University of Virginia Library

Greg's "Rationale of Copy-Text" Revisited
by
Fredson Bowers

Perhaps the most influential textual document of this century has been Sir Walter Greg's "The Rationale of Copy-Text," dating from 1949.[1] The general acceptance of this rationale by Anglo-American editors[2] has led to a new school of editing, especially of works in the nineteenth and twentieth centuries under conditions remote from the period 1550-1650 to which he confined himself. The "Rationale" is a theoretical discourse dissociated from any overall survey of typical problems in the editing of a group of authors, or even of any single writer, although it is clear that the Elizabethan drama and its peculiar problems bulked large in Greg's own experience and helped to shape his thinking. The illustrations, drawn from Shakespeare, Jonson, and Marlowe, are more of the nature of case-histories involving the operation of editorial judgment in a few

selected verbal cruxes than a systematic attempt to link example with precept to illuminate the workings of the rationale in broader terms. Some of its tone and structure may be laid to the fact that Greg seems to have thought of it largely as a coda to and correction of one element in a much larger work, The Editorial Problem in Shakespeare (1942; 2nd ed. 1951); moreover, it was written in the form of a paper to be read at a learned meeting and was published verbatim, with the addition of relatively few footnotes. Finally, although Greg's experience related to the general period 1550-1650 in which he set his argument, actually he confined his examples, with one exception, to plays written within a decade of 1600. He made no specific claims for application beyond the period of his illustrations; and indeed he was not equipped by experience to apply his rationale at first hand to the much more varied conditions of the literary periods that followed. In some part, also, it must be acknowledged, Greg had a wide critic's view of Elizabethan editing and problems of textual transmission (in his day confined almost exclusively to the drama) but a narrower practical experience in critical editing. His two-text edition of Marlowe's Doctor Faustus is a diplomatic reprint, not a critical edition; and his critical text of the play is a modernized one. Thus his actual experience in the nuts and bolts of the editorial process in the period, outside of the Faustus and a vast experience with the Malone Society typographical facsimiles, was confined to his edition of the manuscript texts of Jonson's masque The Gipsies Metamorphosed. However, one must not ignore a very important influence on his experience, that of dealing with the textual problems of the manuscripts in the medieval miracle play cycles (1914) and the theoretical Calculus of Variants (1925) that in some part may have grown from this experience.

Greg emphasized, with his usual generosity, that in the first of its two parts—the actual choice of copy-text—the rationale was chiefly a restatement of McKerrow's position in Prolegomena for the Oxford Shakespeare (1939), and that his own original contribution was confined to the second part—the eclectic treatment of 'substantives' and 'accidentals' with their separate authority in collateral-text or in revised-edition situations. The importance of this second part as supplementing the first can hardly be overestimated, however, for it was the capstone of the total theory and (except for Housman in other circumstances) the most convincing expression written of the principle of eclectic texts. Combined with the reasoned emphasis he placed on the choice of copy-text for the authority of its accidentals instead of its substantives, it is this freeing of an editor from 'the tyranny of the copy-text' in his choice of substantives when more than one authority is present that has exercised

ultimately the paramount influence. Indeed, it may be said that Greg seems to be more concerned with this matter of editorial freedom of judgment than he is with the narrower matter of copy-text choice, which he seems to have regarded as so well established as to require relatively less illustration and argumentation. On the contrary, in modern editing the choice of copy-text has bulked as large as questions of its treatment. In one or other respect, however, Greg's influence has been so marked that in the words of one scholar, "[Greg's] views have never been publicly refuted."[3]

In my opinion these views are indeed irrefutable but only when applied to the specific conditions in which Greg placed them and subjected to the implicit limitations of these rather special conditions. Yet unquestionably they have in theory a wider and equally legitimate scope. Greg himself visualized only a Renaissance textual situation; nevertheless, it is agreed that most of his conclusions are as pertinent to texts of any later period as they are to the age of Shakespeare or Jonson whenever the transmissional conditions are the same.[4] Consequently, for many texts it has been possible to edit Dryden, Fielding, Hawthorne

and a number of nineteenth-century authors according to the same rationale as Shakespeare. This fact is now commonly accepted, but the very success of the new editorial theory has brought dangers in its train.[5] In the hands of scholars knowledgeable in the conditions of the nineteenth century but with no personal acquaintance with the Elizabethan textual situations that are familiar to Renaissance students, the rationale has had a tendency to harden into a mold without consideration of the inherent flexibility Greg took for granted within the framework to which he was accustomed. Moreover, the extrapolation of certain of Greg's statements to conditions other than those he envisaged has had a tendency to be dogmatic in a manner alien to Greg's generally permissive attitude for the treatment of exceptional situations. It is no secret that converts are likely to be more rigid than those born to the faith.

Attempts to explain the purposes of Greg's rationale and its application—as well as some unusual circumstances in which it does not operate—have appeared from time to time since 1949,[6] but it may still be worth while to attempt a more precise analysis than has been available of the Renaissance textual conditions with which Greg was familiar, the intent being to offer background discussion to assist scholars of other periods when they apply Greg's tenets to later literature and different situations. In this process some analysis may be useful of a few ambiguities and half-lights in Greg's own presentation that could interfere with an understanding of his basic thought. Finally, a brief survey may be introduced of a few of the more common present-day textual problems

that offer difficulty. By these means perhaps a more balanced view may be obtained to guide future discussions pro and con.

The fairest presentation of Greg's theoretical position, together with a searching analysis of the development and linking of its several lines of argument, is that recently made by Dr. G. Thomas Tanselle (see, above, footnote 4), an essay that removes any necessity for me to repeat the chief items of the rationale. We can, then, begin in medias res. As good a point as any for an opening of the matter comes in Greg's discussion of the reasons why 'accidentals' have a separate authority in a text, sometimes independent of the 'substantives,' and the consequences that follow in the necessity to choose as copy-text that early document containing the most authoritative accidentals without major regard for the status of its substantives. Throughout Greg's discussion he never mentions the relative authority of a holograph manuscript as against that either of a scribal manuscript or of the printed book that ultimately derived from the holograph with or without the author's supervision. Since Elizabethan authorial manuscripts of works that saw print can be counted just about on the fingers of one hand, it seems evident that Greg saw no point in considering a contingency that was unlikely to arise, particularly when a simple extension of his principles backward in the transmission would easily cover the case. Thus Greg consistently writes in terms of the preservation of accidental authority only in derived documents which by their very nature must imperfectly reproduce the author's own system of spelling, punctuation, word-division, etc., if system there was.

These particular conditions help to explain certain of his attitudes. First, the overlay of the author's accidentals by the transmissional process is more extreme in the Renaissance than in later periods, thus diminishing the authority in this respect of any derived document. Since in its spelling, standards of capitalization, division, and in the somewhat impressionistic rhetorical instead of syntactical system of punctuation[7] the language was in a state of flux, the widest variety of accidentals prevailed within a general framework of acceptability. Johnson's Dictionary imposing standards of spelling was still well in the future. The

manuscripts of the period, except for careful scribal copies for presentation, may have practically no fixed standards at all, and can be especially deficient in punctuation, some almost to the point of unintelligibility. Even the prompt-books for plays are often so erratically punctuated as to leave almost complete freedom to any compositor if he had set copy from them, let alone the actor who recited from the parts transcribed from these books. For many words neither authors nor compositors had any fixed system of spelling but only a general preference for one form or another, if that. (Shakespeare—if it was he who wrote Hand D in the Sir Thomas More manuscript—in a brief scene spells sheriff in five different ways.) The accidental characteristics of compositors were likely to be more advanced than those of authors; it follows that each resetting of type, whether in transferring a manuscript to print or in reprinting the text of an earlier edition, was a modernization.[8] Systematic proof-correction by authors was a rarity. Among themselves compositors differed in their spelling preferences so that a book might exhibit several different sets of characteristics. Whatever effort a compositor or scribe might make to reproduce the wording of his copy, no suggestion is present in the period that any importance was attached to following the author's ordinary run of accidentals.

Under these conditions Greg knew that (in a manner impossible in the nineteenth century)[9] the accidents of a printed book would differ

considerably from those of the holograph, in contrast to the substantives that a compositor might be supposed to wish to reproduce with reasonable fidelity. This fact is at the root of his division between the authority of substantives and of accidentals. The prospects that any printed book (or scribal manuscript) would reflect to any significant degree the author's accidentals are uncertain and variable, in most cases difficult to recover, and if fortuitously recovered difficult to identify.[10] Hence Greg makes no claims for specific recognizable authority in the copy-text accidentals and regards their reproduction as more significant of the time (and locality of a scribal transcript) than of the author, at least in any identifiable form. His discussion (p. 376) of the possibilities in the distant future of establishing a "standard spelling for a particular period or district or author" is in part a red herring, applying more to manuscript than to printed-book texts, and it represents chiefly a cautious attempt to protect his flanks (that it is necessary, instead, "to choose whatever extant text may be supposed to represent most nearly what the author wrote and to follow it with the least possible alteration") from the experiments for synthetic accidentals that were starting in his day among editors of medieval texts. It can be ignored with safety. Why he limited his succeeding statement to manuscripts and did not include printed books in writing "Thus a contemporary manuscript will at least preserve the spelling of the period, and may even retain some of the author's own" is hard to understand except for the speculation that Greg himself felt somewhat more secure about the facts of manuscript transmission (which he had studied) than about the specific details of compositorial transmission with which he had only a general acquaintance. (The considerable scholarship devoted to the spelling characteristics of Elizabethan compositors, principally in the First Folio, has developed mostly since Greg's day.)

At any rate, it is possible to clarify for the modern scholar Greg's conclusion that "it is only on the grounds of expediency, and in consequence either of philological ignorance or of linguistic circumstances, that we select a particular original as our copy-text." This applies almost exclusively to multiple manuscript texts of the same work and in essence is of no concern whatever to a nineteenth-century editor. In fact, Greg introduces this statement mainly as an argumentative ploy to support the conclusion that the editorial treatment of scribal manuscript accidentals must be conservative, through ignorance, whereas that of substantives can be subject to editorial judgment. It has little or nothing to do with the selection of copy-text for works preserved in print. If one is not to modernize an early text (and Greg rejects this proposition for scholarly editions), then the accidentals of some one document must be chosen as a framework for the substantives. Since in such manuscripts as he has in mind we are dealing not with holographs but with documents that may derive from holograph at some distance and will preserve the author's accidentals only in part (if at all) and in details unknown to us, we should select that (manuscript) document as copy-text which best preserves the accidentals of the author's period and district (when possible). It is granted that because of the derived state of the manuscripts we do not possess the philological expertise to restore the exact accidentals of the original period of its holograph. Thus we bow to expediency in selecting the manuscript that seems best to reflect the original conditions, including the linguistic circumstances of district. However the choice of manuscript copy-text is decided, it is based on expediency, in the sense that whatever selection is made from among various documents will at best be a compromise that only imperfectly (if at all) can be supposed to reproduce the holograph's exact characteristics; nevertheless, once chosen, this document's accidentals are to be transcribed and retained with a minimum of alteration except as correction is needed.

What to Greg is the more important point in this discussion then follows: although the transmission of documents may vitally affect their accidentals, it affects the substantives in ways that textual and critical analysis can evaluate, whereas little can be done on the evidence to utilize such critical techniques to recover authoritative accidentals. Accordingly, by the full exercise of editorial judgment these original substantives can be reconstructed on the whole (according to the limits of the evidence) in a manner impossible for the accidentals.

Greg thus emerges from his consideration of early manuscripts to the more pertinent examination of the conditions in which book copy-texts should be selected. In doing so, he comes for the first time to examples that may apply to later literatures. The grounds of expediency in the choice of manuscript copy-text with which he introduced his discussion of accidentals have only occasional application to early printed books[11] and none to those of the nineteenth century. The reason is that in his analysis of the situation in respect to manuscripts Greg automatically assumed the presence of a number of collateral texts for the work, this multiplicity of documents not ordinarily resulting from the progress of the work through different stages of authorial revision but instead from the copying on different occasions of a variety of texts deriving from the lost archetype. Except for a few special cases in the hand-printing era, these conditions scarcely exist in later times independent

of revision, which was not primarily in question in the manuscripts Greg was hypothesizing.[12]

When a book text is never revised or in any way altered by comparison with some other authorially-derived document, only one authoritative document can be identified. Since he ignores holographs, Greg places this single authority as the earliest printed text, the one that derives most immediately from the author's lost manuscript and is thus distinguished from any later series of mere reprints[13] stemming in linear fashion from the one primary authoritative preserved document. This is a 'substantive edition': in the case outlined it is the only substantive edition, and obviously it represents the sole possible authority both for accidentals and for substantives (the wording). It is the only form of copy-text that presents to an editor no problems other than correction.

