigation it is important to attempt some analysis of what were the effects of this transmission, including the work of scribes, compositors, and possible editors. When there is only one text available, as in the plays first printed in the Folio, this investigation is of course somewhat circumscribed, but nevertheless such evidence as there is must be evaluated as to what can be determined about the process of transmission and the nature of the manuscript underlying the printed version.

Some parts of this investigation concern pure textual criticism, for McKerrow was correct, I believe, in disavowing as in any sense bibliographical the speculative enquiry which is necessary to achieve an hypothesis about the nature of a lost manuscript underlying a printed text and the history of its transmission up to the point that we can first read the text in print. But such speculations often need the checkrein of bibliographical logic if they are to escape manifest falseness.

The case is indeed worth a narrower scrutiny, for a principle is present which reaches its fullest flower in the Hamlet problem. If the Folio text were indeed set from a manuscript, independent concurrence of Folio and quarto in any reading would powerfully argue for its correctness. In order to demonstrate the need for emending any such reading we should be forced to argue either that the error was present in some manuscript which lay behind both the manuscript for the quarto and that for the Folio prints, or else that the X-manuscript reading was correct but that both quarto compositor and Folio compositor, or else a scribe copying the manuscript which either used, in addition to the other compositor, independently produced the same misreading. Such cases are not unknown, but an editor must be very sure of his ground and have palaeographical evidence to boot, before he engages himself to alter such a reading. On the other hand, if the Folio directly derived from the quarto,

though with conflation from some manuscript source, concurrence of the two texts in any reading is no serious argument for its correctness, since the reading may perhaps represent an error left uncorrected by the scribal preparer. In fact, it is a paradox of textual criticism, as Sir Walter Greg has pointed out, that when such a derivation of texts can be established, readings in which the two texts concur are much more likely to hide errors, and to be in greater need of emendation, than readings in which the two texts differ.

Of course an editorial problem still resides in the variants, whatever the relationship of the texts, since the source for those in the Folio must still be hypothecated and the authority established. It does not do to take it as axiomatic that Folio variants, coming from an authoritative source, are thereby superior to the quarto in expressing Shakespeare's final intentions. The chain of transmission which may lie behind the manuscript used to correct the quarto could have been such as to afford opportunity for errors to creep in, these errors being faithfully copied by the scribe annotating the quarto. And by no means all such errors will be apparent errors. The majority are quite plausible readings. But the case may be even stronger. Dr. Williams has argued in an unpublished paper that by and large in Troilus and Cressida, where the same situation obtains, the manuscript behind the quarto was a Shakespeare fair copy somewhat revised in the process of copying from the original, and that the manuscript used to revise this quarto for the Folio printer's copy was Shakespeare's foul papers, or original draft. Thus if one accepts a Folio reading when it differs from that of the quarto, the odds are—it is argued—that one is rejecting a Shakespeare revision in favor of a Shakespeare first thought which he had modified. The case is a very interesting one and fortunately is in process of close examination.[*]

When we turn to 2 Henry IV, some of the problems I have cited, such as hypotheses about the nature of the manuscripts which in one way or another must lie behind the quarto text, and the Folio at least in part, are not proper subjects for truly bibliographical enquiry. But in two particular respects bibliography can, I think, be of assistance. To date, the arguments on each side—one for the independent typesetting of the Folio from a manuscript, and the other for the use as copy of an annotated quarto—have rested very largely on textual evidence of a non-bibliographical cast. There is certainly some reason to hope that a minute bibliographical examination of the formal characteristics of both texts would establish the case for direct derivation, or—with more difficulty—by demonstrating the lack of sufficient evidence for a direct relationship, the probability that

the Folio was indeed set from a new manuscript. Then once the exact relationship of the two texts was determined, the problem of the variants could be tackled from a narrow and partial yet a necessary point of view. I refer to a comparison of the variants with a classification of the known habits in departing from copy of which both compositors A and B, in somewhat different ways and degrees, were occasionally guilty. It is very likely that such a minute comparison would establish the strong probability that certain variant readings in the Folio did not emanate from the authoritative manuscript which had an effect on its text, but are instead unauthoritative compositor's errors which an editor should weed out of the text. The investigation of the problem of 2 Henry IV, I think it is obvious, is still only in its elementary stages; and any scholar who tackled it with the proper rigorous method and wide sweep of evidence would be kept happily occupied for a good long time.[2]

