University of Virginia Library

Editing without a Copy-Text
by
G. Thomas Tanselle [*]

In the history of textual criticism, as of most human affairs, a few basic viewpoints have moved through the centuries in cyclical fashion, each losing favor temporarily in one area or another and then returning to prominence in an altered form. The fields concerned with reconstructing the past, like textual criticism, constantly reverberate with alternating claims about the place of judgment in the process. At one moment, objectivity seems possible, and artifacts tell their own stories with little or no assistance; at the next moment, subjectivity is welcomed, and artifacts are a springboard for the historical imagination. In the nineteenth century, for example, the genealogical approach to biblical and classical textual criticism, now associated with the name of Karl Lachmann, emerged from a desire to minimize the role of judgment in combining readings from variant texts and was thus a reaction to the less disciplined eclecticism of many eighteenth-century editors, who often altered texts according to their personal tastes. By the early twentieth century, in turn, there were efforts—as in the work of A. E. Housman—to reinstate an open acceptance of judgment. But even Housman's brilliant advocacy of the subjective element in textual criticism did not prevent several determined attempts in the twentieth century to develop mathematical procedures for weighing variants in ancient texts.

A distrust of subjectivity, in a different form from Lachmann's, came in the twentieth century to dominate the study of medieval writings: the procedure, which was anti-eclectic, involved selecting a "best text" and altering it only at the places that seemed obviously erroneous. Although, for post-medieval literature, the term "best text" has not been widely used, the term "copy-text" as employed by R. B. McKerrow in his 1904 edition of Thomas Nashe meant essentially the same thing. Thus the editing of Renaissance literature that proceeded alongside the "New Bibliography" emerged from this restrictive base—not unexpectedly,

since what was new in the "New Bibliography" (developed by McKerrow, A. W. Pollard, W. W. Greg, and their followers) was the analysis of the physical details of books for evidence that could solve textual cruxes without the use of literary judgment (or at least limit the role of such judgment in making textual decisions). No approach to editing can totally eliminate judgment, of course; and I am less concerned here with the amount of judgment that actually entered into various editors' work than with the general avoidance or acceptance of judgment embodied in their editorial theories. McKerrow, by the end of his life in 1940, had moved toward a greater willingness to allow variants from one text to be incorporated into a chosen copy-text, as he outlined his plans for an edition of Shakespeare. But it remained for W. W. Greg in 1949 to enunciate, in his famous lecture "The Rationale of Copy-Text," a position for the editing of Renaissance drama that approached Housman's of a half-century earlier for the editing of classical literature. Greg, like Housman, elevated the role of informed judgment in choosing among variant readings.[1]

One of the most revealing facts about his rationale, however, is—as the title of his paper suggests—its retention of a concept of "copy-text," a basic text into which alterations (or "emendations")[2] can be incorporated. It now seems time, after another half-century, to move beyond this often useful but nevertheless inherently restrictive concept. That Greg's intention was to liberate editorial judgment is indicated by his warning against "what may be called the tyranny of the copy-text"; but his recognition that a copy-text could indeed tyrannize did not cause him to abandon the concept. He was not quite ready to carry to its logical conclusion the dominant twentieth-century English line of thinking about the textual criticism of post-medieval literature, a line that had become gradually more liberal during the first half of the century—though the position he did take certainly constituted an important extension of it.

Greg's rationale has been much debated in the succeeding half-century, but the discussions have generally focused on his recommendations for selecting and emending a copy-text rather than on the necessity for designating a particular text as copy-text in the first place. To consider the latter issue is not to imply any repudiation of Greg at all; rather, living with his ideas for a considerable period has enabled us to see more clearly their essential direction and has put us in a position to understand how, paradoxically, a reduction of emphasis on "copy-text" actually builds on and completes his argument.

I

The first step in thinking through a new approach to the question of copy-text is to examine whether there is a future for critical editions—editions, that is, in which the editors, using their informed critical judgment, make alterations in the documentary forms of texts that have come down to us (or at least allow themselves the freedom to make such alterations, whether or not they ultimately find any changes to be necessary). All the various "best-text" or "copy-text" theories are concerned with the process of altering documentary texts; but if a convincing argument can be made that no such process is justified, then there is no point considering critical editing further. In the continual give-and-take of arguments over subjectivity and objectivity, some scholars naturally take the position that editions presenting critical texts are less valuable (if granted any value at all) than editions containing facsimile or diplomatic (or computer "hypertext") reproductions of texts as they appear in extant documents.[3] The late twentieth century has seen an energetic resurgence of this view, springing from two seemingly contradictory premises. One—that authors do not control either the language they use or the forces that allow their work to reach the public—denigrates the significance of authorial intention and thus brings renewed attention to texts as they were published. The other premise—that the process of revision and change through which verbal works move is of more valid concern than any single final text that may be postulated—causes increased interest to be directed toward the texts of every prepublication draft and revised edition. Although the former premise tends to reject the author and the

latter to concentrate exclusively on the author, both reflect disaffection with critical editing as a supposedly authoritarian imposition of stasis on inherently unstable material.[4]

