University of Virginia Library

Some Relations of Bibliography to Editorial Problems [*]
by
FREDSON BOWERS

BIBLIOGRAPHY CONCERNS ITSELF WITH EDITORIAL problems not as a usurper of the functions of legitimate criticism, but instead as the necessary foundation on which, in certain investigations, textual criticism must be based and to which criticism must constantly refer for more or less definitive judgments. Bibliography, in W. W. Greg's acute phrase, is the grammar of literary investigation. This position was strikingly advanced in Greg's classic address on the relations of bibliography to literature which appeared some time ago in Neophilologus, and, again, in his "Bibliography—An Apologia" (The Library, 4th ser. XIII [1932], 113 ff.). Any attempt merely to restate Greg's views here would be quite superfluous. For some years, at least among scholars, the inferential identification of bibliography with textual criticism has been so firmly established that to talk about the contribution of one to the other might appear to be like a solemn discussion of the usefulness of the arm to the hand.

Nevertheless, it is possible to suspect that, as more and more scholars have come to deal with these twin methods, the popularizing of their disciplines may have led, by subtle degrees, to a rather over-simplified view of the basic provinces of bibliography and textual criticism, especially as applied to editorial problems. Certainly the theory that every bibliographer is professionally qualified to be a full-fledged textual critic and editor is quite wrong. Correspondingly, there are a number of distinguished

textual critics whom, according to any strict accounting, we might hesitate to call bibliographers.

Some fortunate scholars like Pollard, McKerrow, Greg, and Percy Simpson may happily combine in themselves both functions, although perhaps in unequal proportions; but this two souls in body one is certainly the exception rather than the rule. On the other hand, the most prominent Elizabethan textual critic now living—I refer to Mr. Dover Wilson—often operates in that dark region of the ur-state of a text which McKerrow rightly hesitated to identify with bibliography. One should comment, however, that not all of Wilson's readers—or his critics—are completely aware of the distinction. It is somewhat perturbing, for example, in Hereward Price's admirable "Towards a Scientific Method of Textual Criticism in the Elizabethan Drama"[1] to find his chief whipping boy, Dover Wilson, consistently referred to as representing the bibliographical school. Save for his impatience at various of Wilson's well-known aberrations, Mr. Price—no more than the rest of us—would, I should think, completely discount such investigations. But we may agree with McKerrow and Greg that this scholarship is not usually bibliographical in any sense in which we should be prepared to use the term.

Bibliography is likely to be a vague and misused word because it has come to be employed for too many different purposes. For the present discussion we may rule out enumerative bibliography—the making-up of finding or reference lists of books on various subjects—and use the term to mean analytical bibliography, the investigation and explanation of a book as a material object. Yet even within this narrower limitation there are various strata and subdivisions of research. For example, we have such forms of bibliographical spadework as the identification and recording of type-faces as by Haebler, or of STC printers' ornaments as is now in progress by F. S. Ferguson, with their dates and history.

But when we move from such essential preliminary investigation, and from the recording of material data as a part of publishing history, to another sort of bibliography, more specifically of the kind we designate as analytical, we come to a field in which all our accumulated knowledge of printing practice and history is

devoted to the examination of individual or related books as material objects, with a view to determining the facts of their production. Is there or is there not a cancel in a certain gathering; were the various sheets of a book printed seriatim or simultaneously in two or more sections; did one or two compositors, one or more presses, work on this book, and if so which parts did they do; was a cancel title-page or a cancel in the text printed later or as a part of the continuous printing of the book. Are variants in the sheets of the book the result of different impressions, or do they result from different typesettings or from simple correction at the press. If some of the sheets are of a single impression but certain others exhibit variant typesettings, which setting is the original, and why was the resetting made. If there are variant imprints, which was first through the press. And so on. It will be seen that correct answers to some of these questions depend not only on the simple discovery of the facts themselves, but more particularly on the satisfactory explanation of the ascertained facts. From specific analyses of this kind, some speculatively minded bibliographers may endeavor to evolve new techniques for bringing to light yet further secrets in printed books, techniques which may enable us to attack successfully some general problems which have heretofore been thought insoluble.

Such technical examinations of individual books for their own sake are proceeding very swiftly these days, and they are being matched step by step by the parallel development of techniques for determining with greater precision the relation of the printed book to its underlying manuscript.[2] No textual critic can afford to ignore the results of these investigations. From Willoughby's and Hinman's explorations of spelling tests,[3] assisted in certain cases by evidence from the varying length of the compositor's stick,[4] we are now able to distinguish with some certainty in the STC period and the following years the precise pages of a book set by different compositors. By an equation of these compositors'

habits we can thereupon make certain assumptions about the characteristics of the manuscript from which they were setting. The cooperative efforts which have been made in the analysis of the evidence of headlines[5] have opened up possibilities, in combination with this compositor evidence, for determining sometimes quite minute but necessary questions of the presswork and proofreading. To these may be added—among many other investigations—the classic Carter and Pollard examination of paper and type which exposed the Wise forgeries, Hazen's use of identified paper to detect Strawberry Hill forgeries, and Stevenson's technique for aligning watermarks with presswork.[6] The unexpected uses of bibliography have been demonstrated in Bond's assignment of authorship to certain doubtful Spectator papers, the whole argument resting ultimately on type left standing in advertisements and headings.[7]

Astonishing results have recently accrued from certain research on eighteenth-century press figures, which had been earlier dismissed by McKerrow as of little bibliographical significance. Knotts[8] has added a chapter to the work of Sale, Chapman, and others; and W. B. Todd's latest interpretation of the evidence of these figures in a number of books demonstrates some most exciting things one can learn of printing.[9] From this evidence a whole new area of research has been opened up of such importance for distinguishing impressions, partial re-impressions, and editions, as well as various other important textual matters, that we may well believe in many cases press figures alone will come to be the most valuable tool for penetrating the extreme difficulties of textual problems in frequently reprinted popular eighteenth-century authors.