Richard III is an example of a play in which the necessary analysis has reached an intermediate stage. The main points in its textual history are these. In the late nineteenth century the estimate was correctly made that the basic source for the Folio text was a copy of the sixth quarto of 1622. Much later two further important discoveries were made. It was effectively demonstrated that the source for the text of the first quarto (all later quartos being mere reprints) was not an authoritative manuscript with some direct transmissional link to Shakespeare's own, but instead a memorially contaminated text very likely derived from the actors reciting their parts as they remembered them in order to make up a new prompt book. It was furthermore discovered that the source for the numerous Folio alterations lay in a collation of Q6 against a new and authoritative manuscript of some sort in order to revise the faulty printed text. Finally it was found that not all the play had been so collated and revised, but that several hundred lines divided between a large patch in the opening of Act 3, scene 1, and another at the conclusion of the play, were printed unannotated in the Folio from leaves detached from a copy of the third edition.

By ordinary standards these facts illuminating a most complex case would be taken as sufficient to guide an editor in the recovery of as much of the true text as we are ever likely to know—parts, of course, never

can be recovered and we must rest content with corrupted lines of Shakespeare which we are powerless to restore. But an editor must make every endeavor to learn the how and the why of such a situation, for his estimate of this how and why may influence his treatment of the printed results. Thus it is most important for textual criticism to learn whether the manuscript used to correct the sixth quarto was a playhouse promptbook, as most critics have maintained, or else, as Alice Walker has recently argued, Shakespeare's own foul papers. Any editor will need to come to some decision on this question, for on the answer to it hinges in part his estimate as to the odds for relative purity or else for possible corruption in the Folio revisions. But before we pass to what bibliography can do, let me point out two implications which might be missed by an editor. If Alice Walker is right in her hypothesis (and I find it a very attractive one) that the quarto was not collated and annotated by Heminge and Condell or their agents to serve as printer's copy, but instead the patched manuscript was given to Jaggard to serve as his copy and for convenience he had the collation made in his printing-house so that his compositors could work largely from printed copy, it follows that if the work were done in the printing-house an editor could estimate the annotations as ingenuous transcripts from the manuscript without any possibility of sophistication from the memory of an actor-scribe such as has been argued for some other plays. Moreover, it follows that we must re-examine all other similar printer's copy, which we have automatically assumed was revised and annotated in the playhouse, for at least some of it may also have been collated instead by Jaggard, a fact that would give us a somewhat different view occasionally of the nature and authority of the variants.

At the conclusion of her case for the nature of the manuscript and the history of the text, Dr. Walker makes a very significant bibliographical analysis to guide a future editor. The specific problems are: (1) how many of the Folio variants are authoritative and how many scribal or compositorial errors; (2) how many errors in the quarto text did the collating scribe fail to correct.

Her use of evidence is very interesting. On the basis of the respective shares of compositors A and B in the unannotated section set from the third quarto, she evolves some statistics to assist an editor in estimating the proportion of Folio variants which may reasonably represent compositorial errors and not scribal corrections. She then attacks the problem of the failure of the scribe to correct. With absolute logic she offers the proposition that when Q6 by the history of its transmission from the first edition varied from this first edition, and when the scribe by reference to his manuscript restored the Q1 reading, this is unassailable proof that he had correctly transcribed an authoritative reading from his manuscript.