Greg's use of the term substantive edition is in some part misleading. In note 1 on p. 378 he accepts McKerrow's definition as "an edition that is not a reprint of any other" and states "I shall use the term in this sense, since I do not think that there should be any danger of confusion between 'substantive edition' and 'substantive readings'." On p. 379 the next statement is clear: "Whatever may be the relation of a particular substantive edition to the author's manuscript (provided that there is any transcriptional link at all)[14] it stands to reason that the relation of a reprint of that edition must be more remote." This is an extension of the original reference on p. 378, "in the rare cases in which a later edition had been revised by the author or in which there existed more than one 'substantive' edition of comparable authority." The distinction drawn here between two kinds of editions needs explanation in terms of the examples Greg had in mind if it is to have any application to later periods. In the first case an author would annotate a copy of an early edition with his corrections and revisions and send the marked-up copy to the printer. Greg calls this situation a rare one: it is rare in the years close to 1600 simply because of a paucity of authorially revised editions, and the only examples that he mentions are Nashe's Unfortunate Traveller and the Jonson Folio set from marked copies of the quartos[15] and at least in part seen through the press by the author. Technically the Jonson is not a substantive edition according to McKerrow because, in Greg's phrase, it is a revised reprint of the quarto copy.

What Greg had in mind for the second category of two substantive editions of comparable authority is almost completely an Elizabethan phenomenon when applied to two printed editions and not to a printed edition and one or more collateral manuscripts. For the moment to pass over the matter of bad-quarto texts, including for the sake of argument King Lear, Greg's second category consists of the Quarto and Folio editions of Hamlet, Othello, Troilus and Cressida, and 2 Henry IV. In their desire to print the Folio from what they considered to be the best texts, and perhaps in one case because of copyright problems, the Folio editors did not merely reprint these plays from the available good quarto texts without annotation of their own as in Romeo and Juliet or with minor alteration deriving from the prompt-book as in The Merchant of Venice, Richard II, or I Henry IV. Instead, the Folio version of the text derives its substantives in major part from the authority of a manuscript in a different tradition from that which was the setting copy for the Quarto. However, this manuscript was not itself set by the Folio printer but for convenience he was given a copy of the Quarto that had been annotated to bring the original printed version into general conformity with the variant readings of the manuscript. Technically, therefore, this expedient by the Folio editors disqualified the Folio texts of these plays for consideration as 'substantive editions' under McKerrow's definition since the Folio is in considerable part a reprint of the Quarto except for the marked alterations. On the other hand, if one thinks exclusively of the literary status of the Folio text, fresh authority has certainly entered and both Quarto and Folio could be regarded critically (not bibliographically) as 'substantive editions' even though one text was not revised from another by the authorial post-publication process that produced, say, the revised reprint of the Folio Sejanus from Jonson's Quarto. Instead, the divergences in the versions appear to stem from pre-publication differences between the two manuscripts behind the separate editions, which varied from each other in the state of their texts. Since the manuscripts were either written by Shakespeare or else were scribal copies of a holograph even at some remove—and in at least some of the plays appear to represent different stages in the development or perhaps in the theatrical application of the text—each is therefore 'authoritative' and the printed results represent (again only in a literary and not in a technical sense) two substantive editions of comparable authority.

The phrase comparable authority has little relation in Greg's mind to the critical question of the literary merit of the texts. Instead, authority is a technical term in textual criticism signifying the derivation of any form from the author, preferably by some transcriptional link.

Authority may be inferior as when some major contamination in the transmission has affected the validity of one text, or when the distant relation of an edition to the holograph suggests the possibility of more than usual transmissional corruption; or it can be very strong, as when a printed edition is set directly from holograph. Hence Greg uses comparable authority to indicate that, so far as the evidence goes, in the main the author had a relatively equal concern in the composition of both versions and that each is close enough to the author so as not to have been distorted excessively by transmissional problems. Authorial concern need not have been revisory, although revision is not by definition excluded.[16] This is what he had in mind when on p. 381 he wrote: "McKerrow was, therefore, in his later work quite conscious of the distinction between substantive readings and accidentals, in so far as the problem of revision is concerned. But he never applied the conception to cases in which we have more than one substantive text, as in Hamlet and perhaps in 2 Henry IV, Troilus and Cressida, and Othello." Greg uses the term substantive edition, or its equivalent, again on pages 384-385, "But whenever there is more than one substantive text of comparable authority [footnote excluding 'bad quartos'], then although it will still be necessary to choose one of them as copy-text, and to follow it in accidentals, this copy-text can be allowed no over-riding or even preponderant authority so far as substantive readings are concerned." Finally, on page 391, "Of both [Richard III and King Lear] much the best text is supplied by the folio of 1623; but this is not a substantive text, but one set from a copy of an earlier quarto that had been extensively corrected by collation with a manuscript preserved in the playhouse."

A basic confusion is created by the use in these juxtaposed passages of the term substantive edition (or text) in two contradictory senses. In his statement on Richard III and King Lear he is employing the strict McKerrow definition that he had promised to observe. This holds for his statement about Hamlet, which Greg regarded as set in Q and in F from two independent manuscripts, a view no longer popular. On the other hand, in his remarks about 2 Henry IV, Troilus, and Othello he forgets that he is supposed to be observing McKerrow and instead reverts to an important modification of the term that he had made in the Editorial Problem whereby what he called 'mixed texts' joined McKerrow's editions (not derived from any other) as a separate category of an expanded concept of what constitutes substantive texts. Editions were mixed if one had been used as basic setting copy for a later but in the process had (a) been conflated by some agent with a more or less authoritative manuscript in a different tradition from that used as setting copy for the earlier, or (b) been revised by the author making his own annotations as Jonson had done in preparing copy for the 1616 Folio by marking up his quarto plays. Independently transmitted texts were automatically substantive in the McKerrow manner. Mixed texts that had undergone scrupulous authorial supervision (preferably of the printing as well as the preparation of copy), like Jonson's, were also admitted without other qualification as substantive. However, for mixed texts resulting in the Shakespeare Folio from conflation of the quarto with some other manuscript, Greg believed that the criterion of an alteration in their 'essential character' had to be applied before the Folio could be allowed as a substantive edition. An example of this reasoning comes in footnote 1 on page xix of The Editorial Problem in Shakespeare:

In mid-1949 when the "Rationale" was written it is uncertain whether Greg had been informed privately by Dr. Alice Walker of her researches indicating that F Othello had been set from an annotated copy of Q, although his association of Othello on page 381 not with Hamlet but with the two plays he was convinced were mixed by conflation would

seem to suggest the possibility, or at the least his concern that other discussions to the same effect of Othello had made its mixed more probable than its independent nature. But at least in respect to Troilus and to 2 Henry IV there can be no question that the Folio could not conform to McKerrow's definition of a substantive edition. Hence the passage in the "Rationale" specifying the four plays as substantive represents a lapse in favor of his earlier (1942) inclusion of mixed editions and contradicts his express statement that he will use substantive edition only in McKerrow's sense of 'completely independent.' Thus what Greg is actually saying on page 381 is this: Hamlet is substantive in both editions because each was set directly from manuscript; on the other hand, the three other plays, being mixed texts, come under another criterion for substantiveness—that of a change in F from the essential character of the text in Q—and he is uncertain whether in fact Q and F have been sufficiently modified in their essential character to warrant substantive status.[17] In this paragraph, then, he is definitely not following his definition of substantive as on page 391,[18] and the reader needs a warning to the effect. Moreover, the straying here brings in question what he intended in the other references to substantive texts or editions (there can be no distinction in kind although one is possible in form), for if these four plays (including Hamlet as we now believe) are not substantive in the Folio because set from marked-up quartos, then Greg is left with no illustration of what he had in mind (in the McKerrow sense) in writing of such dual substantive texts. This is anomalous, for the exact circumstances that led him to deny substantive status to the Folio Richard III operate for Othello and the rest.

It is an article of faith that in this period a scribe bringing printed copy into general conformity with a manuscript in lieu of giving the printer the manuscript as setting copy[19] would confine himself largely to the transfer of variant words and would ordinarily ignore accidentals that were not associated with the variant substantives; hence the accidentals of the Folio, it is believed, are in the main Jaggard's compositors' versions of the Quarto's and of no authority since they would lack any transcriptional link with a manuscript that in turn ultimately derived from Shakespeare. Under these conditions, and editor is forced to choose the Quarto versions as copy-texts, regardless of any other considerations (except for 'bad quartos') because they alone have transcriptional links that ultimately tie them with the holographs.[20] The sole editorial question then concerns the authority of the variant substantive readings taken one by one, in which matter Greg is on relatively solid ground even though he is silent on the intricacies of the textual problems that might influence editorial judgment.

Since a textual critic is now left with no examples in the "Rationale" of early printed books with variant texts (not authorially revised editions) set independently from different manuscripts, he will need to supply these for himself. Within this period the ones I am familiar with require a choice between the authority of a printed edition and a manuscript version, scribal as in Beaumont and Fletcher's Beggars Bush, Woman's Prize, and Bonduca, or authorial as in George Herbert's Temple; or else a printed edition and multiple scribal manuscript texts as in the poems of John Donne. An editor can approach the Beaumont and Fletcher plays as if the manuscript were actually a printed edition (or the 1647 Folio a manuscript) since only the two documents are preserved in their different traditions and the choice of copy-text resolves into the usual question of transmissional closeness to holograph versus the problem of the early or late date for the reproduction of the accidentals.[21]

When various manuscript texts of different degrees of completeness are preserved of poems, as with Donne, usually no way exists to draw a family tree for each short lyric, although something can be done in rougher outline for some of the collections as a whole. An editor is forced to consider the printed edition merely as the equivalent of the manuscript that was its copy, substantially subject to the same rules of a reconstructed text that characterize editions of manuscripts. In order to secure general uniformity of accidental texture most editors prefer to select the printed edition as copy-text on just such grounds of expediency as Greg remarked, although some circumstances might suggest a bolder course even when considerable disparity between the accidental systems of different poems would result. However, many manuscript versions of poems would be troublesome to read without as much editorial tidying up as the original printer gave his copy; and so unless one is dealing with holograph, editorial expediency is not necessarily blameworthy since authentic results are so very uncertain when working with derived documents.

A reader unacquainted with Renaissance textual conditions must also be on guard not to draw improper parallels between modern cases and the implications of Greg's use of Hamlet and the other three plays as examples of dual authority. Greg mentions these in context with a consideration of revised editions and the most suitable treatment for them, and hence one may believe that he considers these plays to greater or lesser degree might perhaps come in that category (but see below). He seems most taken with the Folio Hamlet, although no critic that I recall has squarely faced up to the implications of any theory that the alterations that might seem revisory are Shakespeare's or another's. The best case for revision can perhaps be made for Troilus and the least good for 2 Henry IV. At any rate, whether these plays do indeed represent a special class of authorially revised texts is distinctly moot.

Moreover, what Greg did not distinguish were the differences in editorial problems between pre-publication and post-publication authorial revision. Post-publication revision almost automatically requires the hypothesis (usually demonstrable) that an earlier printed copy has been revised by the author. When this occurs, as remarked, no question of copy-text arises for early books because of the assumption that the usual revising author, like Nashe, would pay small attention to changing the accidentals of the printed copy that were acceptable by current standards. After all, he knew that the compositor(s) of the revised edition were going to alter what they set according to their own wishes and that this result would also be acceptable by current standards. In his own period, except for Ben Jonson, Greg could not have conceived of an author—like Walt Whitman or Henry James (or William)—closely supervising the details of revised editions for accidentals as well as substantives.