I should be less than candid if I tried to pretend that the immediate or even the chief aim of these and of many other bibliographical scholars was to serve as the sons of Martha to textual criticism. Without question, a number of bibliographical investigations which eventually proved invaluable to textual criticism were not undertaken with this end immediately in view and almost by accident arrived at their textual applications. Thus a real difference from textual criticism must always be felt, I suspect, in bibliographical research in its purest aspect. To defend themselves against doubters, bibliographers are accustomed to argue that the ultimate aim of all bibliographical research, and therefore its justification, is the definitive account of books in a descriptive bibliography, or the direct application of their findings to textual criticism. But I make bold to say that—at least as I view it—the immediate end of a great deal of research is no such thing. Although I believe that the conventional justification is valid for the ultimate aims, and that this rationale is deeply felt by almost every bibliographical student, nevertheless, even though eventual application is the ideal, ordinarily textual criticism has not provided the immediate spur to the investigation and no specific textual use is often anticipated, at least in the early stages of the work. This attitude develops, in part, because many technical studies that are absolutely necessary yet require no interest in textual criticism on the part of the investigator. In part, because various of these studies are concerned only with a fragment of some total problem, the whole of which must be unraveled before any of it can be made of practical use to textual criticism.

Nonetheless, in spite of the various legitimate reasons which may be advanced to explain the rationale for much technical work in bibliography, the point of view does in fact serve to separate many bibliographers from textual critics; indeed, it would be tempting to say most bibliographers were it not an increasing tendency for a number of analysts, with or without due training, to cross over the line and to concern themselves directly with the editing of texts. Others, though not editors, may develop so great an interest in textual questions that they direct their research specifically towards bibliographical problems which do have an immediate textual application and thus underlie accurate criticism. The record, while respectable, has not been one of unalloyed

success, perhaps, and when bibliographers have concerned themselves with matters of textual criticism that were not primarily bibliographical, they have sometimes floundered pitifully if they lacked the training and the qualities of mind which are necessary for textual criticism in its purest state.

I use this statement of a point too often overlooked, in our current tendency to exalt the bibliographer, as an aid towards a cursory glance at the nature of textual criticism, a subject on which I can speak with little authority and on which I shall anticipate correction. It is perhaps absurd to set up too sharp a distinction between textual critics and textual bibliographers. Anyone, we may suppose, who is concerned with the origin, derivation, authority, and correctness of a text, in whole or in part, is a textual critic, no matter what his method of approach. Some scholars tackle problems through bibliography, some through what Greg has called metacriticism, some through an attempted or actual combination of bibliography and criticism. For purposes of distinction I must deal in blacks and whites more starkly than is perhaps realistic. Nevertheless, I am sure that when we distinguish between a McKerrow and a Dover Wilson, both brilliant examples of different schools, we are in fact implying a greater fundamental difference than exists between the conservative and speculative wings of the same method.

Greg—and in this he is frequently misunderstood—insists that the bibliographer must view a book only as a material object.[10] This is a far cry from the aims of a textual critic. Historically, textual criticism developed long before analytical bibliography; and it evolved its own rules, especially for dealing with manuscripts, both classical and vernacular, according to certain principles which are outside the strictly bibliographical range, if for our purposes we continue to limit the definition of bibliography to analytical bibliography. These principles, or something like them, were also applied to printed books, again before bibliography was more than a gleam in the eye, and—as Greg has shown in his paper for this series—they are still in part applicable to certain problems. Thus there has been established in the past a firm tradition for dealing with texts; and it would be foolish

for us to believe that analytical bibliography has displaced this tradition, or even that in many cases it could. Nevertheless, analytical bibliography has demonstrated that in some circumstances the classical tradition is not completely self-sufficient. We may put it, therefore, that bibliography is neither a usurper nor a poor relation in the field of textual criticism, but rather its foundation, the grammar of the subject.

If, in the interests of brevity, we pass over the important field of manuscript texts and concern ourselves with textual criticism based chiefly on printed books, we may, perhaps, discern three main but by no means mutually exclusive lines of endeavor. The first concerns itself with the authorship, origin, and characteristics of the lost manuscript behind a printed book both in whole and in part. I am thinking first of various studies in the attribution of anonymous works, or of the solution of problems in multiple authorship as in the Beaumont and Fletcher plays, Massinger, and Dekker; and secondly of such studies as Mr. Duthie has made in the bad quarto of Hamlet and more recently in the Pide Bull King Lear. Dover Wilson's reconstructions of the manuscript copy for Hamlet and other texts should be placed here, as also various studies such as Hoppe's of Romeo and Juliet but especially Greg's of Doctor Faustus. As with the Faustus and Lear investigations, according to the circumstances this division may merge imperceptibly with the second, which I take to be concerned with the critical analysis of texts in known manuscripts or printed exempla. Here we might place studies like Grierson's on the text of Donne, Wolf's on commonplace books, Shaaber's on 2 Henry IV. The determination of the order and authority of printed texts after the first also applies, and we must certainly include the very difficult task of separating from compositor's, proof-reader's, and printing-house editor's alterations those true revisions from an authoritatively corrected copy used as the basis for a later edition. The relation of the variants in the Quarto and Folio of King Lear or Troilus and Cressida is such a problem. In all of these divisions various special studies may enter in one form or another. I list as examples a few of the inquiries which Mr. Price sets up as among the subjects for textual criticism: vocabulary tests of authorship, and also of plagiarism, metrical tests including study of broken lines, deficient lines, redundant syllables, feminine endings,

prose as blank verse, and so on, applied to form an opinion of a text.