This happened a certain number of times. On the other hand, when a reading in Q6 differed from that in Q1 by reason of its printed transmission and had been passed over by the scribe without revision, he had manifestly failed to copy the necessary correction from his manuscript, either of the Q1 or of a different reading. This happened a certain number of times. On the basis of these statistics Dr. Walker arrives at an estimate that about 140 non-Shakespearian readings are still present in the Folio text. At least in part, the method is bibliographical, and it needs only a fuller and more rigorous examination of the respective characteristics of compositors A and B, as derived from other plays which can act as a control, letting us see in completeness their wanderings from copy,—it needs only this fuller analysis to provide an editor with the maximum bibliographical knowledge he can obtain before starting his work. What I find very pleasant in Dr. Walker's concluding assertion that her findings should encourage an editor to emend more boldly than is usual, is the paradox that bibliography which will sometimes narrowly limit an editor's discretion to alter his text, in cases like this actually opens the door wide for all the resources of textual criticism.

Certainly this added freedom which bibliography can sometimes afford an editor to be logically bold is exemplified with King Lear, and it is not too much to say that recent bibliographical work has quite upended the conventional editorial treatment of this text, and in some respects outmoded every available edition. Professor Duthie's admirable edition published a few years ago gave the text a much needed thorough analysis. He performed a signal service in disposing of the shorthand theory to explain the peculiarities of the first quarto, and in analyzing the work of the scribe who annotated this quarto by reference to a playhouse manuscript in order to furnish the Folio copy. When he came to attack the problem of the Folio variants in constructing his own text, Duthie felt confident that critical ability alone could distinguish in the main between the authority of quarto and Folio in cases of significant readings. But when he arrived at the problem of the relatively indifferent reading, he confessed that criticism was helpless and that he had no recourse but to prefer the Folio variant since it was present in a text of generally greater authority. Some hundreds of readings in which the choice seemed indifferent, therefore, were arbitrarily accepted from a single source.

Now it is clear that in all cases of variation between quarto and Folio, either the preparer scribe or the compositor must be responsible, and it seems reasonable to assume that if the Folio readings are accepted in all such cases, in a certain percentage an editor is incorporating unauthoritative compositorial departures from copy. Whatever little authority any specific reading may have in the first quarto, with its contaminated text,

the possibility that Shakespeare is being accurately reproduced is at least greater than that a compositorial error in the Folio is accidentally recovering the true reading. However, nothing could be done about this problem until a method was evolved to reach some evidence to assist an editor in separating compositorial variants from scribal corrections, especially for the large body of indifferent choices. A start has now been made on the problem by the method of compositorial analysis. It would not be surprising to find that in over a hundred cases an editor is justified in rejecting a commonly accepted Folio reading in favor of the quarto. For King Lear this is an almost revolutionary prospect, but I have no doubt that the method will be the correct one to follow, no matter how much more complicated it will make an editor's life in the future by immensely broadening the number of choices between variants which he must make.

I am emphasizing the need for the minute study of the compositors who set Shakespeare texts because it is through these workmen that what Shakespeare wrote was set into type and printed in the texts now preserved to us. What these compositors did with the words they found in manuscript concerns us closely, for—cumulatively—it is not impossible that in thousands of small but eventually important ways these workmen have altered the exact reading of their copy. Therefore, the textual bibliographer is deeply interested in finding—for the First Folio—to give an example—precisely which pages and columns were typeset by which compositor, and how many workmen were in fact concerned in the composition of the First Folio text. Possibly there were also compositors C and D. The case is still uncertain. By an analytical examination of the evidence bearing on the presswork, we seek to learn what parts of the Folio were printed on one press and what parts on two presses. We want to recover any evidence as to whether a compositor was hurried or at leisure in composing certain pages; and the method by which the sheets went through the press can tell us much about this. Then we want to know every last thing there is to know about the habits of spelling, punctuation, and of faithfulness or carelessness in following copy for each compositor who worked on the book. This evidence can all be applied in one place or another to help remove the veil which the printing process places between us and the manuscript, and to help recover the at least relative purity of the manuscript reading. Let me illustrate with an example which is not concerned with an especially common problem but which I think is interesting. Some years ago I heard Professor G. B. Harrison deliver a paper on Coriolanus, in which he argued—very convincingly at the time, it seemed,—that the irregular lineation found in this Folio play represented a faithful following of Shakespeare's manuscript. And by showing the results which the rough lines had on vocal delivery as against the conventional