It is odd that in the "Rationale" as distinct from the Editorial Problem Greg is curiously uncertain when he treats Jonson, who of all Elizabethan authors has the best claim to the supervision of most parts of a revised edition, as in his Folio Works in 1616. Greg knows where he stands when it comes to the Folio Every Man in his Humour which, though set from an annotated quarto, was so heavily rewritten as to enforce the choice of the Folio as copy-text, it being a labor of Hercules to work up an apparatus for a text that would endeavor to insert the

Folio accidental as well as substantive revisions in the original Quarto texture. What Greg does not mention, and may even not have recognized, was the fact that for choice of copy-text it was not the question whether Jonson's revision of the Folio accidentals in this play made them on the whole more authoritative than those of the Quarto (a proposition that could be readily paralleled in modern times). Instead, he accepts the Folio as copy-text on what turn out to be grounds of expediency: because of the very thorough substantive rewriting, the accidentals that accompany the altered wording must of course be accepted, and these are so numerous as to make it almost impossible to attempt with any consistency to preserve the Quarto accidentals when revision of the substantives would not necessarily have affected their variation. In this particular example there is nothing wrong with expediency (although Greg does not call it that) as a practical matter, but expediency can seldom operate as a principle. Thus when he comes to other of Jonson's revisions his gingerly treatment is likely to mislead a reader who does not know enough about the circumstances to recognize that at bottom Greg is also applying expediency to cases that are not at all parallel and should have been subject to principle.

In dealing with the Jonson Folio as a unit, Greg notes that although it has been disputed whether Jonson revised the proofs, Simpson was most likely correct in assuming that he did so; but there can be no question that Jonson was responsible for the numerous corrections made while the sheets were in process of printing. He continues: "Simpson's consequent decision to take the folio for his copy-text for the plays it contains will doubtless be approved by most critics. I at least have no wish to dispute his choice" (pp. 390-391). He then appends to this last sentence the footnote "Simpson's procedure in taking the 1616 folio as copy-text in the case of most of the masques included, although he admits that in their case Jonson cannot be supposed to have supervised the printing, is much more questionable." This position about the Folio is difficult to equate with Greg's principles for the choice of copy-text as illustrated by Nashe's Unfortunate Traveller—that identified substantive revision should be introduced into the accidental texture of the earliest edition. It is this pronouncement about Nashe that has been taken by modern critics as an absolute, without sufficient understanding of the complex reasons (including expediency) that led Greg to take the opposite position for Jonson, and not necessarily just for Every Man in his Humour.

In the first place, Greg certainly thought of Jonson's care to supervise the printing of his works as an exception to ordinary Elizabethan authorial procedures. For the Folio Greg mentions three stages of supervision

of the revised edition: Jonson's marking of alterations in the copy sent to the printer; Jonson's revision of the Folio proofs; and finally his correction of the sheets while they were in process of printing. The first and third of these are incontrovertible, whereas the second (which could be argued was perhaps the chief reason for Greg's acceptance of Simpson's choice of the Folio) is so conjectural as to be of small value as a basis for editorial principle. A distinction exists between the second and third that is a product of hand printing only. A scrupulous author could indeed require pre-printing proofs, but they were not at all common chiefly because they delayed production of the volume and in some cases even the output in other respects of the printing shop. That is, since type was in short supply and had to be distributed before the next sheet (in general) could be completely typeset, composition of the Folio would have been suspended while Jonson read such special proofs. Under these circumstances printers ordinarily required authors, instead, to attend the press. This term meant that an author would drop by the printing shop each day and would read whatever proof was handy. Sometimes it might be newly-set type-pages if available, but ordinarily the author picked up a sheet being printed, marked any changes he wished, and the press was stopped while the changes were made, to be incorporated in all subsequent sheets printed. Collation of multiple copies, as was done with the Jonson Folio, will reveal the majority of such in-press alterations, which indeed were the usual method of proofreading even by the printer's own reader. If, then, any suggestion can be dropped that Jonson read proofs other than at the printer while the sheets were being machined, an editor is spared having to take account of quite unassessable evidence and can approach the task of comparing the printed Quarto with the Folio in its uncorrected and corrected formes in a rational endeavor to estimate, as an editor must, how much of the accidental variation was due to Jonson and how much to the printer. The choice of copy-text will depend upon his answer.

In the nature of the case we have no means of knowing precisely to what extent Jonson marked the accidentals for change in the original copy for each separate play; but in certain plays like Sejanus Simpson can point with certainty to Jonson's idiosyncratic attempts to impose a classicizing spelling system on Greek and Latin derivatives and to utilize the apostrophus, a metrical punctuation designed to mark the presence of an extra syllable to be lightly sounded in the line (IV, 337-341). He also gives examples, which appear to be buttressed by the evidence of Jonson's holographs as well as his press-corrections, illustrating some of the subtleties of Jonson's punctuation alterations (IV, 342; IX, 73).

The important question then arises of the degree, outside of the preserved evidence of the press-corrections, to which Jonson was able to enforce the compositors to follow copy in the details he had presumably marked for alteration as well as in those he had not. The evidence suggests that he had the usual Elizabethan luck, meaning that the compositors despite what may have been good intentions by no means followed copy in anything approaching the manner one would expect in

modern times. The evidence is of two kinds: first, the extraordinary amount of press-correction required when printing was in progress (presumably to correct departures from copy in at least some part); and second, the large number of idiosyncratic accidentals lost in the transition from Quarto to Folio even though some were added, not only the classicized spellings but particularly the apostrophus, which Simpson was forced to supply from the Quarto in perhaps half of the total occurrences. Especially when it is considered that in Sejanus specifically, and in general in the other plays also, Jonson did nothing in the press-corrections about the Folio departures from his ordinary spellings (and these were fairly numerous), a problem does indeed arise whether the supervision that he gave the play both in the preparation of the marked Quarto copy and in the Folio press-correction is sufficient to elevate the Folio to copy-text status for plays (unlike Every Man in his Humour) where the amount of substantive revision was markedly less and so had little or no effect on the separate problem of the accidentals that were independent of revision. In this connection Greg's footnote about the masques is surprisingly mild. It does not seem 'questionable' to accept the 1616 Folio as copy-text for portions that Jonson admittedly did not supervise in the printing: it seems quite contrary to Greg's whole position since for the masques, at least, the line is a thin one between the case of The Unfortunate Traveller except for the hypothesized marking of the masque copy. That is, if the control that Jonson exercised by press-correction is missing (and this should have no bearing on matters of copy-text), Greg's mere questioning of Simpson's choice of copy-text for the masques implies his at least partial acceptance of the proposition that Jonson's marking of his copy was sufficiently detailed to change the texture of the Folio accidentals from derived to 'substantive' status. Such a view needs more evidence than has been advanced: it is basically most improbable.

Some hint as to Greg's seemingly paradoxical reasoning may be extracted from a statement he made about Sejanus. Although this play comes in a part of the Folio where Jonson did exercise what control he chose over the printing as well as the copy, Greg allows that "In the case of a work like Sejanus, in which correction or revision has been slight [in comparison with Every Man in his Humour], it would obviously be possible to take the quarto as the copy-text and introduce into it whatever authoritative alterations the folio may supply." He then adds, "and indeed, were one editing the play independently, this would be the natural course to pursue." The coda indicates that somehow Greg felt it was suitable to choose the Folio as copy-text for Sejanus in an edition within the Works that (rightly or wrongly) had opted for the Folio as

the more convenient or even authoritative copy-text document as a whole, even though if one were editing the play independent of its place as a Folio play, the Quarto would be the natural copy-text. Since Sejanus is by no means an exception but rather the rule within the Folio, this seems to me to exemplify either a singularly uncharacteristic waffling in the matter of the Herford-Simpson edition or a complete disregard of his own acute and important distinction between editions of particular authorities and editions of works (p. 384). The Herford-Simpson Jonson was ostensibly an edition of the works which by a mistaken choice of copy-text for many parts turned itself into an edition of the Folio. That Greg gives simultaneous approval to two contradictory positions in respect to the same play confuses the more general application of his principles. And his return, after the discussion of Sejanus, to Every Man in his Humour as a revised edition does not assist matters, since this is admittedly an exceptional case, one in which the Folio's authorial accidentals are dependent upon the amount of substantive revision, whereas in Sejanus, the other plays, and the masques, the substantive revision is insufficient to affect the central copy-text question of how much Jonson was able to impose his wishes (assuming that he had such comprehensive wishes, for which the evidence is partial) upon the accidentals of the Folio text, regardless of the substantives.

On page 390, in which the discussion of Jonson is concluded, Greg in effect decides that in cases of revision no hard and fast rule can be laid down as to when an editor should take the original edition as his copy-text and when the revised reprint. This is not very helpful nor is it necessarily true. What is less than satisfactory is that in the end, given the contradictory state of the Jonson discussion, Greg leaves the reader with a generality and no guidelines except for the almost unique case of Every Man in his Humour and the elementary case of The Unfortunate Traveller. The general impression is that he has not himself thought through the situation, with enough concrete examples in mind over a broad enough scope, to afford the reader any material assistance, even in formulating rules of thumb. It cannot be sufficiently emphasized that Every Man in his Humour is not a useful example of revision either in its own day or later,[23] and what holds true for it need not hold true for other revised editions in which the rewriting of the substantives has

not throughout so vitally affected the forms of the accidentals. Most revised editions have not been subject to such thorough rewriting; hence the editorial problem for them is the extent to which within the areas of unchanged substantives the differences in accidentals in a revised edition may be referred to the author. It is this question, and this question only, to which the choice of original or of revised edition as copy-text must be directed.

Greg's "Rationale" presents some difficulties even in dealing with the texts of his own day. The first has little or no relevance to modern textual problems and can be passed over quickly with only a few words of explanation of Greg's position. On pages 384-385 he remarks that

whenever there is more than one substantive text of comparable authority, then although one of them must still be chosen as copy-text, the substantive readings of this chosen edition have no preponderant authority over those of the other substantive text. To this he adds a footnote stating that he inserts the proviso that two substantive texts must be of 'comparable' authority in order to exclude the so-called 'bad quartos' of Shakespeare and other reported texts whose testimony can in general be neglected.[24] He should then have emphasized on page 391 that his examples of Richard III and King Lear are commonly taken to represent 'bad quartos' and hence that 'revision' of these texts in the Folio is on a very different basis from what a modern reader might suppose. As a result, there is no proper analogy between them and examples of post-publication authorial revision as represented by Nashe's Unfortunate Traveller or Jonson's Sejanus, or by the second and third editions of Dryden's Indian Emperour and the fourth edition of Tom Jones.[25] Richard III and almost certainly King Lear, then, should never be cited by modern textual critics as Elizabethan examples of revised texts in the usual sense that an earlier satisfactory state of the work has been worked over by an author in order to incorporate improvements and corrections.

No comparison with a bad quarto can or should be made—in respect to copy-text—with such plays as Hamlet (presumably) but certainly Othello and Troilus which were also set up in the Folio from an annotated quarto that had been conflated with some manuscript. In these plays the original quarto text derived from an authorial manuscript, whether or not at some remove; hence its accidentals have a link with authority as well as its substantives. However, because of the memorial transmission of the text in a bad quarto, the transcriptional link with the author's manuscript is broken and its accidentals can have no possible authoritative origin. Under these circumstances it would be largely pointless to choose a bad quarto as copy-text and insert in its texture the more authoritative substantives of the good text. Naturally, the Folio's accidentals of Richard III have no presumptive authority either, except as the scribe may have been influenced by his manuscript in new text or (perhaps improbably) in collating fairly close passages that required less annotation. But however doubtful the authority may be, it is at least better than that of the Quarto, which in the case of Richard III does not even represent the period of the original printing. It follows that a reader could be misled at first when Greg writes, "So great and so detailed appears to have been the revision that it would be an almost impossible task to distinguish between variation due to the corrector and that due to the compositor, and an editor has no choice but to take the folio as copy-text." This criterion has no applicability whatever (as it has in Every Man in his Humour) and the case is not altogether set to

rights when Greg immediately adds, "Indeed, this would in any case be incumbent upon him for a different reason; for the folio texts are in some parts connected by transcriptional continuity with the author's manuscript, whereas the [bad] quartos contain, it is generally assumed, only reported text whose accidental characteristics can be of no authority whatever" (p. 391).