Finally, perhaps we may assign as a third general division the orderly bringing together of all this information in an editorial capacity, and the consequent evolution of a modern critical text designed to represent the intentions of the author more faithfully than any single preserved manuscript or printed copy. Whereas the findings of analytical bibliography may or may not be applicable in the first two divisions, the claims of analysis enter full force in the editorial third.

In the first two divisions, it is true, the investigations lead hopefully towards the ideal of a definitive text, but, as with pure bibliography, much textual criticism is undertaken either for its own sake or else to attack only a fragment of the total problem, so that the immediate end of the research may not be the formation of a text. For example, Mr. Duthie did not edit Hamlet after he had solved the problem of the bad quarto. It will be convenient, however, to treat textual criticism and its relations to bibliography in its narrower application to the evolution of an edited text and to the various delicate problems that arise before this text can eventually be fixed. Moreover, the limitation may be extended to cover only the specific problem of old-spelling critical texts, for—as Greg's paper has shown—these are subject to certain criteria which do not always apply to modernized versions, even though the basic problems are shared in common.

At the start one should distinguish certain editorial problems that have no necessary connection with bibliography. An editor of the literature of the past must have considerable linguistic attainments, or ready access to professional advice. Through long familiarity he must grow to be a native in the characteristic thought, usage, speech patterns, and customs of his period. Although bibliography may occasionally assist in the solution of some problems, or offer a convincing after-the-event confirmation, much emendation—or refusal to emend—much estimate of authenticity, must be made quite independently of bibliographical considerations and instead on a philological basis. This aspect has no relation to bibliography, and it requires a discipline and study which leave little time for bibliographical investigations not concerning the problem immediately at hand. Moreover, if

we speak of only one characteristic, the great emendations have been inspired art and not systematic science. One can give a rational palaeographical explanation to derive "a table of green fields" set in type from the crabbed script of "a" babld of green fields," but I beg leave to doubt that this famous emendation would to the present day ever have been arrived at by strictly palaeographic reasoning. Greg, I believe, has no very high opinion of palaeography as giving more than a hint for emendations,[11] although its confirmation is often most valuable. Usually, I suspect, one arrives at an emendation of any subtlety by inspiration, memory, and a strong sense of analogy, and then one brings in palaeography if possible to justify one's conclusions.

Many other considerations of text are too frequently confused as soluble by bibliography. For example, bibliography can establish that edition B. was printed from edition A., and not from an independent manuscript; but if revisions appear in B., there is no bibliographical technique for determining except in isolated cases whether they derive from the author himself, a scribe revising the text of A. from a manuscript, or an editor of some kind. Such problems in one way or another involving emendation or the acceptance of variant readings fall to the lot of the textual critic once bibliography has cleared his way to the limit of its ability.

This critical acumen, which we cannot value too highly as applied to text, is, of course, the product of a keen and imaginative mind; but, again, it is materially aided by a very close acquaintance with one's period in general philological considerations. This acquaintance gradually develops an opinion about speech and imagery which, in mature and thoughtful hands, has its own authority not only in questions of emendation but in any division of textual criticism. We may note that even the conservative, bibliographical McKerrow does not disdain in certain circumstances to write about the authority of variants between editions which best seem to have the greatest internal harmony with an author.[12]

This strong and subtle imagination necessary for close inquiries

into texts may apply itself to studies in attribution, plagiarism, or multiple authorship, or to studies in an author's speech and metrical characteristics, or it may be found in Dover Wilson's various attempts—whether rightly or wrongly—to discover layers of revision in the pre-printing history of a manuscript, or it may be utilized in the scrupulous examination of the sources and methods of corruption in a text which Duthie has demonstrated in Hamlet and in King Lear. Such striking deductions as those Duthie makes about the composition of memorial texts from tag ends of general recollections exhibit a critical virtuosity which has its own discipline and rigorous training.

If I were a textual critic concerned with such matters as I have been sketching, problems involving my total powers as a critic, I fancy I should be rather impatient with a bibliographer who insisted that I should be all this and McKerrow too. And I should be strongly inclined to reply: you are a technician—you do the technical part of this business and I shall apply your findings, taking care that I have studied the principles and general methods of your craft seriously enough so that I can follow your arguments and understand the applicability of what you are saying.

This is roughly what has happened, and as a consequence something like the following rationale is commonly accepted. The bibliographer's function is to prepare the general material of the texts, when bibliographical investigation is necessarily involved; and the textual critic, in the light of bibliographical findings, can then proceed to apply the discovered relationships, and to add his own art, to achieve the finished, definitive result.

Since we are an age of specialists, this separation of function seems reasonable to us. Indeed, in various cases it may work very well, and in some it may even be positively necessary. Yet I must confess it is a position I held with more conviction in the past than I do today, and I anticipate holding it with even less conviction in the future, especially if certain far-reaching bibliographical speculations and experimental techniques for dealing with the accidentals in an old-spelling text ever reach success. This change in attitude, I am conscious, may have been dictated too much by various experiences with amateur editors, from whom the good Lord deliver us, by a tendency to over-emphasize the importance of a close reconstruction of accidentals, and by allowing the

special problems of distinctly unusual texts to bulk too large in my mind so that the uncomplicated, run-of-the-mill variety is obscured.

