University of Virginia Library

The Editorial Problem of Final Authorial Intention
by
G. Thomas Tanselle [*]

Scholarly editors may disagree about many things, but they are in general agreement that their goal is to discover exactly what an author wrote and to determine what form of his work he wished the public to have. There may be some difference of opinion about the best way of achieving that goal; but if the edition is to be a work of scholarship—a historical reconstruction—the goal itself must involve the author's "intention." The centrality of that concept to scholarly editing can be illustrated by W. W. Greg's "The Rationale of Copy-Text,"[1] which, in the quarter century since it first appeared, has established itself as the most influential document in modern editorial theory. What Greg succeeded in accomplishing was to provide a rationale for selecting, and then emending, a basic text in those cases in which the choice was not made obvious by the historical,

biographical, bibliographical, and linguistic evidence available. In such instances, an editor requires some guiding principle by means of which he can maximize the chances of adopting what the author wrote and minimize the chances of incorporating unauthorized readings into his text. Greg's now celebrated solution rests on the position that, if a finished manuscript of a text does not survive, the copy-text for a scholarly edition should normally be the text of the earliest extant printed edition based on the missing manuscript, for it can be expected to reproduce more of the characteristics of the manuscript than any edition further removed; variants from later editions which are convincingly shown to be revisions by the author can then be incorporated into this copy-text. Because authors who revise their work do not always give as much attention to what Greg calls "accidentals" (matters of spelling and punctuation) as to "substantives" (the words themselves)—and because such attention is in any event extremely difficult to determine—the copy-text usually remains the authority for accidentals; and if an editor adopts as authorial certain substantive variants in a later edition, he need not adopt all the other variants in that edition. Following this plan, the editor has a rational means for deciding among indifferent variants (he retains the copy-text readings); and the resulting critical text should be closer to the author's intention than any individual surviving form of the text.

Although Greg did not address himself to the question of a precise definition of "author's intention," it is clear from such a summary that he considered the goal of an edition—and he was speaking of an "old-spelling critical edition"—to be the reconstruction of a text representing the author's final wishes about the version of his work to be presented to the public. In Fredson Bowers's words, the task is "to approximate as nearly as possible an inferential authorial fair copy, or other ultimately authoritative document";[2] or, as he put it another time, following Greg's theory will produce "the nearest approximation in every respect of the author's final intentions."[3] If an author can be shown to have gone over his work with scrupulous care for a revised edition, examining accidentals as well as substantives, the revised edition (as the closest edition to an "ultimately authoritative document") would become the copy-text. Such a situation does not arise in most

instances, but Greg recognized its importance: "The fact is," he said, "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" (p. 390). In other words, an editor cannot avoid making judgments about the author's intention on the basis of the available evidence; the strength of those judgments, in turn, will depend on his historical knowledge and his literary sensitivity.[4] The job of a scholarly editor, therefore, can be stated as the exercise of critical thinking in an effort to determine the final intention of an author with respect to a particular text.[5]

Just what is meant by "author's final intention," however, has not been made entirely clear, although at first glance the concept may seem so self-evident as not to require formal definition. Its use in connection with editing suggests that an editor's task is not to "improve" upon an author's decisions, even when he believes that the author made an unwise revision, and that an editor's judgment is directed toward the recovery of what the author wrote, not toward an evaluation of the effectiveness of the author's revisions.[6] Furthermore, the concept, as a goal of editing, would seem clearly to imply that, when an editor has strong reason to attribute a revision to the author, he will accept that revision as "final" on the grounds that, coming second, it represents the author's considered and more mature judgment. Greg suggests that this procedure is equally valid for dealing with wholesale revision when he writes, "If a work has been entirely rewritten, and is printed from a new manuscript, . . . the revised edition will be a substantive one, and as such will presumably be chosen by the editor as his copy-text" (p. 389).