This whole discussion of Richard III and of King Lear as examples of occasions when "a reprint may in practice be forced upon an editor as copy-text by the nature of the revision itself, quite apart from the question whether or not the author exercised any supervision over its printing" is useless for principle since its practice is so limited and the custom of treating these bad quartos is so established as not to require comment. They have no value in establishing a classification (to set against Every Man in his Humour) of textual situations where the nature of the revision may force upon an editor a reprint as copy-text, "whether or not the author exercised any supervision over its printing" since the choice of copy-text in good texts versus bad memorial versions is established not by the quantity of the revisions but by the nature of the bad original, without authority for its accidentals. The point lies elsewhere.[26] What Greg needs to strengthen his argument are not hypothetical examples to add to Every Man since its editorial problem will so seldom arise; what in fact need illustration are (a) cases of true substantive collateral texts (for which no examples are given) and (b) further analysis of authorially revised editions not conforming to the quantitative criteria necessary to apply to Every Man but instead to the qualitative criteria applicable to Sejanus.

It is particularly unfortunate that the first of these is neglected, for in modern times various combinations of (a) and of (b) may arise to plague an editor. What emerges is that Greg is not very satisfactory in dealing with texts of multiple authority when both are printed editions, and he omits entirely any discussion of the more common case of dual authority when one text is a printed edition and the other a manuscript. One problem that he did not face squarely is that the familiar examples of Elizabethan dual authority in printed editions are mixed, not independent; thus questions of copy-text are not much involved with the examples he gives but instead editorial judgment in the selection of the substantives as between the variants in the two editions.

In fact, Greg does not rule on the copy-text that should be chosen for Hamlet, Troilus, Othello, and 2 Henry IV, and his paragraph about McKerrow's lack of attention to these plays is without point for his argument although he comes to lean on it. In context we may take it that he raised the issue of these plays because in discussing problem texts of the two authorities, and the treatment of readings, McKerrow's Prolegomena had practically confined itself to Richard III. Greg certainly approved of McKerrow's decision to use the Folio as copy-text for Richard III and in general to throw the weight of authority for the substantives on the Folio version as well; but he is interested in what McKerrow's views would have been about the substantives if he had lived to encounter the problems of Hamlet and the rest. Although Greg does not say so, it would seem that he believes McKerrow would have chosen the Quartos as copy-texts and would shortly have found that his rule would not have worked that in essence forbade a choice among the substantives, for he implies that McKerrow would have preferred many Folio substantives for these plays and would have run into trouble when he encountered, also, manifestly superior Quarto readings. Any real application here to copy-text is remote. For instance, on page 387 Greg starts to deal with true revised texts and notes McKerrow's position as earlier described on pages 378-381 but only in respect to The Unfortunate Traveller and according to a principle that McKerrow abandoned; and in contrast he refers the reader to his own emendation of McKerrow's rationale on pages 381-382. However, these pages mention only Hamlet and similar plays and immediately go on to illustrate the tyranny of copy-text in respect to substantives; it follows that his recapitulation on page 384 of his own position is in some part colored by the special cases of texts like Hamlet and the rest where questions of actual revision enter less frequently than the corruption of the lost original copy in the process of multiple transmission. It is once again the lack of distinction on Greg's part between variants that arise in the course of pre-publication transmission and those that result from post-publication transmission that causes difficulty. As a result the one clear statement (apart from the Jonson discussion on pages 389-390) that he makes about the problems of revision, on page 387, is less than comprehensive. He supposes that normal revision for a new edition will be made by an author sending a list of changes to the printer or else a corrected copy of an earlier edition. He then lays down a set of syllogisms to guide an editor in the choice of substantive variation, with one of which—the lesser authority of a revised edition in cases of doubt—I have a serious quarrel. But these syllogisms apply literally only to editions in linear relationship to each other, like the four revised editions of

Joseph Andrews. They need modification (where Greg gives no guidance) for multiple pre-publication authority in textual situations like Hamlet, or Beggars Bush and Womans Prize, or (to jump ahead) Stephen Crane's Red Badge of Courage but more particularly, because of its simultaneous problem of copy-text and substantive authority, Crane's Third Violet, to say nothing of his "Price of the Harness," "Death and the Child," and "The Revenge of the Adolphus."

It is time to recapitulate and to see what emerges for editors of modern documents. Greg's rationale distinguishing the authority of substantives and accidentals is a sound one. However, a modern editor must always be aware that when he follows Greg he is not necessarily dealing with the same conditions and that the frame of reference may therefore alter in subtle ways that end with important differences. For example, when Greg writes "the historical circumstances of the English language make it necessary to adopt in formal matters [accidentals] the guidance of some particular early text" he is considering a problem of no application to late nineteenth- and twentieth-century literature. In Greg's early period in matters of accidentals the language was much more idiosyncratic and in process of flux than is true at the present time when standards of correctness exist that are relatively stable and simple to apply. Given the circumstances of transmission, Greg tacitly has little faith that in most cases more than the general characteristics of the period will be preserved by an insistence on the superior authority of the accidentals in the document closest to the lost manuscript. In multiple manuscripts it is clear that he has no faith at all that authorial characteristics can be recovered. In printed documents he recognizes the chance that something may come through, but it is clear he is not optimistic. The uncertainty about the effect of transmission has its effect on his discussion of the copy-texts suitable for Jonson according to his rationale. In contrast to Nashe, he recognizes that Jonson had some concern for the accidentals of a revised edition in a work like Sejanus and this leads him to sympathize with Simpson's selection of the Folio as copy-text although in the next moment he notes that in many respects the Quarto is the more logical choice. In opting for the Quarto in an independent edition he reflects his general position about Elizabethan accidentals in revised editions; in accepting Simpson's choice of the Folio he indicates his uncertainty as to how much of Jonson's marking of the Quarto as setting copy (and possibly of proof before printing began, as he thinks) might not have come through even though in such an unidentifiable form that it would not be practicable to isolate and transfer editorially from the Folio back into the Quarto copy-text as emendation.

This indecision is symptomatic, it would seem, of a general view that the choice of copy-text on the basis of its accidentals is more a formal than a basically meaningful proposition, a convenience but not a great deal more. No editor of modern literature could feel in this way: in contrast to Elizabethan, the linguistic interest in the spellings of a modern author is minimal; what impels an editor of later works to concern himself with copy-text is the conviction that the accidentals are an inseparable whole with the substantives in transmitting the author's total meaning. That they have a literary interest, not merely a philological, marks a considerable difference in the point of view and very likely in the rationale for the selection of copy-text when revised editions are concerned, as well as the treatment of this copy-text once selected.

Other differences are involved, however. A present-day editor who does not wish to normalize the accidents of his text chooses his copy-text and then generally retains its system not as a convenient guide but in the belief that he is relatively near to the formal features of the lost copy (in case of a first edition as copy-text) or the conviction (if he selects an authorial typescript or manuscript) that he is preserving an authentic example of the writer's own system. Ordinarily he will be right. However, if we inquire in what respects he believes the author's intentions are best preserved (apart from certain habits like word-division), an editor will ordinarily be forced to admit that he is basically referring to the punctuation as a significant guide to, and controller of, shades of meaning. (This is to omit consideration of alterations in syntax that properly come under the heading of substantives or at least semi-substantives.) Some authors may have highly idiosyncratic occasional spellings like Hawthorne's cieling (perhaps best emended, given his period); in other cases the idiosyncrasy is mild and easily preserved, as in the tendency for nineteenth-century American authors to adopt the -our spelling in the English manner for some words but not for others,[27]

or to use s for z as in recognise, or connexion for connection, and so on. Rarely an author comes along like William James with a few pronounced views on spelling, such as his 'reform' tho (not the correct tho') or excentric, and the like. In short, spelling is not ordinarily a part of an author's individual method of expression, whereas in Elizabethan times it is of especial importance to a scholar to attempt to reconstruct because of the possibility that phonetic spellings will guide us to clues about pronunciation or will offer evidence to assist the emendation of compositorial misreadings of peculiar forms.

Modern editing and Greg's rationale join in the wish to preserve as much as possible of the authorial accidentals,[29] but they do so for different purposes (critical versus philological)[30] and with markedly different results, even granting the information from preserved holograph documents about the accidents of a modern author available to the editor. It is a fair interpretation that so long as his major interest in editorial freedom of judgment was satisfied, Greg—knowing the uncertain basis of transmission—was prepared in general to accept as copy-text any document with reasonable claims to authority. He is firm, of course, that in a series of linear reprints without revision only the first edition will do, and he is firm that in cases where the revising author may be supposed to have paid little or no attention to the accidentals the first edition is also the proper copy-text. But for a situation like Sejanus in which

there is evidence for the author's attention to the formal features in marking up the revised printer's copy, he is so much the reverse of doctrinaire as to be almost indifferent whether an editor chooses the Quarto or the Folio. This being so, modern critics are mistaken who apply Greg's rationale narrowly to revised texts and insist that the rationale requires the selection of the first edition (or of a manuscript or typescript) as copy-text, followed by the insertion of revised substantives into this earliest accidental texture. Greg does not lay down the law and is much more permissive than is generally supposed by those who take his remarks about The Unfortunate Traveller as the sole guide. More typical is: "The fact is that cases of revision differ so greatly in circumstances and character that it seems impossible to lay down any hard and fast rule as to when an editor should take the original edition as his copy-text and when the revised reprint. All that can be said is that if the original be selected, then the author's corrections must be incorporated; and that if the reprint is selected, then the original reading must be restored when that of the reprint is due to unauthorized variation. Thus the editor cannot escape the responsibility of distinguishing to the best of his ability between the two categories. No juggling with copy-text will relieve him of the duty and necessity of exercising his own judgement" (p. 390).[31] Modern editors must understand, however, that in Greg's view this emancipation applies only to the treatment of the substantives and that the copy-text accidentals are to be followed conservatively, as a convenience, except in the case of error.

Once again it must be emphasized that to Greg Elizabethan accidentals are not really a part of the total meaning of a work in the modern manner of thought. This important distinction between the concepts has at least two consequences of importance for the present day. First, since the general texture of the accidentals at the present time is much more uniform than in the period with which Greg was concerned, the choice of copy-text transcends the grounds of expediency and must

be recognized as having a critical significance beyond that which Greg conceived. Second, it follows that given the generally greater information about an author's accidentals possessed by an editor of modern documents, and often the wider range of preserved transmitted documents starting with drafts and ranging through typescripts and sometimes even proofs, an editor may be able to exercise his judgment in the choice of accidentals between competing authorities in a manner impossible for Greg to envisage as applying in the early period. It is true that the kind of accidental material on which an editor's critical judgment can operate is less susceptible of demonstration from the test of meaning than that applied critically to substantives. Nevertheless, although certain areas of accidentals may seem generally to be so indifferent in their variation as not to be subject to reasoned choice, a knowledgeable editor can deal with other areas on the same basis that Greg urges for the substantives. This new condition presents fresh problems that have not been much explored and where Greg can offer little guidance.

Generalizations about the fidelity of the printer to his copy are dangerous in any age; but it is at least allowable that from the late seventeenth century when more uniformity in spelling and in standards of punctuation began to be imposed on compositors, the uncertainty that attaches to Elizabethan conditions of transmission begins to clear. Of course, all conclusions based on compositorial fidelity to setting from printed copy (the usual evidence) are not necessarily applicable to setting from manuscript: in any age printed copy, having been styled already, will be followed more faithfully by a compositor than he is likely to reproduce a manuscript which (until comparatively recent times) it was his duty to style on the author's behalf.[32] Nevertheless, the practical

hopelessness an Elizabethan scholar feels about the problem of identifying authorial accidental characteristics imbedded in the variable systems of different compositors can begin to yield to modified optimism by the twentieth century, it would seem, and sometimes considerably earlier according to special conditions. With the change come certain modifications that may need to be applied to the popular interpretation of Greg's rationale.