smoothing out and regularizing of the lines into normal pentameter units, Harrison carried his audience along with him most successfully to accept his point of view that the irregularities were conscious with Shakespeare and represented a part of his dramatic and poetic art. Now, however, that we know something about the compositors, and learn from Dr. Walker that whereas A more often simply turned a long line under in the Folio, compositor B was just as likely to make two lines of it, it is clear that these days before we attempt to apply critical evidence as Professor Harrison did, we must be bibliographical and must break down the typesetting of the play and see where we stand. Thus since both A and B were concerned with Coriolanus, we shall need to distinguish their work and to see if there were any marked difference, such as we find in other plays, between the way the lines were split or were turned over in their respective sections. And only then, after this systematic bibliographical examination had been made and the characteristics of the text established as deriving from the manuscript or from the compositor in small or in large part, can we have a foundation of fact on which to base a critical discussion of the irregular lineation of Coriolanus. But the need to know everything there is to know about the characteristics of the Shakespearian compositors goes far beyond questions of verse-lining and into the very words that we read as Shakespeare's—we hope. For at present we can only hope in many cases of complex doubt—we do not yet have the bibliographical facts on which to base an opinion that is worth very much.

There is another gap in our knowledge of the way in which the printing process may have corrupted the text we have commonly accepted. We have very curiously never made the collation of copies that is necessary to discover which pages of an early Shakespeare text were proofread and what were the changes that took place as a result of the proofreading. I do not know of any source where a student can go to find out the facts about proof-correctors' variants in most of the Shakespeare first quartos, which are often the highest authorities for our text: so far as I know the majority have never been fully collated. This I think is a shocking state of affairs. It means that editors have been content to take their text from any random copy that was most convenient for them, without curiosity as to whether they were reprinting compositors' errors or proof-readers' sophistications. When the text has been based on the Folio the case is no better. In sorrow I quote from the preface to the latest edition of King Lear: "It has recently been suggested that a study of all the extant copies of the First Folio would reveal variants comparable in importance with those in the extant copies of the First Quarto. As no evidence has yet been produced that the Folio is made up of corrected and uncorrected sheets, I am skeptical of this theory; but I have consulted

facsimiles of two different copies of the Folio as well as the two originals accessible in Leeds, without discovering any variants." To be skeptical whether there are corrected and uncorrected formes in the First Folio is to be skeptical indeed, for in 1932 in the basic study of the printing of the First Folio Dr. Willoughby reprinted a facsimile of an actual proof-marked sheet in Antony and Cleopatra, and since then three further proof-sheets have been discovered by Professor Hinman in the Folio. This editor simply had bad luck in finding no variants in the four copies he collated, since I have been shown at least one sheet for Lear in which variants exist. Fortunately future editors can no longer remain in blissful ignorance of the facts about the Folio once Dr. Hinman finishes his collation of the seventy-nine First Folios in the Folger Shakespeare Library by means of his most ingenious mechanical collating machine, which reveals press-variants by super-imposing the images of the pages from two different copies. And while I do not myself expect that variants will be discovered which rank in importance with those in the First Quarto Lear, for the nature of the copy in the two printing houses was very different, nevertheless it would be surprising if the results did not increase our ability to ladle out the dross from the gold in various passages.