These trends have therefore publicized specific programmatic reasons for the production of facsimile and diplomatic editions, besides the obvious general motivation to make widely available the texts of unique manuscripts and scarce printed editions (recognizing, of course, that no photograph or transcription can preserve all the physical evidence of the original—and that all of it may be relevant to textual study).[5] Although facsimile and diplomatic editions are noncritical in that their editors' aim is to reproduce without alteration the words and punctuation of documentary texts, critical judgment is inevitably involved in deciphering handwritten (or poorly printed) texts and in deciding which documentary versions of a work to present (if all are not to be included). Because this decision-making is not intended to alter texts, however, the editions that result are usually thought of as more "objective," despite the number of subjective decisions that may have been involved; and although arguments about the relative amounts of subjectivity and objectivity in editing are normally concerned with critical editions, there is no doubt that those persons who wish to minimize subjectivity in critical editing should logically be drawn to noncritical editions. Facsimile and diplomatic editions, whatever subjectivity may underlie them, are fundamentally different in conception from editions in which the goal is to alter documentary texts according to some predetermined guideline. By not requiring such guidelines, they have seemed to need less discussion over the years than have critical editions; and their increased presence in recent methodological debates is to be welcomed as a partial redressing of the balance. There is no question that they serve an important function in the study of the past.

But any attempt to argue that they are necessarily superior to critical editions, or indeed that they constitute the only legitimate kind of edition, cannot possibly succeed. The two kinds must always coexist, for they represent two indispensable elements in approaching the past: the ordered presentation of artifactual evidence, and the creation, from that evidence, of versions of past moments that are intended to be more comprehensively faithful than the artifacts themselves—random (and perhaps damaged) survivors as they are. It is not possible, in any case, to prevent human beings from interpreting evidence; to ban critical editions would be as futile as to try to suppress any other product that reflects the natural

working of the mind. Critical editions, however, are not merely inevitable; they are desirable. A text reconstructed by a person who is immersed in, and has thought deeply about, the body of surviving evidence relevant to a work, its author, and its time may well teach the rest of us something we could not have discovered for ourselves, even if the reconstruction can never be definitive—and even if, indeed, it places us in a position to criticize its own constitution. Authorially intended texts, which have been the goal of almost all critical editions in the past, cannot be expected to reside, in perfect accuracy, in surviving documents—or perhaps, for that matter, in any documents that ever existed. But the fact that they are not—and possibly never were—fully available in physical form does not deprive them of the status of historical events. (The same could of course be said of texts as intended by publishers or any of the other individuals that had a hand in the production process.)[6] Some people may not be interested in reconstructing such events, but their lack of interest cannot render the effort invalid.

There is another reason that critical editions are essential: they are demanded by the very nature of verbal works. Like musical and choreographic works—and unlike works of visual and plastic art—verbal works employ an intangible medium. Any tangible representation of such a work—as in letterforms on paper—cannot be the work itself, just as choreographic notation or traditional musical scores are not works of dance or music. The media involved—language, movement, and sound—being intangible, these works can be stored only through conversion to another form, which in effect becomes a set of instructions for reconstituting the works.[7] Any instructions—indeed, any kind of reproduction or report—may be inaccurate, and thus every attempt to reconstruct such works (or versions of works) must include a readiness to recognize textual errors in their stored forms. Understanding the difference between works and documents also enables one to see that a concern with the integrity of discrete versions is not incompatible with the adoption of variants from different documents.[8] Reconstituting works (or versions) in intangible

media is a critical activity, and universal agreement about their makeup cannot be expected. But if we wish to experience the texts of works (or versions), and not simply the texts of documents, we must leave the certainty (or relative certainty) of the documentary texts for the uncertainty of our reconstructions.[9] Every act of reading is in fact an act of critical editing: we often call critical essays "readings," and critical editions are also the records of readings. The editors who produce them earn the respect of other readers to the extent that their work reflects historical learning and literary sensitivity; but no reader is likely to agree with every decision made by any other reader, even the most respected critical editor. The process of critical editing is the ineluctable, if unending, effort to surmount the limitations of artifacts in the pursuit of works from the past.

II

Having established the necessity of critical editions, one can then consider what procedure should be followed in making the critical judgments they entail. Some kind of guideline is required if the operation is to be disciplined and historically oriented. Otherwise, textual decisions would simply reflect the editors' own preferences, and the results, which would not necessarily be without interest, would not be an attempt at historical reconstruction. Editorial taste is indeed essential, but an edited text should reflect, not the personal preferences of the editor, but the editor's judgments regarding the preferences of the author, or the author in conjunction with others, at a given moment. Simple as this distinction is, it has probably been the root of more textual disputes than any other single point. But since critical editing must rest on editorial judgment, a sufficient guideline would seem to be one that states the goal toward which that judgment is to be directed, perhaps indicating what ancillary information should be taken into account but without placing limits on the judgment itself: to say, for example, that an editor's literary sensitivity, informed by biographical, bibliographical, and more broadly

historical research, should be employed to attempt to determine the constitution of texts at particular past moments.