Obviously, when no anterior documents are preserved, the first edition is the substantive text and in lieu of further revision must become the copy-text instead of any later reprint. When revised editions are present, the modern editor has the same problem that exists in the Elizabethan choice of editorial method to deal with revision as between Thomas Nashe and Ben Jonson; that is, he must attempt to find evidence that will permit him to evaluate whether the author confined himself chiefly to the revision of substantives or else included accidentals as well and on more than a casual basis. If the evidence suggests that substantives were the author's major concern, then it is probable that the edition nearest to the lost holograph—that is, the substantive first edition—should be retained as the copy-text. In this event the accidental variants can be winnowed for possible authorial changes as against publisher's and printer's, these assumed authorial markings to be inserted in the copy-text on the same basis as the revised substantives. The process is materially aided in the twentieth century by the fact that generalizations about transmissional variation appropriate for the age of Shakespeare are of lesser validity now.

When an editor decides that an author revising an edition has concerned himself in a meaningful way with marking accidentals for alteration, he has Greg's blessing for choosing the revised edition as copy-text. This choice is not one to be made lightly, however, nor without concrete documentary evidence relating either to the work itself or to other works with which the text under consideration shows common characteristics presumably arising from similar causes. For instance, a few examples of the marked periodical copy are preserved that William James used for his book collections, enough to demonstrate that he was every bit as concerned to modify his accidentals as his substantives before typesetting began. However, when the book shows further accidental as well as substantive changes from this revised copy, an editor must decide whether the accidental changes are again of a piece with the substantives or else may be mainly attributed to the printer. Comma for comma no editor can demonstrate that every individual difference is authorial; but enough general Jamesian idiosyncrasies come through to indicate that James was in the habit of heavily revising his book proofs even after he had carefully marked the printer's copy, and that these proof changes concerned the accidentals as well as the substantives.[33] As a result of a full analysis of James's revisory methods, in the James edition the volumes in which the periodical articles constitute the original texts and a book the revised reprint of these articles offer a critical text based on the final revision as copy-text.[34] Evidence for any one of these works is buttressed by identical evidence for the others. In its nature, for example, the evidence revealed by the collation of journals and book is practically interchangeable between Pragmatism and The Meaning

of Truth, indication of James's customary scrupulous revision of book copy and then of the proof of both volumes, even though no corrected book proof-sheets have been preserved as documentary evidence.[35]

It can be said, then, that the important evidence should be sought in two parts before a revised edition can be made the copy-text with any confidence: first, firm evidence that the author revised the formal features of his text as well as the substantives whether in the press copy or in proofs or in both;[36] and, second, some evidence when obtainable about the degree of fidelity given by the printer to the press copy, this as an assistance in estimating the influence of the author on the altered accidentals.

The pertinent passage in Dr. Robson's essay (Editing Nineteenth Century Texts, pp. 116-117) runs as follows:

My unease is really with the argument that Mill's 'approval' constitutes such overwhelming authority in a revised-edition copy-text that an editor may decline to emend by reference to earlier authority such accidentals as he felt were demonstrably not Mill's. This is a far cry from a computer concordance that might or might not serve as the evidential basis for an overall attempt at deciding each and every detail between the printer and Mill. I seriously doubt that for punctuation such an 'ideal Platonic text' is possible for any author. I am suggesting, chiefly, that half a loaf is better than none if one takes accidental authority as seriously as substantive authority.[37] If an editor chooses to apply to modern authors the criterion of linguistic expediency Greg felt forced to defend, then our positions are so far apart that argument can scarcely be heard across the gap. It does seem to me, however, that if in Mill's case the accidentals are 'relatively unimportant,' the assertion should be illustrated and the text modernized. Certainly if there is to be one rule for creative writers and another for economists and philosophers and historians, perhaps the subject is worth some interested discussion to clarify the respective positions. In my own experience, I should say that the accidentals are more important in the text of the philosopher William James than in the several prose fiction writers I have edited; that is, if the attention James paid to them in print in contrast to the others is a criterion.

If any discussion about the importance of accidental fidelity in editions of writers of different kinds were to develop, one of the first problems would be to clarify the role of 'authorial approval.' Strictly speaking, all a proponent can assert is that the writer has tacitly 'approved' the accidentals of

the press copy by declining to alter them in the preparation of a revised edition.[38] This is a severely limited but logically defensible position although painful to anyone who would like to see identifiable authority in accidentals restored. However, the printed result may differ from the setting copy. In some cases variation may be so indifferent that no decision is possible whether the writer or the printer was on the whole responsible (although in such neutral circumstances the printer is usually blamed). In others, an editor may feel strongly that the writer has intervened, whether in the lost copy or in proof. In still others, the variation is of a nature readily assignable to the printer. In this last category, and perhaps in the first, the question then arises whether by the act of reading proof (if this act can be established) and failing to alter the printer's differences from the original 'approved' copy the author has made a further act of approval that can cover these variants, on the simple ground of non-interference whether or not in fact he may have noticed their variation.[39] I believe this second proposition is less defensible than the first; nevertheless, it is equally pertinent and hence affects the first. The question often overlooked by advocates of authorial blanket approval is this: which style was approved—printer's copy or final result? What if an author did not read proof for a revised edition but was confident that the publisher could handle the situation. Is he then responsible for having 'approved' the printed result as well as the press copy? The whole matter goes much deeper, however, and into areas too large to discuss here. Prominent is the question whether an author's acceptance of publisher's or printer's styling constitutes 'approval' sufficiently valid and meaningful to be utilized by an editor as a basis for judgment about copy-text.

More copy-text problems are raised by the preservation of manuscripts and typescripts than can be covered here, especially since the textual situation may be complicated by a subsequent revision of the original substantive document. Textual situations may range from editions produced directly from holograph or from authorial typescript

with little authorial attention in proof, to professional typescripts with their peculiar problems of authority and 'approval,' and on to circumstances of extreme complexity in which an author may so rework a manuscript by revising the typescript made from it as to destroy in major part the efficacy of manuscript authority except for a casual accidental here and there, as with D. H. Lawrence's The Rainbow (private information from Dr. Charles Ross). Or as happened with James's A Pluralistic Universe, the author may have revised the typescript of the book once, and then independently further revised parts of another copy of the typescript to prepare copy for the periodical publication of some of the lectures before the book was issued, and finally read proof on each form without reference to the other. The manuscript of A Pluralistic Universe is the only pre-publication document preserved; but James's treatment of it, as manifested in the considerably altered journal and book texts, suggests the view (supported by other parallels) that he really regarded his manuscripts as drafts both for substantives and for accidentals, to be revised and perfected by reworking any typescripts made from them but with the major revision likely to occur in galley proof (those being spacious days). Under these conditions (as worked out in detail in the ACLS edition for the Harvard University Press, 1977), the finally revised book was selected as the most appropriate copy-text.

It will be seen that we are here in an area of substantially different problems (not always so recognized, as remarked in the discussion of Greg's examples of revision) in which revision is pre-publication and not, except as a separate problem, represented by such post-publication revision as is found in James's Pragmatism or in Mill, or in Nashe and Jonson as against the example of pre-publication revision in Troilus and Cressida. In pre-publication revision one may need to deal with such transmissional variation as that created by one or more typescripts (Stephen Crane's Active Service) and, inevitably, with variation from manuscript or typescript due to the compositor(s) of the first edition (The Red Badge of Courage), with an eye out for publishers' readers. Frequently documentary links in the chain of transmission are wanting and their reconstruction difficult if not impossible except on such a hypothetical basis as to be nearly valueless for evidence. Under the more extreme of these circumstances, any preserved manuscript or early typescript material assumes an almost overriding importance unless it appears to the editor that the author has entered the transmissional process at a later stage than the preserved documents and in such a significant manner affecting the accidentals as to promote some perhaps final document like the first edition to superior overall authority in the choice of

copy-text based on the accidentals in the classic Greg rationale. This last was the case with A Pluralistic Universe, and it resulted in the defensible choice of the first edition as copy-text over the manuscript on the grounds that the evidence indicated that the majority of the accidental differences could be imputed to the author. On the contrary, this evidence was wanting in the textual history of Hawthorne's romances The Blithedale Romance, The House of the Seven Gables, and The Marble Faun.

The reasons for the selection of the manuscripts as the copy-texts for editions of these Hawthorne works have been fully discussed in the textual introductions of the Centenary Edition and pinpointed further in my "Practical Texts and Definitive Editions," Essays, pp. 412-439). Basically, several thousand accidental differences in each text appeared between the printer's-copy manuscript and the first edition. On the whole these could not be imputed to Hawthorne (a) since they often ran contrary to the established uniform characteristics of all the manuscripts taken as a group, (b) since even without reference to the above it would be absurd anyway to believe that Hawthorne made so many accidental alterations in proof, and (c) since the amount of accidental variation from manuscript differed among the identified compositors who set the first editions. The general authority of the manuscript accidentals as against those of the first edition thus being established as a working hypothesis, and the manuscript selected as copy-text, the copy-text accidentals were reproduced with relatively few exceptions—these deriving mainly from first-edition correction of positive errors and oversights and from a few forms that seemed characteristic of Hawthorne and conjecturable as his proof-alterations like the similarly slight alteration in proof of the substantives. In these examples of direct linear derivation, the editor had two primary documents of printer's-copy manuscript and printed book. Missing, and their details to be reconstructed by conjecture, were the intervening proofs. Since Hawthorne was not a copious rewriter of his works in proof, the want of these was not such a serious loss as with an author of a different kind. The evidence of the proofs was not really required to demonstrate how much the compositors had altered Hawthorne's intricate parenthetical punctuation system in the interests of simplification although they would have been useful, of course, in positively identifying whatever few accidental (and substantive) changes he did make.

Other examples may hold as well even though the linear transmission has not been so fully preserved in the documents. Stephen Crane's manuscript of The Red Badge of Courage was professionally typed; a copy of this lost typescript, further revised by the author and probably looked

over by Ripley Hitchcock, the Appleton editor (although the depth and extent of any correction are obscure),[40] was given to the printer of the first edition. Missing are not only the proofs (only a negative loss since Crane was a careless and reluctant proofreader) but, more important, the marked-up typescript the original of which certainly contained corruptions that were passed on to the first edition.[41] Viewed as a whole, the first edition is more likely than not to smooth over the idiosyncrasies of Crane's accidentals. It follows that the manuscript reproduces in a more faithful manner than any other preserved (or conjecturally reconstructable) document the peculiar characteristics of Crane's accidentals, eminently worth saving for their flavor and their stylistic effect—often affecting rhythm and pace—and hence it is the most suitable copy-text. In this particular example the existence of the collateral newspaper texts does not affect the copy-text authority, for these stem from a second copy of the typescript in an unrevised state and are at best useful when they agree with the manuscript against the book, or with the book against the manuscript, in establishing the typescript readings and thus in helping to isolate post-manuscript alteration as authorial or as unauthoritative. These newspaper versions have been so seriously condensed from the full version and in part editorially rewritten as to be of little use in emending accidentals, as are the newspaper versions of Crane's Third Violet, however.

The above examples concern different aspects of authorial prepublication intervention in the transmission of a text which has a direct linear relationship between the preserved authorities, even with

missing intermediate documents whose essential features can be reconstructed without insuperable difficulty. Greg's rationale, based on post-publication revision, operates with equal efficiency on simple conditions of pre-publication revision as in Hawthorne; but for more complex cases one must include his recognition (even though generally expressed) that conditions may enforce the selection of a later revised document over an earlier authority if the accidentals appear to be more authoritative. However, questions of multiple authority, as remarked in his references to Hamlet and allied Shakespearean plays, are not satisfactorily resolved in the "Rationale." In multiple authority as found in modern editorial problems the main point of investigation is to establish as specifically as possible whether diverse authority is what may be called technical (or mechanical) in the sense that variation is due exclusively to the transmissional process or whether one branch is weighted by authorial revision. In the first case all documents may be technically of equal accidental authority if at equal distances from the archetype; or they may be of unequal technical authority if some documents are further removed than others. But no matter how equal or unequal the distances, no fresh authority has entered the transmission at any point either directly or by reference. In the second case, a decision needs to be made whether authorial intervention has increased the authority of one line chiefly in respect to the substantives (only casual attention probably having been given to the accidentals) or whether there is evidence that the alteration of accidentals bulked sufficiently large so that accidental variation must be treated with as serious a scrutiny for authority as substantive differences.