Whatever the cause, I have here what is only a selection of examples to illustrate that a textual critic, when he is himself incapable of applying advanced bibliographical techniques to every detail of an old-spelling text, can seldom achieve absolute authority in his results, and may indeed be led into serious error through the false confidence induced in him by the notion that bibliography has sufficiently prepared the way before his labors have begun.

This hypothetical critic can ordinarily be prepared only to follow and apply bibliographical arguments: his training has not prepared him to evaluate their correctness on technical grounds. A small but rather interesting example occurs in Thomas Southerne's Disappointment of 1684. Here the case for a cancel to abridge a censored scene was very plausibly advanced in 1933, and in 1946 was vetted by a good bibliographer who made some necessary modifications but in no way questioned the central thesis. Yet in the last sixteen years if any textual critic had treated the scene in the light of this apparently authoritative evidence, without testing it bibliographically for himself, he would have been quite wrong, for no cancel exists and the scene is not abridged.[13]

Secondly, this critic may often be forced to enter upon subjects where rigorous bibliographical investigation has not yet been made, although he may not be aware of that fact. A really egregious case of false bibliography has only recently been corrected by W. B. Todd's study of The Monk.[14] The full story is very involved, but the point is brief. If between 1935 and 1949 any critic had blindly relied on the accepted ordering of the publishing history which pseudo-bibliography had set up, he would have been led to evolve a text which treated Lewis's revisions as the original readings, and the readings which Lewis had discarded as in fact his later revisions. Q. E. D. Don't trust all the bibliographers.

This is a pretty fix for a textual critic, since he is damned if he does and damned if he doesn't. And in fact, in many cases no solution for his dilemma is possible unless he is himself also a trained bibliographer who is capable of re-examining technically a fouled-up problem. Yet he must do something. One course is to trust only some bibliographers. The very best butter. Nonetheless we may well remember in this connection Curt Bühler's favorite quotation of the observation made by A. W. Pollard on his colleague Robert Proctor, a great pioneer in incunabula studies, "that in matters of bibliography he would not have taken the results of an archangel upon trust." As an example we may survey two specific but fortunately minor instances in which Mr. Duthie in his new edition of King Lear was misled by untested bibliography. A small part of his arguments concerning certain peculiarities of the quarto rests on Greg's speculation that two compositors with different habits might have been engaged with the book. Now in this case Greg was not speaking ex cathedra as the result of a detailed bibliographical examination, but only in terms of possibilities. Fortunately the point was of no very great consequence for Duthie's conclusions, because he accepted the speculation on trust and did not test it. We know now, however, from a recent bibliographical study, which can be confirmed by even stricter bibliographical evidence,[15] that only one compositor set the book. These peculiarities, therefore, require another explanation.

In the second instance the point of discussion is the mislining of the verse at the opening of Act III, Scene 2, the great storm scene. Greg took the view that the early part is mislined because the compositor was setting the text as prose, and it was not until he reached a certain point that he recognized he was dealing with verse and thereafter proceeded to line correctly. Duthie, naturally, accepted Greg's explanation as 'bibliographical,' and quite properly added that of course the compositor must have gone back to insert capitals at the beginning of each line in the prose section to give it the appearance of verse that it assumes in the printing. However, a single piece of bibliographical evidence demonstrates the mechanical impossibility of such a procedure. Throughout Lear the compositor of the quarto used a short stick for verse and

a longer stick for prose; and he never sets prose in his short or verse measure. Since the opening lines of this scene were set in the verse measure, not the prose, the explanation for the mislineation must be sought elsewhere, for from the start the compositor intended to set verse. But other consequences resulted from this failure to see the bibliographical evidence. Because the implications of the short measure were not recognized, in these lines Duthie over-conservatively retained a corrupt lack of punctuation which succeeds in destroying, almost completely, not only a successful flower of rhetoric but also one of the mightiest images of this great storm scene.[16]

If the critic finding himself thus abused hurls a curse on this treacherous science of bibliography and betakes himself to his own estimates—to what Greg calls metacritical evidence—he may find himself in a rather vulnerable position. A simple, though typical, case occurs in The Dumbe Knight of 1608 which is preserved with either one of two title-pages, the major difference between them being that one gives the author's name whereas the other does not. Metacritics worked up a pretty romance about this play, conjecturing that Gervase Markham, the author, was so disgusted when he found his play printed with an inferior under-plot by another hand that he withdrew his name from the title. On the contrary, bibliography demonstrates that the title with the name is the true cancel leaf so that the name was added rather than excised.[17]

Vulnerability increases when the problem is one of any complexity. Here is a small problem yet one in which an old-spelling editor must make a decision. The inner forme of text sheet A of George Sandys's Christs Passion, a translation in 1640 from Grotius, was reset in the course of enlarging the edition. One typesetting of this inner forme agrees largely with the invariant outer forme in making up capital W's from two V's, whereas the other uses regular W's. Aha, says the critic, I'm not so bad at bibliography myself: obviously the setting of inner A which is consistent with outer A must be the original for me to reprint. The only trouble is, he is wrong. A rather technical interpretation of the evidence

of the headlines demonstrates that this setting must have been printed a sheet or so later.[18] The inconsistent forme is the authoritative original which must be taken for the copy-text, and it is probably inconsistent because we may be dealing with two compositors casting off copy and setting by formes in order to get the book started in a hurry.