The fear that encouraging editorial judgment would amount to licensing personal preference, however, has repeatedly caused textual theorists to impose restrictions on the field within which editorial judgment is allowed to operate. The "best-text" theories are one result, and they have been curiously persistent, despite the ease with which their essential illogic can be exposed. In ruling that texts can be altered only where they are obviously incorrect, this approach seems to imply, incredibly, that texts are likely to be correct wherever they are not manifestly incorrect—a patent absurdity. A. E. Housman in 1903 pointed out this flaw most memorably, in the preface to the first volume of his edition of Manilius, and he wittily added, "assuredly there is no trade on earth, excepting textual criticism, in which the name of prudence would be given to that habit of mind which in ordinary human life is called credulity."[10] Perhaps the most basic way of stating the incoherence of the "best-text" approach is to observe that it begins with a belief that documentary texts can be improved through editorial intervention (otherwise there would be no reason to allow the correction of clearcut errors) but proceeds to cast doubt on the usefulness of such intervention (otherwise it would be allowed to operate more widely). If there were a chance that editorial judgment could correct a text at places not obviously incorrect, there would seem to be no reason not to sanction the effort; and one must therefore conclude, with Housman, that these places are assumed to be correct. Actually, however, there is no point looking for an explanation of this unreasoned approach other than a reflex reaction: the belief that a restriction of judgment was required to improve on the unscholarly eclecticism of the past.

McKerrow was undoubtedly caught up in this reaction when in 1904 he chose the second edition of Nashe's The Unfortunate Traveller as his copy-text, to be altered only at obviously erroneous points, on the grounds that it contained some revisions by the author (2: 197). He was not happy with all the readings that this policy forced him to retain, but he felt that he had "no choice" in the matter; when he stated that an editor could not "pick and choose among the variant readings of his author's works those which he himself would prefer in writings of his own," McKerrow did not admit the possibility that choice among variants could be performed on any other basis.[11] Thirty-five years later, in

his Prolegomena for the Oxford Shakespeare (1939), he took the significant step of recommending a limited eclecticism: he asserted that an editor could best approach the goal of reconstructing "an author's fair copy" by using as copy-text the earliest authoritative edition (which would supply spelling and punctuation closer to the author's than a later edition would be likely to contain) and inserting into it the substantive alterations from a later edition (p. 18). But once editorial judgment had determined the presence of an author's hand in a later edition, McKerrow would allow judgment to operate no further, claiming that "we must accept all the alterations of that edition, saving any which seem obvious blunders or misprints." He was clearly still under the spell of the "best-text" fallacy; and it was at this point that Greg stepped in to observe that alterations in a later edition may come from various sources and that it is essential to make discriminations among them.[12]

Greg's primary purpose in "The Rationale of Copy-Text" was to provide a sound argument for greater editorial freedom of choice.[13] "I am only concerned," he unambiguously proclaimed, "to uphold his [the editor's] liberty of judgment" (p. 386); at another point he called it "disastrous" to "curb the liberty of competent editors" (p. 388). He asked why, if judgment was to be admitted in distinguishing possible from impossible readings (as it was in the "best-text" approaches), "should the choice between possible readings be withdrawn from its competence?" (p. 381). The judgment of an editor, he answered, "is likely to bring us closer to what the author wrote than the enforcement of an arbitrary rule." Curiously, however, Greg's strong endorsement of editorial freedom in regard to substantive variants was not extended to what he called "accidentals" (spelling and punctuation), for it was the function of the copy-text, in his view, to provide the accidentals. He made a clear distinction between the nearly automatic acceptance of copy-text accidentals and the use of judgment in dealing with substantives: "the copy-text should govern (generally) in the matter of accidentals,

but . . . the choice between substantive readings belongs to the general theory of textual criticism" (pp. 381-382). He was well aware of the fact that, if one considered it possible to evaluate the authority of accidentals, there would be no need to designate any text as a "copy-text" possessing presumptive authority. As he said in his sketch of the history of textual criticism at the beginning of the essay, "So long as purely eclectic methods prevailed, any preference for one manuscript over another, if it showed itself, was of course arbitrary" (p. 374). The "purely eclectic methods" he referred to were those founded on personal taste; but the point would be equally valid for any approach in which all choices among variants could be settled through the exercise of some kind of judgment. Greg's acceptance of McKerrow's idea of a copy-text, therefore, was founded on a belief that there was usually insufficient evidence for reasoning about accidentals.