The purest example of technical multiple authority that I know of comes in Stephen Crane's "The Price of the Harness," sent in manuscript (now lost) from Cuba to his agent in New York who had a professional typescript (now lost) made from it. One copy of this typescript was sold to the Cosmopolitan in the United States and the other to the British magazine Blackwood's. Crane could have read proof on neither. Because of their immediate radiation from the same typescript with no opportunity for authorial revision, both printed versions have technically the same authority for the accidentals (and in this case for the substantives). The choice of copy-text is one of convenience only and meaningless in principle. It follows that the accidentals of the critically edited text must be drawn freely from both versions according as the editor judges one or the other to reproduce Crane's particular characteristics (as filtered through the typescript) the more faithfully. The same freedom of choice is required for the accidentals as Greg adjures for substantives, and the reproduction of the accidents of one authority

in Greg's and the modern conservative manner would be illogical and anomalous.[42]

Crane's novel Active Service in some part illustrates unequal authority in two arms of radiation. One copy of the typescript that Cora Crane made from the lost manuscript is preserved, this being the setting copy used by Heinemann for the first English edition. A second copy was sent to the publisher Stokes in the United States, who—dissatisfied with Cora's bad typing—had a fresh typescript made from it to use as setting copy for the American edition. The manuscript being lost, the preservation of one copy of the original typescript is basically all that matters to an editor; since Crane's proof-corrections are few or nonexistent, the typescript must be the copy-text and there is no true radiation by the American branch but only derivation. Hypothetically, of course, if the typescript had not been preserved, the two editions would have radiated from this lost document, which would need to be reconstructed from the evidence of their multiple authority, although with lesser weight given to the Stokes edition because of its more distant relationship to the archetype (if this fact could itself have been recovered). Actually the case is more complicated than the preservation of the typescript suggests. From time to time the preserved typescript copy-text is wanting or defective, at which points the Heinemann edition becomes the copy-text as one step nearer to lost authority than the Stokes; but in these places it is proper to correct the Heinemann copy-text by reference to the accidentals of the radiating American edition if these are thought to preserve the formal features of the typescript more faithfully than the English. In Chapter V of the novel by an extraordinary accident the Heinemann typescript is entirely missing but is replaced by the corresponding pages of the American transcript, which had been used in this place as the setting copy for the English edition, the American edition by mirror image being printed from the lost section of the original typescript. In this chapter both the preserved typescript and the Stokes edition radiate at equal removes from the lost original and, as in all situations of multiple authority, the choice of copy-text becomes one of expediency (or convenience)—in this case the American typescript may be taken to have preserved the accidental features of the original more accurately on the whole than the Stokes edition set from Cora's original.

Whenever Cora's typescript is preserved, questions of authority may

slightly weight the matter of copy-text even though—for Crane, whose revision of accidentals was minimal either in the pre- or post-publication stage—these are likely to concern the substantives more than the accidents. That is, the early ribbon part of the original typescript has no written-in authorial revisions, whereas the later carbon section is occasionally corrected and revised. Because this revision agrees substantially with the distinction between ribbon copy and carbon, there is a chance that the copy sent to the United States (although in batches) contained such revisions in its early carbon section which would have been perpetuated in the American edition. One may still feel slightly uncomfortable about this possibility since some of the evidence is contradictory although on the whole against the presence of such authority in the earlier part of the Stokes edition. Even if authority were to be demonstrated, however, the likelihood of Crane's formal alterations, such as they would be, being recognizable after passing through another typing and then the American compositor(s) is so minimal as not to disturb the choice of the preserved typescript as copy-text throughout except for Chapter V and a few minor gaps.

Distance from the archetype is not an automatic rationale to apply without a full evaluation of the evidence. For instance, one copy of the typescript of Crane's Third Violet was sold to a newspaper syndicate which set it up in proofs that in turn were distributed to six known subscribing newspapers as setting copy for their compositors. The other example of the typescript, almost a year later, was used to set the book, somewhat revised by Crane in the interval. In editing this work it was most convenient to choose as copy-text the radiating arm represented by the book, in part because it was one stage closer to the lost typescript than any newspaper, in part because during the revision Crane could have altered any of the typescript accidentals that he did not like (not a probable hypothesis for any extensive alteration but still a possibility), and in part because of the difficulty of utilizing as copy-text a synthetic reconstruction of the lost syndicate proofs made from the evidence of the six newspapers. This latter operation would yield an exact account of the substantives but a less exact even though tolerably full account of the accidents. (Of course, insofar as the syndicate proofs can be reconstructed, they and the book are at equal distance from the typescript: the only problem remains the fact that we know the book's accidentals precisely, whereas in various details some of the proofs' punctuation, for example, must remain conjectural owing to the newspapers' conflicting evidence.) On the other hand, the book shows the effects of publisher's editing (probably) and of compositorial styling (certainly); as a result, in many respects the accidentals of the newspapers reconstructing

the lost proofs are closer to Crane's characteristics than the book's despite the latter's more immediate derivation. As a result, a critical text produced from such radiating authority had to be very much a combination of book and reconstructed syndicate proofs, the total evidence attempting to reproduce as closely as possible the accidents as well as the substantives of the lost typescript, the farthest back a reconstruction can penetrate.

In similar manner the periodical texts of four of William James's Pluralistic Universe lectures in the Hibbert Journal could be balanced against the book to reconstruct in many respects the accidentals of the lost typescript from which each derived at equal distance. Any choice between journal and book as copy-text for this work would be superfluously theoretical; nevertheless, in the editing process the reconstructed accidental characteristics of the lost typescript were of real concern since some of their variance from the preserved manuscript could be attributed to James's own revision.

It may now be possible to sharpen some of the essential differences between the Elizabethan conditions with which Greg's rationale was contrived to deal and those of later periods to which editors now attempt to apply the formulas, with whatever conclusions can be drawn about the modifications that appear to be required in principle or in procedures. Insofar as the single matter of copy-text is concerned, Greg's rationale and his illustrations are centered on the problem whether a linear post-publication revised edition or the original substantive edition makes the superior copy-text, this choice having nothing to do, however, with the second matter of the editorial judgment that will select the most authoritative substantives from the two editions. Throughout he ignores the question of holograph versus first edition as copy-text simply because in his period this problem seldom arises. Even so, an example lay to his hand in the Jonson Masque of Queens which Simpson had edited in volume VII of the Works (1941) in a diplomatic transcript of the holograph. The problems in The Masque of Queens in effect differ little from those facing an editor of Hawthorne's House of the Seven Gables, even though the transmission of the text is not the same and there are added complications in the Jonson: in each the question arises of the extent to which accidental variants that appear to be authoritative can be identified in a restyled printed text and utilized in a critical edition that takes the manuscript as copy-text. In Hawthorne these are the proof-alterations he made; in Jonson the revisions occurred in the ancestral working papers he sent to the printer after making the

preserved fair copy from the unrevised text,[43] with some chance that he also read proofs. The same also applies to the problems met with in A Pluralistic Universe when a manuscript is preserved but not the printer's-copy typescript made from it, a typescript that contained accidental as well as substantive revision in James's hand (corresponding to Jonson's revisions in his working papers except that the James is the more complex problem owing to the intermediate non-authorial typescript whose details are only partly reconstructable where they differed from the manuscript). In the Hawthorne the editor chose the manuscript as copy-text since the identifiable and even the probable authorial accidental revisions in proof were vastly outnumbered by the printer's variants from the manuscript copy. On the contrary, the editor of James chose the printed edition as copy-text on the mirror-image of this situation; that is, the evidence that the printed edition was the culmination of a series of authorial revisions during its pre-publication transmission from original manuscript and constituted a more authoritative document, on the whole, than even the holograph representation of the accidentals. Naturally, possibilities of this nature could not have been contemplated by Greg even if he had tackled the more elementary cases of holograph versus printed edition as, in some part, in The Masque of Queens. It follows that any modern attempt to impose an absolute rule that an author's manuscript is sacrosanct as copy-text finds no support in Greg and has been extrapolated from his illustrations of the transmission of accidentals from one printed edition to another, a quite different matter.

Indeed, Greg less understandably (since these are in greater supply in his period) fails to mention the problems of copy-text that arise in the choice between a printed edition and a scribal manuscript as in Fletcher's Beggars Bush, Woman's Prize, and Bonduca, as well as Suckling's Aglaura. Again, these occur in the considerable area in which a reader draws a blank in Greg—that of pre-publication textual history that may or may not involve authorial revision but that ends in producing at least two different documents with claims to authority as copy-text.

This leaves wide open for an editor any Elizabethan parallels between scribes and modern typists, or other interveners in the transmission of a text such as the friends who assisted T. S. Eliot or Charles Dickens in the reading and alteration of proofs. The limitation of problems of copy-text to the specific post-publication circumstance of an author revising one edition to produce another fails to offer any guidance in the important questions of multiple authority affecting decisions of copy-text as found either in Fletcher's plays or in Crane's Third Violet. In these days in which recording media are joining print as authoritative documents, and a poet reading his own verses may be subject to unofficial as well as official tape recording, something of the sixteenth- and seventeenth-century conditions of multiple copies of variant texts, with their derivatives, are being reproduced although it is true that these have only a tangential relationship to accidental variation except as it could be reconstructed from pauses and the like. Nevertheless, questions of copy-text based on the matter of accidentals may not prove so simple in these cases.

What we come down to in the end is the conclusion (which actually has far-reaching consequences for the relation of the "Rationale" to the editing of modern authors) that Greg's interest in the accidentals of a text was minimal compared to his concern for the free exercise of editorial judgment in respect to the substantives. Sixteenth- and seventeenth-century conditions of transmission, whether through scribes or compositors, left small chance, or at least much uncertainty, that very much of the authorial accidental characteristics were preserved in a print—and the usual absence of authorial manuscripts to act as a guide meant that what had been preserved could not be distinguished from compositorial accidentals. In these circumstances there is little choice but to take as copy-text the document (it is simpler to think basically in terms of documents and not of editions) that is closest to authority in these respects, obviously the first edition when all antecedent documents (scribal or authorial manuscripts) have not been preserved. After this decision Greg's subsequent concern is only with the specific case of Every Man in his Humour in which extensive authorial revision has so altered the substantives as to enforce general acceptance of the Folio accidentals that accompany the altered wording. (All procedures for bad quartos which Greg considers are basically of no value as evidence for the present day.) The general conditions of Greg's period, then, encouraged his view that the chief importance of the formal features of the copy-text outside of their unknown and unknowable relation to the author was to establish the most linguistically suitable texture for the substantives—not the most authoritative accidentals on a demonstrable

basis but the nearest to the general texture of the period of the work's composition. This attitude then leads to the principle that an editor can even scrap the copy-text punctuational system and substitute his own, although he should give "due weight to the original in deciding on his own."[44] "Much the same applies to the use of capitals and italics" (p. 385). This contrast helps to point the difference between the linguistic interest for Greg of the spelling and the lack of critical interest in punctuation and rest. It is interesting to notice that for an editor of later literature the importance of these two concerns has been precisely reversed.

Greg's natural pessimism about the specific authority of accidentals tends to widen the gap between the editorial treatment he envisages as suitable for them and for substantives and ends by removing the accidentals in major part from the function of the editorial judgment appropriate for the substantives. Once again, Greg's concern for the post-publication transmission of texts in linear revised editions prevents him from considering other conditions (even though they may seem to constitute special cases) in which something can be done with the choice of copy-text between two competing non-linear authorities as in the manuscripts and the printed texts in the Beaumont and Fletcher 1647 Folio, and in the possibilities for emending the copy-text accidentals from other authority. In this latter respect he shows practically no interest in the emendation of the accidentals that should demand editorial attention even in the linear original and revised texts of Sejanus. It is this ignoring of the possibilities that exist in some early works for the judgmental treatment of certain classes of accidents on a par with the substantives that creates the gravest difficulties when the "Rationale" is narrowly applied to the editing of modern works. One can be a rebel to Greg's general though not absolute advice to seek one's copy-text in the earliest authority, as is Dr. Robson's editorial board for the Mill's edition, and still fall into the trap of preserving Greg's conservative views on the opportunities for authoritative emendation of the copy-text accidentals.