We do not need the example of nineteenth-century editors basing Shakespeare texts on the Pavier Quartos to indicate some of the larger dangers of inferences about the relations of texts made on purely metacritical grounds. But lest we feel too confident about our present abilities to cope with matters more properly the province of textual criticism, such as our ability to construct a correct family tree of editions on the basis of their readings alone, we may consider the case of the ninth edition of Dryden's Indian Emperour in 1694. Certain of its readings are drawn apparently at random from those which are unique in the seventh edition and are not found in the eighth. Others derive from equally random readings unique in the eighth and not found in the seventh. There are three reputable critical interpretations of this phenomenon: (1) the ninth edition comes from an independent manuscript; (2) the seventh and eighth editions were collated at the printing house and the conflation of readings thus results from editorial intervention—or one might even introduce collation of one of these editions against a prompt copy or other manuscript; (3) the ninth is set from a lost edition which may be placed in between the seventh and eighth and on which the eighth is also based. This last would be a favorite with critics accustomed to dealing with manuscripts, and it is certainly the most plausible.

At least the first two of these quite reputable explanations might cause some critical perturbation, since fresh authority could have been introduced into the text, and the various new readings in the ninth not found in either the seventh or eighth might demand critical acceptance on their merits. However, when analytical techniques are applied, the answer is not hard to find. Two compositors were employed on the ninth edition, but—perhaps because only one press was available—instead of chopping the reprint up into two sections to be set and printed simultaneously,

they set the book seriatim but working in relay. One would compose from four to five type-pages, or only one or two, and then be relieved by the other while he distributed his type, and so on in turn. Since the eighth edition was a paginal reprint of the seventh, what actually happened was that one compositor set from a copy of the seventh, and the other from a copy of the eighth, and so they worked merrily along in relay to the confusion of scholarship, conflating the two editions to form the ninth. The answer is very easy after bibliographical analysis.[19]

I do not wish to infer that in all cases bibliographical evidence is applicable. To be properly bibliographical, evidence must concern itself with only certain relations between preserved printed or inscribed pieces of paper. For example, a form of textual criticism linked with bibliography can usually decide on an over-all basis whether variations in a later edition are compositors' variants or editorial revisions; and pure bibliography can demonstrate whether or not this later edition was otherwise set from a copy of an earlier. But whether these revisions were the author's or someone else's can never be decided by bibliography under ordinary circumstances; that is for textual criticism pure and simple. The case is usually not demonstrable by any form of bibliographical evidence. Some critics may deal extensively with evidence which is probably bibliographical at bottom because it is founded on the peculiarities of printed inscriptions on paper; but the inferences they draw may have no relation to the laws of bibliographical evidence. As I shall indicate in a moment, strictly bibliographical deduction is not always possible from bibliographical facts. It is this common confusion about the difference between the strictly bibliographical and the metacritical interpretation of the evidence offered by a book as a material object which places, wrongly I believe, a number of Dover Wilson's ingenious arguments in the field of bibliography. Larger examples may be cited. The Taming of the Shrew problem can have no bibliographical basis, for whatever relation existed between A Shrew and The Shrew antedates their printing, and the manuscripts behind the printed copy of each have no bibliographical relation. Somewhat less clearly, perhaps, the formation of a critical text of Hamlet

is not, basically, a bibliographical investigation. There is no direct relationship between the three printed texts.

But when pure textual criticism concerns itself with problems in which bibliographical investigation is possible, it can seldom be equally definitive as in this other field, for critical interpretation of evidence is at best inferential, and the logic of the argument is frequently reversible. Until the problem was directly tackled by bibliographical methods, the question of whether Troilus and Cressida in the Folio was set from an annotated copy of the quarto or from an independent manuscript had yielded no convincing answer from the critical approach. Yet Mr. Williams has made the problem seem like child's play,[20] and we may find his study very illuminating for the relative validity of bibliographical and critical methods in such situations. Similarly, Mr. Hinman's authoritative study of Othello and its second quarto contains a most ingenious bibliographical solution of another long-vexed critical problem.[21]

There are other problems involving the relations between editions. Greg, I think, has remarked that given two editions only, the critical method could never satisfactorily establish their relationship or indeed, if they are similarly dated, as with certain of the Pavier quartos, their priority. I may be making this more positive or detailed than his original statement, but when we realize that Greg is not thinking in terms of inferences establishing high probability but instead of absolute proof, then we must admit that true demonstration is impossible because of the reversibility of the critical evidence. Is a variant a correction or rationalization in one edition, or a corruption in the other—frequently the case may be argued either way.

An example is Dryden's Wild Gallant, which has two editions in 1669, the year of its first appearance in print. One is clearly a paginal reprint of the other, but the question is—which? On the one side we have Macdonald setting up his number 72a as the first edition, whereas Griffith, backed by Osborn, argues for 72b, or Macdonald's second edition. As a part of a fresh bibliographical examination of this play I have elsewhere tried to show in detail

too lengthy to be summarized here that while the basic fact on which each hinges his argument is bibliographical, the arguments from this fact are only inferential and therefore do not conform to the strictest requirements of bibliographical reasoning, that is, to a mechanical demonstration for which there is no possible alternative save in the realm of purest fantasy.[22]

It will be pertinent, however, to consider more fully in this play certain kinds of evidence which, for a very good reason, have not been published. The exposé, I hope, will offer an enlightening example, though at my own expense, of what bibliographical evidence is and what it is not. I can speak very feelingly on the subject because when I came to the problem as an editor I was completely booby-trapped, and at first I arrived at certain conclusions on grounds which I gullibly persuaded myself were bibliographical only to have the whole argument blow up in my face when at length the true evidence became apparent.