It is true that, in many instances, the simple interpretation of

"final intention" to mean that intention reflected in the last alterations made or proposed by the author is workable enough and results in no ambiguity as to the aim of the editorial process. Nevertheless, such an interpretation does not answer certain theoretical questions which can assume practical importance in the remaining instances. Two basic kinds of situations particularly require further consideration: cases where the editor must distinguish authorial alterations from alterations made by someone else and must decide what constitutes "authorial intention" at such times; and cases where the editor faces alterations unquestionably made by the author but must still decide which readings represent the author's "final intention." In what follows I shall offer some preliminary comments on these two situations. But it is necessary to begin with at least a brief consideration of the meaning of "intention" for this purpose and with some recognition of the critical implications of attempting to discover "authorial intention."

I

The question of the meaning of "intention," both in general terms and in relation to works of art, involves many complex philosophical issues and has been widely debated.[7] Probably the best-known and most influential discussion of this subject in relation to literary works is W. K. Wimsatt and Monroe C. Beardsley's 1946 essay, "The Intentional Fallacy," which takes the point of view that the author's intention is irrelevant to the process of literary interpretation and evaluation.[8] Although the essay is not as clear as it might be in distinguishing

among kinds of intention, it has become a classic statement of the position that the critic should not be influenced in his examination of the literary work itself by any information relating to what the author thought he was saying in that work. Other writers have argued the opposing view—notably E. D. Hirsch, Jr., whose Validity in Interpretation (1967) supports the position that the meaning of a work is the meaning put there by the author.[9] Discussions of this kind, however, regularly take the text as given and focus on the activity of the critic as he faces that text; they do not raise the question of the authority of the text itself, apparently assuming that the text in each case is the text as the author wished it to be.[10] Of course, a corrupt text could equally well be the subject of critical analysis;[11] but the question of the bearing of authorial intention on interpretation would hardly arise unless the text is assumed to be what its author wished.

One might at first conclude, therefore, that such discussions of intention are irrelevant to editorial work, conceived of as operating at an anterior stage and providing the material for the critic to analyze. If, in other words, one could assert that the editor's task does not involve critical decisions but only the recovery of factual information about what word or mark of punctuation the author wanted to have at each point in his text, one could then say that any effort to understand or assess the "meaning" of the text is an entirely separate matter and that the possibility of an "intentional fallacy" applies only to this interpretive and evaluative activity.[12] It is immediately apparent, however, that the job of the editor cannot be so regarded. If the aim of the editor is to establish the text as the author wished to have it presented to the public (and we shall postpone any consideration of other possible editorial aims), he cannot divorce himself from the "meaning" of the text, for, however much documentary evidence he may have, he can never have enough to relieve himself of the necessity

of reading critically. Suppose, for example, that the only extant text of a work is a fair-copy manuscript in the author's hand. The editor in such a case cannot simply reproduce the text mechanically, without thinking about its meaning: there is always the possibility that the author, through an oversight or slip of the pen, did not write down what he meant to write, and the editor who is reading critically may be able to detect and correct such errors, or at least some of them. It is an act of criticism, however elementary, for an editor to recognize that where the author wrote "the the" he actually meant "to the." In other instances it may be equally obvious that the author cannot have meant what he wrote, and yet it may be impossible to say with certainty which of several possible corrections conforms with what he had in mind. Yet the editor will probably find it necessary to make some correction, since the reading of the manuscript is plainly wrong. When two or more texts of a work exist and there are differences between them, there may be no conclusive evidence to show which differences are the result of the author's revisions and which are not. Yet the editor must decide which of the readings to accept at each point of variation. These decisions are based both on whatever external evidence is available and on the editor's judgment as to how the author was most likely to have expressed himself at any given point. This judgment in turn is based on the editor's familiarity with and sensitivity to the whole corpus of the author's work and on his understanding of the individual work involved. He may be specifically concerned only with the author's intended meaning in one sentence, or even one phrase, but the interpretation of that sentence or phrase may depend upon the author's intended meaning in the work as a whole.