This attitude has had, apparently, a serious effect on modern editing where the amount of information available about authorial characteristics equals and normally surpasses that preserved for Jonson and puts matters on quite a different plane from Shakespeare and other

typical early writers who bulked large in Greg's mind. The effect is found, of course, precisely where Greg's own gaps occur: the possibility that the accidentals of a revised edition may be more authoritative than those of an earlier edition closer to lost original authority, and even that a printed first edition may be more authoritative in its formal features than the holograph manuscript from which it ultimately derives. That either exception is possible save under the rarest and most special of circumstances has been vigorously denied by some modern converts to Greg's textual theories. In fact, this denial is justified only in cases of a limited authorial interest in the revision of accidentals, these being the commonest conditions an editor may encounter in dealing with revisions. But the less common although by no means highly exceptional examples of the contrary, as occur in the writings of Walt Whitman and William James, are being supported as more editors tackle writers outside of the limited scope of nineteenth-century American fiction. James Joyce and D. H. Lawrence come to mind.

Until editors have had more experience with the special cases of revised editions as copy-texts it is not possible to lay down more than a few specific guidelines, with the advice to seek out and study the examples of this editorial procedure in order to acquaint oneself at first hand with its theory and practice. However, it would seem that several criteria must be satisfied before a revised edition can become a superior copy-text.[45] First, evidence must exist, or be conjectured with major probability, that the author revised his text not only in the substantives but with more than casual attention to the accidentals; second, evidence should be sought that the printer of the revised text was relatively faithful to his copy, or was made so by scrupulous authorial proofreading, so that the author's accidental texture has not been restyled in any thoroughgoing way. If these two criteria are met, then a third may be examined. Briefly, an editor needs to examine his collations of the variant accidentals to determine what categories are manifestly authorial and what are manifestly compositorial. These two groups should be isolated, for since they constitute the classes of accidentals that are as

subject to critical selection or discard as are the substantives, technically they need have no more influence on the choice of copy-text than do the substantives. This is an important point in the revised rationale I am suggesting.

The copy-text is chosen on the basis of its accidentals according to Greg, for in his opinion these constitute a body not subject to the selective judgment he advocated for the substantives. The choice based on the accidentals is still (and no doubt invariably) sound in my opinion for any period although the reasons for the significance of accidentals in a text have shifted materially since the Elizabethan period. When conditions in later authors approximate those that influenced Greg to advocate as a general rule the earliest document (meaning the one closest to the authorial archetype), the rationale is also valid in both its parts. For instance, the evidence strongly suggests that in collecting the 1837 Twice-Told Tales from their various newspaper and magazine appearances Hawthorne performed a minimum of revision in the printer's copy and in proof; hence the accidentals of the originals, set from holograph, are more authoritative than those of the book reprint, although not necessarily reproducing exactly what would have appeared in the lost manuscripts.[46] This was post-publication revision of the kind that Greg recognizes. In pre-publication revision, the evidence suggests that the vast majority of the variant accidentals in the printed texts of Hawthorne's romances belong to the compositors and that some of the relatively few accidental alterations he may be supposed to have made in proof from the forms of the printer's-copy manuscripts are often unrecognizable. In earlier literature the case of the revised second and third editions of Dryden's Indian Emperour may again be cited as examples under (b) of an author whose accidental revisions may be thought (on some concrete evidence from the press-corrections in one forme of the second edition) to have been significant from time to time; but they are insufficiently idiosyncratic to be recognized among the far larger

number of compositorial variants, and therefore the best copy-text is either the scribal manuscript of a slightly earlier version or else the first edition.[47]

Again in earlier literature, it would seem that whereas Sejanus meets the first test, in that Jonson may be taken to have revised copy for the Folio with some care for its accidentals, the second test—the general fidelity of the Folio printer—appears to be failed. Thus a doubt is raised, although somewhat less than with Dryden, about the amount of compositorial departure from copy in the Folio text. When one proceeds to the third test, in order to see if the doubt may be resolved, an editor finds it possible to isolate several ranges of Jonson's idiosyncratic accidentals that beyond question represent his markings in the Quarto printer's copy. These may be put aside (like the F press-corrections) since for the moment they should not affect the selection of copy-text. The clearcut compositorial variants are less useful, representing as they do only the Folio departure from recognizably Jonsonian characteristics found in the Quarto copy. At this point an editor may discover that the remaining accidental variants (chiefly punctuation) are still fairly numerous. If surveying these he comes to believe that on the whole they are more likely than not to represent the F compositors' variants, then he should choose Q as his copy-text since the Q accidentals in this particular category will probably reproduce more authority as a group than those in F. The set-aside idiosyncratic accidentals can then be inserted in the Q copy-text as authorial revisions holding the same status as the F authoritative substantive variants. On the other hand, if the editor is still uncertain about the neutral category of the accidentals and has some evidence from parallels to take it that, although each separate one is not identifiable, the probability rests that on the whole they are more likely in F to reproduce Jonsonian markings in copy (and perhaps undetected proof-correction changes if this evidence has not been exhausted)

than compositorial variation from Q, then revised F becomes the natural copy-text. If so, it is the duty of the editor to correct those accidentals in F that appear to be unauthoritative by drawing on Q for the corresponding versions.

To repeat, the probability on what evidence is available about the author and his characteristics (and the printer) that a majority of the generally unassignable accidental variants is authorial or compositorial should be the ultimate determinant in the selection of copy-text as between two authorities, whether an original and a revised edition, or two collateral editions.[48] For example, it was this general sense of the authority of the indeterminate class of accidental variants between the partially reconstructed syndicate newspaper proofs and the first edition of Crane's Third Violet that finally dictated the selection of the first edition as copy-text. Under other circumstances when no question of revision was present, the Boston Museum version of Hawthorne's "Ethan Brand" appears to have received less house-styling than another publication in the Dollar Magazine (both radiating at equal distance from lost printed sheets from an unpublished magazine set from manuscript). Although the Dollar Magazine clipping was the printer's copy for the later appearance in The Snow-Image (1851), no evidence is preserved to indicate that Hawthorne had prepared it with any revisions. Hence an editor may believe that the copy-text offering maximum general authority, such as it is, is the Boston Museum (Centenary Edition, XI, 417-419). Compositorial studies showing unequal levels of variation from copy according to the stints were among the other evidence in Hawthorne's romances that led to the establishment of the manuscripts as the most trustworthy source for the accidentals as a whole. With a more complicated transmission from preserved manuscripts (in large part) through periodical publication to book collection, despite considerable substantive revision in the process the manuscript accidentals for Hawthorne's Our Old Home (1863) remained on the whole the more authoritative, also. In another situation the grave

difficulty of accepting in the first edition of The Red Badge of Courage a number of what appeared to be sophisticating accidentals not necessarily Crane's and far more numerous from the manuscript than could be attributed to his marking of the typescript printer's copy and the book proofs, led to the selection of the manuscript as on the whole the best repository of the authoritative accidentals of this work.

As a marked contrast, when one approaches such a typical William James text as A Pluralistic Universe, each criterion for copy-text may be satisfied in favor of the general authority of the revised edition's accidentals. Briefly, preserved documents for other works in its period demonstrate the frequency of James's attention to accidentals in preparing printer's copy and in the revision of a book's galley-proofs. Second, the evidence of Some Problems of Philosophy indicates that the Riverside Press was exceptionally faithful in setting the accidents of his copy except for a few categories of housestyling that can be readily isolated. In the third test, although some of James's idiosyncratic accidentals had clearly been inserted in the book as deliberate alterations of the typescript (and manuscript), a large number of the book's punctuation changes could not be so precisely assigned to James in clearly defined categories; nevertheless, in the majority of cases they were either consistent with his favorite although not invariable practices or else were, at the least, not inconsistent. Various of the book's changes could be easily interpreted as James's attempts to make consistent and formal the more informal system of a manuscript that was in effect not much more than a draft which James had always intended to revise thoroughly before publication. Since evidence within the book and the journals existed for James's concern in revising his accidentals during the stages of the copy's transmission from manuscript through typescript to final print, and since the majority of the relatively indifferent variants did not seem to be assignable to the printer either on internal or external evidence, the book became the natural copy-text. The chief editorial problem for the accidentals, then, was the identification of the Press's styling so that it could be removed in favor of the authorial forms from the manuscript but with some reference as well to the possibility of post-typescript revision in the Hibbert Journal publication not transferred to the book. All this was pre-publication revision of course. Post-publication revision of the same nature and with the same problems occurs in the essays in Pragmatism, The Meaning of Truth, and The Will to Believe in which journal articles were revised to serve as printer's copy for the book and the revision of the text in both its aspects continued in the book's galley-proofs.

Pre-publication revision during transmission can affect the choice of copy-text when the option lies, as usually, between a preserved manuscript or typescript and the first edition. (Marked proofs for a book are preserved much less frequently than other antecedent documents, but even unmarked early proofs may be valuable for demonstrating what differences from the setting copy resulted from the transmission.) The authorial revision in some intermediate stage(s) needs then to be reconstructed. In the simplest cases when a manuscript was the printer's copy as in the Hawthorne romances or in Lectures VII and VIII of James's Pragmatism, this may be no more than the reconstruction of the proof-alterations. In more complex cases, as in James's Pluralistic Universe, a lost worked-over professional typescript plus extensive authorial proof-alteration needs reconstruction as well. In post-publication revision, ordinarily only the two printed documents (original and revised) are available, as in Fielding's Tom Jones or Joseph Andrews, Hawthorne's Mosses from an Old Manse, or Lectures I-VI of Pragmatism. Here the problem is to reconstruct the lost authorially prepared printer's copy and to attempt to evaluate the role of subsequent proof-corrections, if any. (Normally these proof-corrections could not be distinguished in the book's text from the marking of printer's copy, of course.) The circumstances will dictate the choice of copy-text. If with Hawthorne's Marble Faun—where bibliographical analysis can distinguish between Hawthorne's proof-correction and the printer's variants for major parts of the first edition—the decision is made that the printer's departures from copy in the accidents completely outweigh the possibility of the author's proof-alterations, the manuscript becomes the copy-text. If with James's Pluralistic Universe the decision is made that James's alterations in the typescript and in the proof are more numerous than the printer's departures from copy not only in respect to the idiosyncratic and hence identifiable accidentals but also to the more

neutral forms, the first edition becomes the preferred copy-text over the manuscript. If with Locke's Of Human Understanding the decision is made that Locke not only revised the accidentals as well as the substantives for the second edition but also used this revised text as the physical basis for a further revision of both features by marking copy and reading proofs for the revised fourth, then with Professor Nidditch one would choose this fourth edition instead of the first as copy-text.

As every editor knows, the choice of copy-text is important not only as a means of preserving major authority in the accidents as a whole but also as providing an editor with a working hypothesis to that end when he is faced with transmissional variation whether pre- or post-publication. Greg believed that the pull of the copy-text authority could operate with the substantives as well, and he advised an editor to rely on the copy-text instead of a revised edition in cases where the choice of a substantive reading was perfectly balanced. This is often sage advice in the sixteenth and seventeenth centuries when the incidence of compositorial substantive error is high and any revised edition is likely to be a mixture of wrong and of right readings.[50] It may be less useful later.

The emphasis must rest, of course, on the complete neutrality of the readings, making it impossible in the editor's opinion to select one or the other on reasoned critical grounds. In such cases of exact balance Greg's mistrust of Renaissance compositors leads him to favor the original reading. We may suppose he had in mind the hypothesis that the original was presumptively authorial. If one then deliberately chose the indifferent reading of a revised edition, the chance entered that one was rejecting authority for a compositorial misreading or memorial contamination, especially since an apparent motive for authorial revision may seem to be wanting. Since the critical judgment could not affirm the authority of the new reading, it would seem better to stick with the authority one knows than to gamble on the unknown quality of the variant which, moreover, offered little incentive to forsake the comfort of the assumedly authoritative known. This line of reasoning gambles, in effect, that the original reading was not a misprint or corruption and the revised reading an authentic correction, but the reverse.