It seems evident, nevertheless, that Greg was not entirely comfortable with the idea of restricting the role of judgment in any aspect of the editorial procedure. He inserted the word "generally" in his directive that "the copy-text should govern (generally) in the matter of accidentals"; and he insisted that the copy-text should not be "sacrosanct, even apart from the question of substantive variation," enumerating instances in which it is "within the discretion of an editor" to alter copy-text accidentals. He even went so far as to say that, in regard to "graphic peculiarities" (by which he meant some practices of spelling and punctuation), "the copy-text is only one among others" (pp. 385-386). If, therefore, copy-text accidentals may be altered whenever one believes there is good reason to do so, just as copy-text substantives may certainly be, the role of the copy-text turns out to be that of supplying readings (of both substantives and accidentals) whenever there seems no other basis for deciding. Greg would never have insisted that any reading should be retained simply because it was present in the copy-text, if an editor's informed judgment pointed to a different choice; such "tyranny of the copy-text" (p. 382)[14] was what he was striving to eliminate from textual criticism. Thus if one is to fall back on the copy-text (for accidentals as well as substantives) only when there is no other way to choose, the key element in his copy-text procedure is determining when two readings are in fact "exactly balanced" (p. 386)—or, to use the more famous term that he also employs, "completely indifferent" (p. 387). If, in an editor's

view, there are no completely indifferent alternatives, then there is no need for a copy-text.[15]

Greg himself was, however, somewhat tyrannized by the idea of copy-text, for his essay includes this statement: "whenever there is more than one substantive text of comparable authority, 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 as far as substantive readings are concerned" (pp. 384-385). Thus at the very moment when he emphasizes the necessity for freedom of editorial judgment in regard to substantive variants from documents of equal "extrinsic" (that is, genealogical) authority, he places a mechanical, unreasoned restriction on judgment applied to the accidentals. When the surviving texts of a work form an ancestral series and a copy-text is chosen for its position in the series, there is a justification for falling back on the copy-text when the variants seem indifferent. But when the documents do not form such a series, and when two of them seem evenly balanced in authority, there is no reason to give more weight to the accidentals in one; the fact that there is often little basis for making decisions about accidentals does not in such an instance justify assigning authority to the accidentals that happen to be in a single document. Greg criticized editors for "abdicating" the "editorial function" if in the case of substantives "of equal extrinsic authority" they relied "on some arbitrary canon, such as the authority of the copy-text" (p. 384). But he was giving them contrary advice for the accidentals.

It was after this passage, however, that he sanctioned the alteration of copy-text accidentals—presumably in both kinds of situations, both where texts can be ranked on genealogical grounds and where two or more authoritative texts appear unrankable. Thus the overriding point is the necessity for editorial judgment, which must operate regardless of the relationships among the documents; and the idea of a copy-text (feasible for ancestral series but meaningless for texts of equal authority,

which offer no basis for generalizing about their readings as a whole) is not really central to the argument. Although Greg wished to warn editors about "the mesmeric influence of the copy-text" (p. 383), he did not entirely escape it himself, for his injection of the concept into his discussion ironically interfered with the full expression of a theory of editorial freedom for scholarly editing.

III

The voluminous commentary engendered by Greg's essay has largely been concerned with the applicability of his approach beyond the field of Renaissance drama and with the appropriateness of concentrating on final authorial intention.[16] Most commentators have spent less time scrutinizing the general principles underlying Greg's recommendations than evaluating how useful those specific recommendations seem in different contexts, and thus there has been almost no questioning of the necessity for a concept of copy-text. The basic meaning of the term "copy-text" has remained stable from McKerrow's time onward—that is, the documentary text used as the basis for a scholarly edition. But the way in which a copy-text is selected and emended has undergone a remarkable shift during the twentieth century; and when the phrase "copy-text editing" is now used it signifies not what McKerrow would have understood by it but generally something close to what Greg meant by it. Being a critic of "copy-text editing" —and there have been many such critics in recent years—is likely to mean disapproving of the elevation of texts as completed by authors over texts as they emerged from publication or theatrical production, or objecting to the preference for the author's final intention rather than an earlier intention, or indeed decrying the practice of eclectic editing itself. But criticizing "copy-text editing" has not meant attacking the idea of copy-text. Whether or not editors of post-medieval writings have been successful in avoiding the "tyranny of the copy-text" as envisaged by Greg, they have not escaped the tyranny of the concept of copy-text.