It is clear, then, as soon as one starts to talk about "intention," that various kinds of intention need to be distinguished, and many of the recent discussions of intention in literature do attempt to subdivide the concept. Thus T. M. Gang differentiates between "practical intention" (intention "to achieve a certain result") and "literary intention" (intention to convey "a certain significance"); John Kemp distinguishes between "immediate intention" (that which a man "intends, or sets himself, to do") and "ulterior intention" ("that which he intends or hopes to achieve as a result of doing what he does"); Morse Peckham discriminates between "mediated intention" ("a statement or other sign") and "immediate intention" ("metaphorical extension of mediated intention into the area of 'mind'"); and Quentin Skinner, borrowing terms from J. L. Austin's How to Do

Things with Words (1962),[13] speaks of "illocutionary intention" (what a writer "may have been intending to do in writing what he wrote") and "perlocutionary intention" ("what he may have intended to do by writing in a certain way"), as well as of "intention to do x" (a writer's "plan or design to a create a certain type of work").[14]

Of such classifications of intention, one of the clearest and most useful has been set forth by Michael Hancher.[15] In his view, "author's intentions" can be divided into three types: (1) "programmatic intention"—"the author's intention to make something or other"; (2) "active intention"—"the author's intention to be (understood as) acting in some way or other"; and (3) "final intention"—"the author's intention to cause something or other to happen." The first refers to the author's general plan to write a sonnet, say, or a realistic novel;

the third refers to his hope that his work will change the reader's viewpoint, say, or bring wealth to himself. The second is the one which concerns the meanings embodied in the work: "Active intentions characterize the actions that the author, at the time he finishes his text, understands himself to be performing in that text" (p. 830).[16] Hancher's argument is that the first and third kinds of intention—programmatic and final—are indeed irrelevant to the interpretation of a literary work but that the second—active intention—must be taken into account in the interpretation (and evaluation) of the work.

Before pursuing the implications of that argument, we should pause long enough to note that what editors in the tradition of Greg are likely to call "final intention" does not correspond to what Hancher here calls "final intention." Rather, the intention with which editors are concerned is Hancher's "active intention," the intention that the work "mean (and be taken to mean) something or other" (p. 831). The fact that an editor, as briefly suggested above, must examine both the author's intention to use a particular word and the author's intention to mean a particular thing in the work as a whole—indeed, must make decisions about the first in the light of the second—is adequately accommodated in Hancher's concept of "active intention." Hancher's initial illustration of the concept includes comment on Hopkins's intention in using "buckle" in "The Windhover" as well as on his intention for the meaning of "the whole action of that poem." Because an intention regarding the meaning of a work as a whole may not always seem distinct from a programmatic intention, Hancher later recognizes that a programmatic intention may "involve a kind of active intention" (p. 836) but distinguishes between such "projected active intention" and the "active intention that ultimately defines the meaning of the completed text" (p. 837). Therefore "active intention," as he defines it, does cover the authorial intentions with which an editor has to deal. Accordingly, whenever I speak of "intention," unless otherwise specified, I am referring to the kind of intention included in Hancher's concept of "active intention"; even when I use the term "final intention," in conformity with the common practice of editors, the word "intention" still refers to the same concept (and not to what Hancher calls "final intention")—though just what "final" may mean in the phrase remains to be examined later.

It can readily be inferred from what I have already said about Hancher's conclusion that he does not belong to that group of critics

who believe in "semantic autonomy" (to use Hirsch's phrase)—the idea that a verbal construction carries its own inherent determinate meaning regardless of what meaning was intended by the author. As Beardsley puts the idea, "texts acquire determinate meaning through the interactions of their words without the intervention of an authorial will."[17] Hancher's argument for the relevance of active intention to literary interpretation shows that he defines "the" meaning of a work as the meaning intended by its author.[18] It is difficult to refute such an argument without taking the position that the language of a literary work operates in a different way from the language of ordinary discourse; yet that position cannot convincingly be taken so long as it is impossible to draw a distinct line between works which are literary and works which are not.[19] Language, after all, consists of symbols, which must be invested with meanings if they are to mean anything. At the same time, a reader does not have access to an author's mind, and, if he understands a text to mean something, it is (at least to begin with) as a result of certain conventions of language which both are following.[20] Yet texts (or utterances) do not have to be complex in order for the conventions involved to be capable of more than one interpretation. The possibility is raised, therefore, that the meaning or meanings a reader finds in a text do not correspond to the meaning or meanings which the author intended.[21] To reject "semantic autonomy" (or "immanent meaning") is not to deny that texts are capable of multiple interpretations. Indeed, the fact that multiple interpretations are possible is a refutation of the idea that a text embodies a determinate meaning.