It must be emphasized that in this matter of advocating the authority of the copy-text for evenly balanced substantive variants Greg was writing theoretically and he offered no illustrations. However, one may speculate that what he actually had in mind was less likely to be examples of linear revised relationship as in Nashe or Jonson but instead the more serious problems found in plays like King Lear where the Folio copy-text readings do not always seem superior to those of the Bad Quarto, and where there are a number of cruxes on which editors divide. Like King Lear Shakespeare's Othello involves transmissional prepublication problems, although of a different sort, that involve scribes as well as compositors, and was probably also in Greg's thought. However, the complexities of bad quarto and 'revised edition' have no relation to modern problems, nor is the Othello situation, or that of 2 Henry IV (as we dimly conceive them) very likely to arise. For any pertinence to other than special Renaissance problems one must confine the case, practically speaking, to linear transmission, either from manuscript or typescript to printed edition or else to an edition revised from the text of an earlier. Under these conditions, the closer one comes to periods where compositorial accuracy improves—especially in the setting from printed copy—the more the authority grows in favor of variants in a revised edition and the more likely it is that an indifferent variant in the revised text is authorial, not compositorial.[51] If so, a very real question arises whether Greg's advice is a good editorial principle to adopt under changed conditions from those of Renaissance compositorial and scribal free-wheeling.

Evidence for either position is hard to find that can be called demonstrable, for if the answer is known between two seemingly balanced variants then they are no longer truly balanced and the case becomes hypothetical. For instance, since neither occurs in his holograph papers it would take the evidence of a computer concordance of Fielding to learn whether in Joseph Andrews Fanny had more likely 'laid hold on

the Girdle which her Lover wore for that purpose' (273.13-14) or, as the revised third edition has it, laid hold of; or whether between the first and third editions Fielding had come to feel that 'was the Subject of your Contention anywise material' had better be changed to anyways (and then in the fifth edition to any ways). Ultimately these difficult decisions involving idiom may come to be less conjecturally based than what songs the sirens sang;[52] but more certainty is possible in other matters that might at first sight appear to be equally balanced. For example, beginning with this revised third edition a number of alterations affecting small irregularities of modification and coherence are improved, such as 48.21 first edition 'She was a poor Girl, who had formerly been bred up in Sir John's Family' altered in the third to been formerly bred up, a change seemingly of some indifference and possible as the compositor's memorial failure. A more obvious example of the same is the original 'whom he observed not to be fallen into the most compassionate Hands' (61.25-26) which appears in the third edition as to be fallen not into; or 'If it was only our present Business to make Similies' (45.30) altered in the third to was our present Business only to make. If an editor stuck to the copy-text in such readings, taking it that the third-edition order resulted from memorial failure or a purist compositor intent on making the style more formal, he would quite definitely be wrong, for (a) the third edition exhibits half a dozen more cases scattered sufficiently to make it unlikely that all were the doing of one compositor,[53] (b) similar improvements are made in the fourth and even in the revised fifth edition, (c) the identical kind of alteration is a feature of the revised fourth edition of Tom Jones.

Contiguity may play a part in the evidence an editor seeks, for in a revised edition it may seem more likely that a cluster of variants results from authorial marking than from a sudden spate of compositorial error. For instance, in Joseph Andrews 57.8-18 the text of the first two editions reads:

In the revised third edition three substantive variants occur: ever he for he ever, should have for should have had, and Circumstances for Circumstance. Each one of these seems about evenly balanced but their clustering pyramids the advisability of adopting the third-edition readings as authorial alterations.

An author's special revisory interests will sometimes indicate that one or other of a choice—revised edition or original copy-text—is the more authoritative. Fielding's concern for clarifying his modification was only one of his interests in revising both Joseph Andrews and Tom Jones. Another was to weed out certain old-fashioned usages in the preterites of verbs, like altering bid to bad (five times) and begun to began. This is also a feature of the revision of Tom Jones in the fourth edition, in which tore is changed to torn, bore to borne, and begun to began (four times).[54] Other concerns manifest themselves in revisions. In Tom Jones the first-edition copy-text reads, 'yet so discreet was she in her Conduct, that her Prudence was as much on the Guard, as if she had had all the Snares to apprehend which were ever laid for her whole Sex' (36.9-37.2). This seems unexceptionable, and the revised fourth edition's if she had all could easily be an eyeskip. An editor might be tempted to retain the copy-text unless he had observed that the fourth edition alters had had to had nine additional times: clearly Fielding was making a special revision throughout the novel. On the other hand, when in Joseph Andrews or Tom Jones the revised editions slip and substitute has for hath, or does for doth, an editor may safely retain the

copy-text, there being dozens of examples in the two novels of Fielding's revision to his preferred forms hath and doth.

Not only substantives but forms of words ordinarily classed among the accidentals but better listed as semi-substantives because of their linguistic interest may offer serious problems because they are as subject to compositorial as to authorial modernization. In Tom Jones whether the first edition Recipe or the revised fourth's Receipt is authorial cannot be known in the absence of the word in preserved holographs. In the fourth edition Fielding seems to have been concerned to alter older and ambiguous forms to modern, as in Council to Counsel, and errant to arrant (see the textual notes to 111.39 and 190.15 on p. 1014), although he missed wave for waive (212.2). This being so, the question arises whether the invariable change (except for its single use as a name) of Ostler to Hostler in the fourth edition is compositorial. In earlier works the word had been printed as Hostler, but it is hard to believe that the workmen setting the first edition in different places altered manuscript Hostler to Ostler. Since there seems some reason to conjecture that Ostler was what appeared in the manuscript, it is difficult to know whether the fourth edition's changes represent Fielding's or the compositors' modernization, and the same for the change from hollowing to hallowing some half dozen times, especially since hollowing is the form printed in Fielding's earlier works. A conservative editor may prefer to stay with his copy-text, a more adventurous one to suspect that Fielding may have been the modernizer.

In any large work like Tom Jones or even Joseph Andrews, intermediate between Elizabethan and modern printing, small verbal differences will of course occur that seem to be perfectly balanced especially when the differences are slight. An example might be cited in Joseph Andrews (27.25-26) where the first-edition copy-text reads, 'She plainly saw the Effects which Town-Air hath on the soberest Constitutions' but in the revised second edition the Town-Air, an easy printer's sophistication but a possible authorial correction or revision. In Tom Jones the ubiquitous problem of these and those rears its ugly head, as at 54.32, 74.28, 385.10, etc., variants that seem paralleled by Joseph Andrews 242.2, which may or may not support the otherwise more doubtful variant in the same work at 126.12. It is particularly difficult to attack such problems, often of misreading, when the compositors cannot be identified. A misreading problem in which one word is very likely right and the other wrong is posed in Joseph Andrews when the captain pulled the chair from under Parson Adams as he was sitting down, so that 'he fell down on the Ground; and thus [or this as in 4-5] completed Joke the first, to the great Entertainment of the whole Company' (245.37).

Perhaps in the category of Fanny's having laid hold on or of Joseph's girdle but slightly more possible as an authorial revision is the question whether Lady Booby's soul was 'tossed up and down by [or with in 5] turbulent and opposite Passions' (287.6). When choices of reading involve such exquisite factors, and the problems of compositorial misreading or memorial corruption in original or revised edition are complicated by the possibility of finicky authorial revision, the presumption of error in the reading of the revised edition that led Greg to suggest the advisability of retaining the copy-text is not quite so firm even though it may sometimes be operative.

The differences may be observed in the Wesleyan-Clarendon editions of Joseph Andrews and Tom Jones. For the first, in my role as textual consultant I recognized that each edition through the fifth had been revised—not in any sense thoroughly but by a polishing and refinement of the language and syntax not always in a conspicuous manner—yet it also seemed clear that the fidelity of the compositors was suspect beginning with the third edition, the doubt increasing as the identifiable authorial revision dwindled in the fourth and fifth editions. This uneasiness about the authority of various small changes, combined with the powerful influence Greg's precept about indifferent choices then exercised on my thinking, led me to suggest the advisability of marked caution in the acceptance of minor differences in the revised editions, advice that I now see led to too many retentions of copy-text authority and too few inclusions of neutral-seeming revisions. Later, doing the Tom Jones text from scratch as my own responsibility, I conceived a higher opinion of the fourth edition's faithfulness to copy despite a number of small lapses by its workmen, and a lower opinion of the value of Greg's advice in periods later than the Renaissance, especially as applied to linear revised editions set with greater general accuracy than is expected in Elizabethan reprints. As a result, in Tom Jones proportionally far more indifferent variants are inserted from the revised edition into the first-edition copy-text than in the conservatively treated Joseph Andrews, although still perhaps not quite enough (because it is difficult not to suspect small idiomatic changes in a revision). In short, eighteenth-century are not Elizabethan compositors; and from this experience I conceived a distrust under some circumstances for the general principle of retaining the copy-text reading in cases of balanced variants. In later periods it is even possible that the advice is dangerous since it may foster an attitude too conservative for the changed conditions. When a modern author gives evidence in his revision that he has been concerned with small things as well as great, I now believe that an editor is better advised to give the benefit of the

doubt to the readings of the revised edition provided he has some reasonable faith in the general fidelity of its compositors to copy[55] and believes that the author was capable of making the changes in question. On the whole, this applies to accidentals as well as to substantives and hence may lead to a selection of the revised edition as copy-text, in which case an editor may return to Greg's conservatism but in the reverse direction.

In the Elizabethan period the nature and extent of authorial revision is sometimes in doubt, and variation between editions representing some authority (as possibly in Hamlet or Othello) may arise not from authorial intervention at some stage but from different textual traditions complicated by scribal transcripts (and in the drama by theatrical tinkering). Under such circumstances one needs to be conservative and to require a possible 'revision' (especially arising in prepublication transmission) to cross the balance line before a supposedly authoritative original is to be altered—like fretful for fearful porpentine. But the climate of opinion changes with post-publication authorial revision of an earlier print. Joseph Andrews and Tom Jones in the mid-eighteenth century offer a halfway house for the work of a practising author in ordinary conditions whose copy was transmitted by compositors still subject to the faults inherent in hand setting. (Alexander Pope's careful attention to his revised texts in preparation and in proof is quite another matter, just as Walt Whitman is another matter.)

By the time one reaches the nineteenth century, although still in the period of hand setting with its greater opportunities for compositorial sophistication, the accuracy in following copy increases for the substantives and with it the odds that these variants in a revised edition are the more likely to be authorial in an overwhelming majority of cases. In Hawthorne's Twice-Told Tales, for instance, in "The Gray Champion" seven authorial substantive revisions appear to have been made in the 1837 collection from the periodical copy but no other variants, and the tale was reprinted in three other editions to 1853 without further substantive difference. In "The Minister's Black Veil" the only unauthorial substantive variants are two that crept into the fourth-edition

reprint; in "The May-Pole of Merry Mount" seven substantive revisions appear in 1837 as against one misprint; in "Mr. Higginbotham's Catastrophe" Hawthorne made twenty-nine revisions in 1837 although here an estimated four printer's errors appear simultaneously; in "The Great Carbuncle" five revisions appear but no unauthorial substantive changes in 1837 (four occur in the unrevised third edition of 1851), and so on. By the time one reaches William James, no more than one or two unauthoritative substantive variants (if that) are likely to appear in any one of his collections revised from journal articles, although there is a good possibility that in the revised The Will to Believe (1897) it was the printer who was responsible for the consistent change of amongst to among.

In fact, it is not until one encounters a textual situation as in James's The Meaning of Truth in which the revised edition is itself the copytext that Greg's advice, paradoxically, proves to be sound for modern literature, since the cases are few in a carefully proofread book like this where the substantives of the text (and not just some accidentals) may need correction by reference to an earlier authority. However, the scrupulousness of a James does not seem to be required to reverse Greg's principle for any period after the Renaissance—and it may be that even in this period the application needs more testing with the readings of authorially revised linear editions. One may flip a coin or for lack of a better reason adhere to the copy-text with a complicated textual situation like that in Othello, or King Lear,[56] but for normal conditions when in genuine doubt the odds may be taken to favor the in-line revision over the reading of the earlier copy-text. The wheel comes full circle, of course, when the revised edition is the copy-text.

Notes