A textual collation of the two editions quickly established that there had been no real rewriting between them, but that in a score of places—if 72a were first—corrections, possibly even revisions, had certainly been made in 72b, a few of which could have come only from the author himself. On the other hand, if 72b were first, in various readings the text had undergone a corruption rather more serious than one might expect in a first reprint. Mr. Osborn in his notes to Griffith's argument had confidently pronounced B. the first on the evidence of this textual degeneration in A.[23] This is a critical touchstone which experience has shown to be sound, provided the further inference is made that the author had no part in producing the reprint. However, I approached the Gallant fresh from a study of The Indian Emperour, in which without question Dryden had directly concerned himself with revising not only the second but also the third edition, the year before and after The Wild Gallant respectively. With this example before me, I felt hesitant to accept Osborn's conclusions without further inquiry, for the second inference could not be automatically applied. Indeed, I came to feel that there was just about as much chance that the careless errors of A. had been rectified

in B. by an errata list sent to the printer as that A. represented a careless and unauthoritative reprint of B. And I still do not think that the priority of one or other edition can be positively demonstrated on the readings alone. Whether one is corruption or the other correction is not demonstrable in the very strictest sense once we know that Dryden had concerned himself at this date with correcting reprints of at least one other play.

I approached the two editions, therefore, not on the basis of the respective 'goodness' of their readings, but in search of some material evidence that one had been printed from the other. Here are some of the highlights of the preliminary case I evolved for the priority of A.

In the A. edition, the shortened name Will is almost invariably followed by a period to indicate that it is an abbreviation for William, and this period substitutes for other necessary punctuation such as commas, semi-colons, or even question marks. In B., Will is treated as a simple familiar name and no period is ever found except once—and significantly this once is the first time the name appears. Inadvertent following of copy by B. in this single initial instance seemed the best answer.[24] Given this hint, I continued looking for what one might describe as fossils in B. of strong characteristics present in A. but not in B. Twice in A. one compositor set the extreme contraction b's for be's, and, once later, h'll for he'll. In B. the two h's are normalized to the conventional he's, but the fossil h'll retained. More evidence appeared of what was apparently inadvertent following of copy. For example, the two compositors of A. spelled the conjunctive adverb than indifferently as then or as than. In B. the invariable spelling is than, even when then appears in A., except for two cases, and one of these turned out to be in the uncorrected state of a press-variant forme in B. where the A. spelling then originally appeared in B. but was changed by the proof-reader to than. There were various other instances in which the isolated appearance in B. of marked characteristics in A. seemed significant, especially since the two pairs of compositors were different in each edition.

Moreover, on the textual principle that the 'harder' reading is the original, there were at least two cases which seemed to point to the priority of A. In the first, the form wall in the phrase in A. "she wall write" seemed to derive more plausibly from a manuscript colloquialism wall or wull than from a misprint of the B. "she shall write." In the second, Constance is deceiving Lord Nonsuch that she is with child, and when he demands the name of the father she responds that she does not know. He exclaims, "Not know! went there so many to't?" and in A. she answers, "So far from that, that there were none at all, to my best knowledge, Sir." In B. the repetition went substitutes for were, and it seemed a plausible hypothesis that the 'harder' reading was were, and that memorial failure in the B. compositor carried over the repetition of went from Nonsuch's line. That the A. compositor in such a circumstance, where the repetition seems most natural, saw went in his B. copy and set were seemed more difficult to believe.

There is not time to go into this next evidence, but in the light of general experience that irregularities between catchword and following word in an original edition usually tend to be normalized in reprints, it seemed to me at first highly probable that certain irregularities in the A. catchwords were much better explained on the belief that it was the first edition rather than the second.

Finally, I attacked the gap of eight pages in the pagination which appears between sheets G and H in both editions. In one edition this identical error could have resulted only from slavish following of copy. Hence if I could establish a reason for the gap in one edition, but no reason in the other, I felt I should have my original. Analysis disclosed that edition A. was set by two compositors in sections of one or more full sheets and that their respective sections can be identified not only by the different lengths of their sticks but also by the difference in the use of skeleton-formes which is associated with each.[25] In B. there is no evidence for any interruption in the presswork between sheets G and H, and indeed it seems likely that one compositor set the last pages of G and continued over into the first pages of H as

part of his stint. On the contrary, in A. the gap in pagination occurs between one of the clearly marked sections where the compositors shifted, and it seemed a reasonable inference that when the compositor of sheet H returned to duty and came to make his first impositions, he forgot the system under which he was working and miscalculated the pagination.

I had, then, among various other pieces of evidence, the apparent survival in unique forms in B. of strongly marked compositorial habits manifested in A., among such fossils being the abbreviation of Will, the survival of the form h'll and of a random then spelling. I may also mention a few cases in B. of the very common practice in A. of using a semi-colon for a question mark, the setting in A. of a period in over a dozen cases where a necessary question mark was used in B., and a very odd but marked use of the full -ed ending in A. for the elided form apostrophe d in B., an expansion difficult to account for with such frequency in a reprint. Then there were such cases as this. The colloquial form u'm for them is absolutely consistent throughout both editions except for one single late use of them in A. where u'm appears in B. The probability seemed much higher that B. had normalized A. than that A. had wilfully departed from its copy B. There were also two cases where proper names were found in italic in B., as customary, but in roman in A., a reversal hard to account for. Finally, the evidence of the catchwords possibly, but almost certainly that of the mispagination, seemed to point to A. as the original.