How, then, is the author's intended meaning to be discovered? In answering that question, one is inevitably drawn back to the work itself as the most reliable documentary evidence as to what the author intended. If he made no statement setting forth his intention, one has

nowhere else to go for direct evidence (though of course one can take into account various historical and biographical circumstances); and if he did make a statement, it may, for a great variety of reasons, not be accurate. As Morse Peckham has pointed out, any attempt, by the author or someone else, to explain the intention of a work ("an utterance") constitutes an inference about an event which took place in the past; the author's account carries greater weight "only because he is likely to have more information for framing his historical construct, not because he generated the utterance."[22] Furthermore, as William H. Capitan has noted, "what an artist gives us as his intention is subject to the artist's limitations in putting his intention into words."[23] The position has been well stated by Quentin Skinner: Hirsch, who does not recognize as "meanings" any meanings other than the author's, decides what Wordsworth "probably" meant in "A slumber did my spirit seal" by turning to "everything we know of Wordsworth's typical attitudes during the period in which he composed the poem" (p. 239). Contemporary statements about these attitudes may of course be useful, but are not Wordsworth's poems the chief source of information about his attitudes? And if a given body of writings can provide such evidence, is it not possible that a smaller body of writings—or even the single poem—could provide it? As Hirsch admits, "A poet's typical attitudes do not always apply to a particular poem" (p. 240); so one is inevitably thrown back on the poem itself. I do not think it necessary to review here the various arguments for or against "semantic autonomy"[25] in order to make my point: all I am suggesting is that a rejection of the notion of "semantic autonomy" is not incompatible with the position that the work itself provides the best evidence of the author's intended meaning.

The bearing of these ideas on the task of the editor is worth making explicit. At the start, the editor has settled one important question through his definition of his goal: he is concerned with establishing the text as intended by the author, and thus he has no doubts about the relevance of the author's intention to his undertaking. But then he recognizes that the most reliable source of information about the author's intention in a given work is that work itself. He will take other information into account, but he must always measure it finally against the very text which is the subject of his inquiry. The editor may at first feel that his job is different from the critic's in that he is concerned with establishing intended wording, not with explicating intended meaning. That is, he may think (in Austin's terms) that he is dealing only with the author's locutionary act, not his illocutionary act. But he soon realizes that his discovery of textual errors or his choice among textual variants involves his understanding of the intended meaning of the text. For if either of two alternative words makes sense at a given point, the determination of which the author intended clearly involves more than his locutionary act.[26] Greg's

rationale tells an editor what to do when he is at an impasse, but it does not eliminate the need for judgments; indeed, it relies on them. So the editor finds himself in the position of the critic after all. Merely because he has already decided that his concern is with the author's intention does not mean that the issue of "semantic autonomy" is irrelevant to him, for he, too, will be turning to the text itself as his primary evidence.

The key to the use of the work as evidence of its author's intended meaning must lie in the approach which the critic (editor) takes.[27] One critic may believe that he has found through internal evidence the most satisfactory explication of a work and may not be interested in whether or not this meaning was intended by the author.[28] Another critic, who wishes to find the author's intended meaning, will read the work in the light of all the historical and biographical evidence he can locate and may thereby eliminate certain meanings as ones which the author could not have intended;[29] his interpretation of the text is thus limited by certain external information, but his positive evidence still comes from the text itself. If I hastily dash off a message for someone and the recipient finds it ambiguous, he will attempt to rule out certain meanings on the basis of what he knows of me and of the circumstances which occasioned the message; what he finally concludes to be my intended meaning, however, cannot be based merely on what the external evidence suggests I would be likely to say in that situation but must rest on the words I actually did use. Furthermore,

if I mistakenly wrote one word while intending another, it may be that the external evidence would cast doubt on my use of that word, but any correction of the error would have to be justified by the context. The work itself is the controlling factor in statements made about its meaning, whether or not those statements aim at elucidating the author's intended meaning.