Because no one, except for the New Variorum editors, has been willing to engage himself to the undoubted drudgery of collation, it is a paradox that we know more about the variable factors in the text of certain relatively minor Elizabethan playwrights like Field and Massinger than we know about Shakespeare's text. But collation of the authoritative texts is not all that is needed. It is true that usually there is no textual service to the discovery of variants in late and unauthoritative reprints of the basic texts. But I should like to point out one overlooked situation in which it is important for us to know if the text was indeed altered in any respect during the course of printing a late and unauthoritative edition. Let us take Richard III as our illustration, a play—it will be recalled—in which an example of the sixth quarto was conflated with a manuscript to serve as printer's copy for the Folio. When the Folio deviates from Q6, therefore, we may take it that in ordinary circumstances the different reading came from the manuscript. But I have already commented on the fact that not all such variants will have this authority, for it is positive that some will have been made by the Folio compositors. These, of course, should be thrown out in so far as we can develop the bibliographical means to enable a textual critic to isolate them. But even if we then disregard variants made by scribal blunder (which will not always be easy to detect), one further class may remain, even though only a handful of variants could be concerned. If the formes of Q6 were proofread while the sheets were being printed, we should have some variant readings

in this edition. And if some of these variants were present in the particular example used by the scribe to prepare copy, and if he left some unaltered (as we know he left a number of errors unaltered that can be detected), we should have still more concealed errors masquerading as true manuscript readings of authority. I do not know for Richard III whether this situation exists, nor does anybody else, least of all any editor of Richard III, for no one has taken the trouble ever to collate a number of copies of Q6 to determine whether there are press-variant readings in this edition. Yet no bibliographical editor would dare to tackle the text without making this elementary study, even though he anticipated what he would doubtless find, purely negative or trivial results.

It is time for me to summarize what the bibliographical method can still contribute to assist with establishing Shakespeare's text.

First, it can re-survey every doubtful case, such as Hamlet, Othello, and 2 Henry IV in which there is still controversy whether the Folio text, which differs considerably from that of earlier quartos, was set directly from a manuscript, or else from a copy of some quarto edition which had been annotated by comparison with a manuscript or, as I believe for 2 Henry IV, had been used as the basis for a manuscript which served as Folio copy. This is not an academic question, for the principle of one's whole treatment of the text rests on the answer to the central problem: what exactly was the nature of the Folio printer's copy.

Secondly, by means of the necessary collation and the interpretation of the variants made in printed texts while they were going through the press, the bibliographical method can discover and classify these proofreaders' alterations with a view to recovering the true Shakespearian readings, or at least to distinguishing the certainly non-Shakespearian.

Thirdly, since the bibliographical method is concerned with the effect of the printing process on Shakespeare's text, it can make searching analytical studies of the relationship between the compositors (those intermediaries by which the text was transmitted to us) and the actual printing. Analytical bibliography can still find material to work on in Shakespeare by investigating the speed of presswork which, if suddenly accelerated, might affect the accuracy of the compositor hurrying to keep up with the press. A combination of analytical and textual bibliography, moreover, is necessary to distinguish in every Shakespeare text the number of compositors engaged on it and the exact sections for which each set the type. Then, by a series of extremely rigorous studies, the characteristics of these compositors must be analyzed in other printed texts where proper comparisons may be made; and finally everything that we know about the care or the negligence with which they treated what they were typesetting,

including what we can learn of their characteristic habits of altering text, must be applied to their work in the Shakespeare canon. Only when all this mass of investigation is done, perhaps years hence, can we say that we have honestly and thoroughly squeezed every last drop out of the possible evidence for deciding in every minute detail, as close as we can come to it, what Shakespeare actually wrote. I do not assert that this bibliographical approach can displace the critical approach, for the two methods must work hand in hand supplementing each other. But I do argue that as scholars we have been most delinquent in failing to utilize the bibliographical method to its fullest extent and in preferring the easier delights of armchair critical speculation to the rigors of first ascertaining the fullest body of facts by bibliographical investigation. Only when all the facts are known can we proceed to the necessary critical synthesis which will make of our attempt to recover the maximum of what Shakespeare truly wrote more than a thing of shreds and patches tailored according to our personal preferences.

Notes