I hope I have made this case reasonably convincing, even in an incomplete and digested form, because I am ashamed to confess that initially—before I began to prepare the text for an edition— it had me thoroughly convinced. Yet I hasten to point out that all these conclusions are quite wrong, that Griffith and Osborn are right. Edition B. indeed precedes A.; and this case I have presented for Macdonald's order, though based mostly on 'bibliographical' facts, has been one only of inference and probability but never of true bibliographical demonstration.

This, briefly, is the real evidence for B. as the first edition which my subsequent investigation revealed. Only one press-variant forme emerged from collation of eleven copies of A., but from ten copies of B. I finally turned up seven formes which had

undergone stop-press correction. Of these, two are indifferent since the uncorrected state agrees with A. for one, and the other is the variant act-heading noticed by Professor Griffith in the apparently unique Texas copy. Of the remaining five, all—in their significant B. readings—have an agreement between the corrected state of the formes and A. There is no need to elaborate the bibliographical argument here.[26] Almost inevitably, A. must have been set from B.

Finally, to clinch the case, this evidence developed. On one occasion in A. the verb tells in the phrase she tells me is misprinted as tel's. We can demonstrate the source of this misprint in A. when we find that in the corresponding place in B. the second letter l in tells inks so very slightly in all observed copies that only its tip is visible without a magnifying glass; and since this inked tip would almost inevitably be taken as an apostrophe, the compositor of A. read it as one and faithfully followed what he thought was the spelling of his copy. Correspondingly, in three places in A. where rather essential punctuation is missing, we find that the actual commas are so lightly inked in some copies of B. as scarcely to be seen. The evidence of the tel's alone is sufficient on which to rest one's whole case for the order of the editions; but especially when so powerfully confirmed by the evidence of the press-variant formes in B. followed by A., one could take the dispute to a court of law and secure a judgment for the priority of B.

From this detailed example I draw a moral which applies to much of what I have been trying to say in this paper. Textual and pseudo-bibliographical evidence can seldom if ever afford more than a high degree of probability, and this is essentially different from positive demonstration. A very plausible chain of inferences can be built up, if the person is as stupid as I seem to have been; and if only the same line of evidence is employed to attack such a case, nothing but an indecisive stalemate can result. On the contrary, strictly bibliographical evidence such as that about the tel's and the press variants—evidence which usually appears in

texts if one digs deep enough—crosses the line of probability into something close to the field which in science would be regarded as controlled experiment capable of being reproduced. Instead of high probability we have, in fact, practical demonstration on physical evidence of a mechanical nature, demonstrable by a mechanical process, evidence like the prose and verse measures of King Lear, and this is what Greg means by bibliography's treatment of a book as a material object. To every point in my construction of the case for 72a as the first edition, alternatives could have been suggested, and therefore any credibility it possessed depended on the cumulative bulk of the inferences. Yet when actual bibliographical evidence enters the scene, one little tel's sweeps all before it.

I draw the further moral as this applies to editing. Such minute yet crucial evidence on which the case hinges would normally be discoverable only by an editor. For example, the ordinary tests which a descriptive bibliographer would apply could not determine the truth; hence, if a textual critic, faced with the problem of corruption versus correction, felt that the standard bibliography of Dryden had sufficiently set up the material for him to proceed with 72a as his copy-text (and he could very well do so), he would be most seriously misled if he failed to come upon the bibliographical points which destroy Macdonald's case.

In this paper I seem to have given a considerable number of examples where bibliographers were wrong. I have not intended to cast doubts on the validity of bibliographical findings, or to speak in a way to discourage the faint-hearted from depending upon bibliographical evidence. Quite the reverse. What has been paramount in my mind is, first, that bibliography may sometimes be imperfectly practised; and unless a textual critic is himself enough of a bibliographer to make his own discoveries independently, or else to submit existing bibliographical conclusions about his material to the acid test of his own bibliographical re-evaluation—unless he can do this, in many cases he will be living in a fool's paradise, either believing that there are no bibliographical problems, or else that the bibliographers have authoritatively worked over the material for him and he can go ahead on his own line without further consideration.

Secondly, this principle seeps down from the very top, or the

choice and treatment of the copy-text, to the very bottom of the last, minute detail of an old-spelling critical edition which must be exhausted bibliographically before pure criticism can properly operate.

Thus I think we may freely say that the bibliographer's text is by no means always the best that can be contrived if the editor has not got the mind of a true textual critic—only perhaps that such a text does the minimum of harm. On the other hand, as a practising bibliographer I do object mightily to critics who often nullify the brilliance of their substantive text by failing to observe bibliographical principles when they engage themselves to an old-spelling edition. I should be inclined to set up four operations which an old-spelling editor should perform, for the interpretation and application of the evidence discovered by these investigations lie at the heart of a sound edition. It is astonishing to a bibliographer to find how often various of these have been omitted by an editor who otherwise has exhibited every desire to be scrupulous in the details of his work. First, the determination of authority in all early editions and, on this evidence, the proper choice and treatment of the copy-text as we have heard it described by Greg. Second, the collation of multiple copies of each authoritative edition to disclose proof-corrections for analysis. Third, the analysis of running-titles for the interpretation of the presswork, a matter which is closely linked with the fourth, or compositor analysis, this last being positively essential, at least in Elizabethan texts, for any consideration of the variants between two or more authoritative editions in a direct line of derivation.