The scholarly editor is in the same position as the critic who is concerned with the author's intended meaning. Regardless of how many meanings he finds in the text, the scholarly editor makes corrections or emendations on the basis of the one he judges most likely to have been the author's intended meaning. Hancher speaks of a science of interpretation, in which the critic's aim is to determine the "authorized" or intended meaning, and an art of interpretation, in which the critic's aim is to find the most satisfying meaning according to his own "norms of value."[30] Some critics would protest that all meanings are part of an intended complex combination of meanings, intended in the sense that the author, whether consciously or unconsciously, created a structure in which they could be discovered. But this appeal to "subconscious intention," as T. M. Gang has indicated,[31] implies a universal set of relationships between consciously produced patterns and their subconscious origins—relationships which, if they are always in operation, cannot be specifically intended and are always available for anyone to discern. One need not deny that such meanings can be found in the work and that they may be valuable (and offer revealing insights into the author's personality and motivation) in order to believe that "intention" cannot be a useful concept if it is made so broad as to cover all potentially discoverable meanings. It is certainly true that neither the author nor anyone else can construct an explanatory paraphrase which is the exact equivalent of the work itself; but

it does not follow that the intended meanings of the work are inexhaustible.

An editor could, of course, emend a text so that it would, in his view, be a more successful expression of that meaning which he finds most valuable in it; but his activity would have nothing to do with the author. In Hancher's terms, he would be engaging in the "art" of interpretation—or, rather, editing on the basis of that kind of interpretation. A scholarly editor sets as his goal the reconstruction of the text intended by the author. In Hancher's terms, he is engaging in the "science," not the art, of interpretation—but it is still interpretation and entails critical thinking. His defined approach is what controls the use he makes of what he finds in the work. He will probably find more than one meaning there, but his specialized knowledge places him in a privileged position for assessing which of them can most reasonably be regarded as the author's.[32] The text he produces can, like any other text, be the subject of critical speculation by those who have no interest in the author's intention; but it can also serve the needs of those critics who are concerned with the work as the product of a particular mind. That an interpretation by one of the former turns out to seem more satisfying to many readers than an interpretation by one of the latter has no bearing on the importance or desirability of the task which the scholarly editor has set himself.

These considerations suggest, first of all, that editing is a critical activity and that the scholarly editor cannot avoid coming to terms with the critical problem of authorial intention. Second, there is a specific and clearly defined aspect of the broad concept of "intention" which is the appropriate concern of the scholarly editor—the intention of the author to have particular words and marks of punctuation constitute his text and the intention that such a text carry a particular

meaning or meanings. Finally, the scholarly editor will amass all the evidence he can find bearing on each textual decision; but, whenever the factual evidence is less than incontrovertible, his judgment about each element will ultimately rest on his interpretation of the author's intended meaning as he discovers it in the whole of the text itself. What controls the editor's freedom of interpretation is his self-imposed limitation: he is concerned only with that intention which his knowledge of the author and the period allows him to attribute to the author.

II

An editor who has given some attention to such preliminary questions has at least begun to reach an understanding of "authorial intention." But there are a number of further questions which arise as he attempts to make judgments in the light of this conception of intention. Perhaps the most common editorial situation is that in which the editor must decide whether a given variant reading is a revision by the author or an alteration (conscious or inadvertent) by someone else. In these cases, at least one intermediate stage of documentary evidence is lacking, and the editor is trying to determine from the surviving material those changes which the author made in that now missing document. He must also face the question whether it is ever possible to think of changes not made by the author as nevertheless fulfilling, or contributing to, the author's intention.