Perhaps the most instructive example is Fredson Bowers, since he was the most prolific and influential editor of this century in the English-speaking scholarly world. He was also the person primarily responsible for the extension of Greg's ideas to the editing of post-Renaissance literature, and therefore one might think it unreasonable to expect that he would have been critical of "copy-text." Yet he was an ardent believer

in the importance of critical judgment in editing, and he did not hesitate to point out what he saw as limitations in Greg's rationale; thus his failure to question the need for a copy-text does show what Greg had called the mesmerizing power of the concept. As cases in point, one may turn to two of his essays from the 1970s, "Multiple Authority: New Problems and Concepts of Copy-Text" (1972) and "Greg's 'Rationale of Copy-Text' Revisited" (1978),[17] which are among the most trenchant analyses that Greg's ideas have received and constitute Bowers's principal reconsideration of his earlier, less measured, response to the "Rationale."

The first of these essays, after sketching Greg's contribution[18] and some of Bowers's own applications of Greg's rationale to later literature, takes up instances of thoroughgoing authorial revision, a subject not dealt with by Greg in much detail. In such cases, Bowers says, when "the two texts are parallel enough for a comparison of the accidentals to be pertinent," an editor "can make some essay at treating the accidentals of the revised edition on the same critical basis as its substantives" (p. 462). What the editor must not do, Bowers insists, is "to succumb to the tyranny of the copy-text." If all variants—accidentals as well as substantives—are treated critically, Bowers recognizes that "in an ideal state" the editor "would arrive at compatible results without regard for the choice of copy-text" (pp. 462-463):

A major element in this "practical convenience" turns out to be conciseness of the apparatus. Actually the choice of copy-text would not necessarily affect the length of the apparatus, though it could frequently do so if one were using the style of apparatus Bowers generally employed. But the main point is that if a shortening of apparatus is accomplished by presenting less information (as it is in Bowers's system), one is in the odd position of claiming that the purpose of selecting a copy-text is to withhold textual evidence from the reader.[19] In fact Bowers proceeds to point out, both in a footnote on this passage and in the following paragraph, that only in an ideal world would the choice of copy-text not affect the resulting edited text—but in that case the choice is not simply "motivated by practical convenience" after all, since the text is affected by it. Yet at the end of the paragraph, Bowers comes back to the matter of convenience, saying that a revised edition should be chosen as copy-text "only in cases of the sternest necessity when to select an earlier document would pile up lists of emendations of staggering proportions" (p. 464). Exactly what function a copy-text plays is finally unclear in this discussion, but the idea that there must be a copy-text of some sort seems never in doubt.

In the second essay, six years later, Bowers gives Greg's "Rationale" a more critical reading, examines its applicability to authorially revised editions of later periods (in the light of his own further experience), and finds that he has a "serious quarrel" with Greg's belief in "the lesser authority of a revised edition in cases of doubt" (p. 123). Although he is convinced that Greg's idea of choosing a copy-text for its accidentals is "still (and no doubt invariably) sound . . . for any period" (p. 147), he concludes that the greater accuracy of compositors in periods after the Renaissance means that indifferent substantive variants in revised editions of the last three centuries may generally be regarded as more likely to be authorial than compositorial (p. 155). There is an irony here in Bowers's partial return to McKerrow's belief in the whole body of substantive variants of a revised edition. Bowers's position is more flexible, but there is considerable similarity between McKerrow's idea of adopting all the variants when some are certainly authorial and Bowers's newly formed conviction of the likelihood "that an indifferent variant in the revised text is authorial" (p. 155). Bowers even admits the possibility that this view may apply to accidentals as well as substantives (depending on one's judgment of compositorial fidelity) and thus that the revised text could become the copy-text (p. 160). But given his recognition that

copy-text accidentals as well as substantives can be altered by the editor (e.g., pp. 128-129, 148-149), along with his willingness to make distinctions even among what he labels "indifferent" variants (rendering them in fact no longer indifferent),[20] one begins to wonder what is left of the concept of a copy-text as a fall-back position. That this concept is the ostensible focus of the essay reveals its tyrannizing powers, for the real subject is—as it must be in an approach that stresses judgment—the process of active decision-making. Bowers was willing—as he put it at the end of the essay—to "reverse Greg's principle for any period after the Renaissance"; but he was apparently not ready to demote the idea of denominating one text as the copy-text in every instance.

This reluctance is most dramatically revealed in his treatment of so-called "radiating texts"—texts that do not form an ancestral linear series but instead represent independent lines of descent from a common source. This kind of textual history is frequently encountered in the study of the manuscript traditions of ancient texts, but Bowers's attention was drawn to it by his editing of Stephen Crane, whose syndicated newspaper pieces provide a classically pure example of the problem: the various original newspaper texts of a piece are independently derived from (and equidistant from) the master copy supplied by the syndicate office to the subscribing newspapers. In the absence of such master copy, the syndicated appearances—"radiating" from the lost original—constitute the only evidence for reconstructing what the syndicate office sent out (which is, in turn, at least one step removed from Crane's lost manuscripts). In the fifth and sixth volumes of his Crane edition (Tales of Adventure and Tales of War, both 1970), Bowers first dealt with this situation extensively.[21]