These few demands are not especially severe and need not require any extraordinary technical training. But I place them as the basis for any bibliographical preparation of an individual text by an editor once the larger questions have been settled. And until they become standard procedure, at least in cases of any complexity where their chief value is found, our old-spelling texts are not going to be definitive in any real sense. Critical brilliance can settle many a substantive crux, although by no means all of them; but if these excellent major substantives are placed in a semi-substantive old-spelling background which is not itself accurate, then the reason for old-spelling texts degenerates into sentimentality or ostentation. Moreover, by an accurate background

of accidentals and minor substantives, much more should be implied in many cases than the mechanical ability to copy and then proofread accurately from an old edition. If we are to learn to reconstruct the accidentals and minor substantives of an author's manuscript with as much care as we labor over reconstructing his correct major substantives, a whole new and delicate biblio-critical art is involved in which criticism by itself can never be a sufficiently scientific instrument.

I shall try to summarize the main points at which I have glanced. First, both textual criticism and analytical bibliography in their purest states are, in my opinion, independent arts which have no necessary relation in their disciplines, or frequently in the subjects with which they concern themselves. Textual criticism must deal with words and their meanings, with stylistic and linguistic considerations, and with the basic questions of authority in texts, both in whole and in part. Pure analytical bibliography, on the other hand, deals with books as material objects formed by the mechanical process of printing. In many investigations it is not concerned with texts as such; but when it does approach texts it endeavors to treat them not from the literary or critical point of view, which is that of the 'goodness' of readings, but instead as pieces of paper mechanically impressed with certain symbols. The mechanical relation between these sets of symbols is thereupon its chief concern.

Second, although the two methods are essentially independent, textual criticism cannot controvert accurate bibliography in its findings when the subject is one on which bibliography can properly operate. However, the two often join in attacking certain problems of texts. These problems are by no means limited to editing, but nevertheless the construction of a critical text is most commonly the point at which the two methods cross.

Third, even though the first and the last approach to a text must be the bibliographical one, neither method can achieve definitive textual results in any detailed manner if utilized without reference to the other. On a broad scale it is bibliography which establishes the physical facts of the derivation of texts from one another and which wrings from a given book every last drop of information about the mechanical process of its printing that may be of service in determining the relation of the printed result to

the manuscript or printed copy used by the printer. Still on a broad scale, it is the function of textual criticism to evaluate the authority of this manuscript and then to proceed to the correctness and the authority of the words in the text in the light of the ascertained physical facts which bibliography has furnished. Thereupon, in the process of fixing the text of a modern critical edition in every possible detail, the two methods are often so closely conjoined that an attempt to separate their respective functions would be futile.

When we come to criticize the average edited text produced today, we find that the usual source of error lies in the insufficient bibliographical training of the editor. Either he has little concept of textual problems, and—scornful of the minutiae which sometimes concern the bibliographer—is content to have any kind of a text, because what he really wants is only a peg on which to hang his annotations; or else he is overconfident in his purely critical abilities to solve any problems which may arise. In either case he has ventured on a delicate literary task without knowing his grammar, and hence his results can seldom be definitive. Sometimes even greater harm is caused by a little knowledge than by none. It is enough to make the angels weep to find Saintsbury, for example, throwing out various correct readings in Scott's text of The Wild Gallant, derived from edition 72b through the Folio and a later edition, to substitute absolute corruptions from 72a under the illusion he was restoring the purity of the text from the first edition. Or to find Davenport in the new edition of Joseph Hall's Poems declining the editorial responsibility to make a choice of readings, and basing his text on the derived reprint of 1598 for the first six books of Virgidemiarum, even though it was slightly revised, rather than the purer 1597 first edition.[27] It is really the hardest job in the world for a bibliographer to convince a critic who is beginning to be conscious of old-spelling problems that an author did not set his own type, seldom proof-read his book and if he did cannot be taken as approving every minute detail of its accidentals, and that a printed book is a fallible second-hand report of the author's manuscript, not a facsimile of it set in type.[28] One of the chief functions of textual bibliography

is to try to pierce this veil of the printing process and to restore, however imperfectly, the authority of the manuscript, which we know only through its printed and thus secondary form.

On the other hand, a bibliographer who tackles the problems of text with insufficient critical and philological training is also in danger of false judgments in decisions which are not material and are therefore not strictly the business of bibliography. The usual result is that he may retreat to the narrowest conservatism to avoid having to face up to problems which are not wholly factual. And this timidity is as unbalanced as a critic's rashness in proceeding without the counterweight of bibliography. Such critical uncertainty may lead to good reprints of a single authority for a text but by no means to a true edition of the best text of an author.

It seems to me that for an ideally definitive work an editor must combine in himself the knowledge of both methods, and the training to put both into practice, or else he must resort to an almost impossible attempt at collaboration. Collaboration can be effective between a bibliographer and a literary critic who will handle all problems of biography, attribution of authorship, literary estimates, and who will write the critical introductions and illustrative notes. In such a case each has his relatively independent responsibilities; and indeed Greg is inclined to recommend this procedure for many editions. But I find it harder to imagine save in the most exceptional circumstances an effective collaboration of bibliographer and critic on the minutiae of the text.

I disclaim bibliography as the usurper of editorial privilege, and indeed I am concerned to aid the textual critic against the increasing pressure which editor-bibliographers are exerting on him. But if this potentially most valuable kind of an editor is to produce texts which in every detail will stand up under the increasingly exact and rigorous standards which are now being applied to this form of scholarship, he must learn his bibliography with a thoroughness not previously thought necessary. Only by this wider extension of scholarship can texts be achieved which will not need to be done all over again by the next generation.

Notes