His procedure, at points of variation in punctuation among the radiating texts, was generally to adopt the reading present in the largest number of them (though he recognized that not every instance could be handled on a quantitative basis); and then he selected as a copy-text the single newspaper text that contained the largest number of the readings that he had decided to adopt. As he says in discussing "The Pace of Youth," the first of the pieces in Tales of Adventure, the Dayton Daily Journal text "has been selected as copy-text for its general concurrence with majority opinion" (p. cxlii). This use of "copy-text" of course shifts the meaning of the term from what Greg (and Bowers in earlier discussions)

had meant by it—for here it does not designate a text that can be accorded presumptive authority by virtue of its genealogical position, since all the radiating texts are equidistant from the copy furnished by the author. Instead of a copy-text that helps one to resolve cruxes, it is now one that is selected after the cruxes have been resolved. The problem emerges even more clearly in Bowers's account of the next piece, "One Dash—Horses," where he designates the syndicate master proof as the copy-text (even though it does not survive), because it is "the earliest recoverable archetype," and he is then driven to use the term "physical copy-text itself" for the one newspaper text that is "perhaps the closest in its accidentals to the lost master proof" (p. cxlvii). Tales of War also includes instances in which Bowers, after pointing out that several newspaper texts are "technically" of equal authority, chooses one as copy-text for its "convenience"—the convenience stemming from the fact that the text chosen (to quote his statement in one case) "is generally less in need of accidentals emendation" (p. cxi).[22]

The conceptual imprecision in this shifting use of "copy-text" probably had little effect on Bowers's final text, since his copy-texts in these instances were chosen not for their authority but for the extent of their agreement with what he had already decided the text should contain. Nevertheless, the fuzziness surrounding "copy-text" here does have practical consequences for the reader, through its effect on the apparatus. Because the list of emendations provides a record of editorial alterations made in the designated copy-texts, and because the "historical collation" records only substantive variants, the variants in accidentals among the radiating texts are not fully reported. Bowers does say, in his headnote to the emendations list in both of these Crane volumes, "when as in syndicated newspaper versions a number of texts have equal claim to authority, all are noted in the Emendations listing" (5:211; 6:335). What this statement means is that a full record of variants in accidentals is normally[23] provided for each place where the designated copy-text is

emended; but any variants in accidentals at places where the copy-text is not emended remain unreported. Yet the documentary authority of those variants is the same as that of the reported ones, given the genealogical equality of all the radiating texts. Thus the reader is deprived of some of the evidence from the texts that collectively form the primary authority in these cases. That Bowers himself sometimes thought of a group of radiating texts as a kind of collective copy-text is suggested by an instance in "One Dash-Horses" where he retains the designated copy-text's capitalization of "North" but says that it "may most properly be regarded as an emendation" (p. 199)—and then includes it (with "stet") in his list of emendations (p. 213). In short, the importation of a concept of copy-text into these situations is more productive of confusion than of clarity.

Bowers's primary theoretical discussion of radiating texts occurs not in the Crane edition but in the latter half of his essay on "Multiple Authority," where it constitutes his principal supplement to Greg—since it, unlike most of his other treatments of copy-text, focuses on a situation that Greg did not attempt to cover. But his discussion, pioneering as it is in many ways, is weakened by his retention of the idea of a copy-text, especially after having in effect shown why it is irrelevant. He emphasizes, quite properly, the role of critical judgment, pointing out that for radiating texts it must be applied as fully to accidentals as to substantives.[24] A statistical approach, he rightly asserts, is not sufficient, for an unconventional authorial reading might be normalized by many compositors and "preserved only by the dogged or indifferent few"; "the minority reading may sometimes need to be adopted on critical grounds" (p. 468). This cogent reasoning implies the importance of recording all the variant readings in radiating texts; and it is therefore revealing of his divided mind that he proceeds to advocate choosing a copy-text "requiring the least amount of apparatus" (p. 471).[25]

He recognizes the reasons for dispensing with the concept of copy-text altogether in such situations: if there are eight radiating texts, he says, the copy-text is "only one of eight equal witnesses, of no greater presumptive authority than the other seven" (p. 470); the choice of copy-text, then, is "theoretically indifferent" and "largely a question of convenience"; it is made only after the critical text has already been constructed (p. 471), and it is "not integral to the editing" (p. 485). He is therefore receptive to the possibility of having no copy-text; yet he maintains that a copy-text be chosen "to serve as the physical basis" (p. 471) for a critically reconstructed text, without making clear why the "physical basis" must be equated with a single document, when in fact it is a group of documents. The idea that there must be a copy-text was so firmly rooted in his mind that the only alternative he saw to selecting a single radiating text was to have a "non-extant copy-text" (p. 471), in Crane's case the lost syndicate master proofs as reconstructed from the radiating texts. To cling to a concept of "copy-text" even if it comes to mean the same thing as the scholar's reconstructed text is surely the ultimate in being tyrannized by the idea of having something called "copy-text." At the end of the essay Bowers says that to try applying Greg's rationale to radiating-text situations "would establish a real tyranny of the copy-text"; but Bowers (like Greg before him) was not able, even while warning others about this tyranny, to shake himself entirely free from its bonds.

IV

Two years after Bowers's essay, I called attention to the inappropriateness of trying to employ in radiating-text situations the kind of apparatus that had become standard, under Bowers's influence, for use with linear genealogies. Although he recognized that Greg's "Rationale" did not apply to radiating texts, he apparently could not bring himself to accept fully the fact that an apparatus reflecting this rationale would not do so, either. It is noteworthy that when, in "Multiple Authority," he shrank from the notion of a "non-extant copy-text" his stated reason was not conceptual but the belief that "the problem of what form an apparatus

would take for such a non-extant copy-text is acute" (p. 471). Actually, as I tried to show, there is nothing complicated in simply listing all the variants (in both substantives and accidentals) in all the radiating texts and keying this single list to the newly constructed critical text.[26] Such a list emphasizes that the critical text has been built up from the evidence of all the radiating texts and was not produced by making alterations in one of them. The only sense in which the latter operation could be said to have occurred is that some form of reproduction (xerographic, say, or photographic) of one of the texts may have been marked to produce a document that could be handed to the printer of the new scholarly text. Bowers was of course well aware, in other discussions, of the significant difference between a copy-text in Greg's sense and a printer's copy for a scholarly edition.[27] Yet his recommendation to choose a single documentary text from a group of radiating texts, and call it a "copy-text," amounts to equating "copy-text" with "printer's copy"—while at the same time elevating it undeservedly to unique historical status by reporting its variants fully (and not those of any other text).

There is no escaping the fact that radiating texts equidistant from their common ancestor provide no text to serve as copy-text; and it is further true that this lack does not prevent an editor from constructing

a critical text—as Bowers did before he chose a copy-text—or from preparing an apparatus recording all the rejected variants. Once one accepts these points, one may begin to see that the thinking involved in radiating-text situations is applicable to all critical editing. Every choice made among variants in radiating texts is an active critical choice; no reading is settled on by default, for there is no text that offers a fall-back position. When the variants in radiating texts seem "indifferent," an editor may of course choose a reading from the text that supplies the largest number of other readings; but the decision is still an active one, in which one of the factors taken into account is the apparent general reliability of a particular text. The process remains one of building up a new text rather than making changes in an old one. If this idea—that critical editing is constructive rather than emendatory—were also applied to texts in linear genealogies, the role of judgment might more clearly be seen as dominant, and any practical guideline (such as Greg's rationale) might be better recognized as an aid to judgment, not a brake on it. An editor, for example, who is following Greg in according presumptive authority to the text closest to a lost manuscript, would—following this plan—speak not of "retaining" readings from that text but of "selecting" them. The notion that wisdom supports the idea of sticking to a copy-text when two readings are indifferent places the emphasis on the preservation of a documentary text; but the idea that a critical text emerges from active choices made among the variant readings (along with, of course, the editor's own inventions, when necessary) emphasizes editorial judgment.

The difference between the two may at first seem slight, a matter of labeling: in the former, one lets a copy-text reading stand if the variant is indifferent and there is thus no compelling argument for altering it; in the latter, two readings that might otherwise be indifferent are not actually so, for the fact that one of them comes from a text of superior genealogical standing provides a reason for choosing it. But the difference between these two justifications for the same decision is not superficial: it goes to the heart of what critical editing is. The key point is not whether an editor would make the same decision by following Greg's rationale or by designating no copy-text but still following Greg's argument for the presumptive authority of the text closest to an authorial manuscript. No two editors can be expected to make the same choices by following either of these systems in any case. The important point is that the former approach places a rule above reason (as any recourse to a fall-back position must do), whereas the latter restructures the problem so that the editor's decision (even if it is the same decision) results from the positive step of taking a reasoned action.

The controlling images of the two approaches are those of initially

full and initially empty sheets of paper. If one chooses a copy-text, then in effect one begins with filled sheets and proceeds to alter the text present on them; but if one has no copy-text, one begins with blank sheets, so to speak, and fills them by placing one word after another on them, drawing those readings from the relevant documents (and, on occasion, from one's own mind). I am not suggesting, of course, that an editor should actually write out a text in this way; but I do believe that an appropriate analogy for the critical editor to have in mind is the medieval scribe taking readings from various manuscripts as he prepared a new manuscript.[28] It is ironic that in classic Lachmannian textual criticism this scribal practice is called "contamination," for some such process (supplemented by the editor's own inventions) provides our only hope for rising above the limitations of individual documents. Although the followers of Lachmann may often have deluded themselves about the objectivity of their own methods, their process of "recension," with its aggregating impulse, in many cases avoided the pitfalls of the "best-text" family of approaches and therefore holds a lesson for editors of works in the modern languages.

The constructive approach I am outlining subsumes all the various points of view that can be taken toward the goal of editing. There is no reason, for example, why an editor interested in uninfluenced authorial final intention could not still follow Greg's rationale—which has proved itself effective for this purpose. But instead of treating one text as a copy-text, an editor would use the genealogical position—and thus the presumed authority—of that text as a factor in weighing each variant reading. Sometimes this factor would be decisive; sometimes other factors would be. The difference between this procedure and the conventional one is subtle but crucially significant. Genealogy is taken into account, but with judgment clearly in the dominant position. If, instead of uninfluenced authorial intention, one preferred to emphasize, say, the text that was the joint product of the author and the publisher's staff, one would then have a different attitude toward some of the first-edition readings that vary from manuscript readings (those, for instance, that seem to reflect house-styling). Or if one wished to focus on a stage that preceded a final version, the bulk of the text might be drawn from the document that provides the best evidence of the existence of such a stage,

but not without a serious weighing of the claims of variants in other documents (since versions of works cannot be equated with the texts of particular documents). What I am suggesting is not a supplement or an alternative to Greg's rationale but an overarching framework that encompasses its goal and that of other rationales as well. Obviously one could rewrite Greg so as to focus on goals other than authorial intention; but the resulting series of copy-text rationales, each aiming at a different end-product, would still not avoid the restriction on editorial judgment inherent in Greg's rationale. What is needed instead is a framework that liberates editorial judgment from the concept of copy-text while being neutral in itself as to the goal toward which that judgment should be directed.[29]

Despite the salutary emphasis of Greg and Bowers on editorial freedom, many editors are still—as Greg and Bowers were—in thrall to the notion, now about two centuries old, that responsibility in scholarly editing is, at least to some degree, incompatible with freedom of judgment. A passage in Bowers's "Multiple Authority" illustrates how inhibiting this attitude can sometimes be. Bowers claims that the evidence of the radiating newspaper texts of a Crane story enables one to attempt reconstructing the syndicate master proofs but offers no justification for pushing on back to Crane's manuscript. It is not the editor's concern, he says, "whether in their recovered form these proofs agree or disagree with Crane's habits of punctuation, spelling, and so on"; and he continues, more emphatically, to say that "it is not an editor's business to print what he may be morally certain the manuscript reading would have been when the evidence indicates strongly that the recovered proof read otherwise" (p. 473).[30] There may be good reason, of course, to be satisfied with having the text of the syndicate proofs; but if one is really interested in what Crane wrote in his manuscript, and if one's knowledge and judgment make one "morally certain" of being able to reconstruct it, why should one be prevented from doing so by the fact that one is going back two or more steps behind the preserved documents rather than just a single step? The "law" of "documentary evidence," to which Bowers appeals, is surely misapplied if it outlaws the responsible use of the historical imagination. The very existence of critical editing depends on recognizing that documentary texts may legitimately be overruled by

informed judgment. Whether an editor is justified in attempting to reconstruct any given stage in a text's history depends on how the task is approached, on how the editor proposes to overcome the limitations of the documents; no a priori ruling can decree one stage to be a valid goal for the exercise of historical judgment and another to be inappropriate.

On many other occasions, however, Bowers not only granted, but openly welcomed, the dominance of judgment—as when, in his edition of Tom Jones (1975), after saying that the textual situation was one in which "Greg's classic theory of copy-text must hold" (p. lxx), he declared that the operation of emendation "is a critical process almost exclusively" and that in such a process "the editor shoulders his proper responsibility" (p. lxxi). In a 1985 address he described Greg's rationale as "a discretionary principle, to be applied flexibly," and as a "liberation" from "mechanical conservatism," complaining that in America it was often used to justify "avoiding the unknown terrors of eclecticism."[31] Both Greg and Bowers unquestionably believed in the liberty of editorial judgment, but in their procedural statements they yoked this belief to a strategy that sprang ultimately from a contrary view, for they obviously carried with them just enough of an inherited distrust of judgment to make them not quite prepared for completing the long historical movement toward the full reinstatement of critical judgment in editing.

What I am proposing here is a way to take that step without abandoning the responsibilities of scholarship. It might be called "constructive critical editing" to distinguish it from an approach that emphasizes emendation. To see critical editing as an activity of rebuilding rather than repairing forces the judgment to play its central role in recovering the past. All historical reconstruction requires judgment to enable one to decide what can be accepted as facts and what can reasonably be inferred from them by an informed imagination. Experiencing verbal works as communications from the past entails this kind of reconstruction not only because they are past events but also because they employ an intangible medium, language. Reading necessarily involves the use of judgment in the extracting of a work from a document. If editors' readings, enshrined in editions, are to be exemplary, they must arise from an active embracing of judgment—which is, after all, the only thing we have to rely on.

Notes