The basic situation can be illustrated by Sherwood Anderson's A Story Teller's Story (1924). The only surviving prepublication text of this work is the typescript printer's copy, which bears revisions by three people: Anderson himself, Paul Rosenfeld, and E. T. Booth, the publisher's editor. One has direct evidence, therefore, for assigning the responsibility for each of these alterations; but the first printing of the book (Huebsch, 1924) contains additional changes, not marked on the surviving typescript and presumably entered on the now lost proofs. Deciding which of these changes were made by Anderson is the central task in editing this book.[33] What the editor has to do is to familiarize himself with all the available relevant evidence—bibliographical, historical, biographical. He may then find that some of it is

convincing enough to dictate certain decisions. For example, on (or just before) 28 October 1924 Anderson wrote to Rosenfeld explaining why he had cut out some material about Waldo Frank.[34] Since several paragraphs about Frank are present in the typescript but not present in the first impression, one can conclude that this is the deletion referred to and that it was made by Anderson on the proofs. But for most of the alterations in the first impression there is no such compelling evidence; most of the editor's decisions must finally be critical judgments, resulting from an evaluation of what evidence there is, from an understanding of Anderson's habits of revision, and from a familiarity with and sensitivity to his style and ideas. Even the deleted Waldo Frank passage leads the editor to a related judgment, for that deletion is only a part of a considerably longer deletion made in the first impression. Because the entire passage concerns Anderson's reactions to various writers, one may conclude that he probably eliminated all of it, and not merely the part about Frank which he happened to mention to Rosenfeld. But that conclusion is a judgment, supported by a critical argument, not by verifiable facts.

The same observations can be made about situations involving variants between printed editions. In these cases the missing documents are the author's marked copy of the earlier edition (or its proofs, or whatever served as printer's copy for the edition set later) and the marked proofs of the later edition. If no document survives which antedates the proofs of the earliest setting of the text, then of course one is dealing with texts which have already been subjected to the routine of the printing- or publishing-house. Thus the essential difference between this situation and the one described above is that here the editor is working at a greater remove from the author's fair-copy manuscript or typescript; but his approach to the problem remains the same. For instance, neither the manuscript of The Rise of Silas Lapham nor the proofs set from it survive; and the history of the early printed texts, which vary from one another substantively at a number of points, is such that one text might contain the later readings in one part of the book and another the later readings in a different part. So for any given variant, the editor must first try to determine the order of the readings and then decide whether the later one could be an authorial revision or correction. At one point in Chapter 19 Irene's complexion is described as "snow-white" in the serialized magazine text and as "colourless" in the first book edition

(set from proofs of the magazine text); since the publication schedule for the book made it highly unlikely that the book proofs of the last part (Chapters 19-27) were given a proofreading by anyone outside the printer's or publisher's offices and since there was an opportunity for second magazine proofs of this part to be gone over later, one can reasonably conclude (barring the unlikely possibility that "colourless" is simply a compositorial error or that the book publisher's editor engaged in this kind of revision) that the magazine reading "snow-white" is the later reading here. Deciding that it was in fact Howells's alteration is of course a matter of judgment, but a judgment made within the limits imposed by the factual evidence.[35] In the case of Moby-Dick, those limits are wider, because it is known that the publisher's reader for the English edition made numerous substantive alterations and that Melville also had the opportunity to make revisions for that edition; distinguishing the two categories can result only from critical judgments as to which kinds of changes are likely to have been made by a somewhat pedantic reader concerned with expurgation and which are more characteristic of Melville.[36] Fredson Bowers makes the same point in connection with Fielding's Miscellanies, where some parts of the first edition were set from marked copies (not extant) of printed periodical texts: "This is a critical process almost exclusively, with only occasional bibliographical guidance, in which the editor shoulders his proper responsibility to separate the author's intended alterations from the verbal corruption that inevitably accompanies the transmission of a text."[37] Sometimes a statistical analysis of internal evidence can be of material assistance in making a critical choice among variants: tabulating the pattern of recurrences of unusual spellings and other features in Shakespearean texts can help to determine which characteristics of those texts derive from the compositors' preferences and which from the printer's copy itself;[38] or examining each variant in the syndicated appearances of Stephen Crane's stories and dispatches in the light of the quantitative evidence (how many times each reading turns up) can help to establish the reading of the syndicate's master proof. Such evidence must then be subjected to critical scrutiny: the fact that only one out of six newspaper texts of Crane's "The Pace of Youth" reads "clinched" at a point where all the others read "clenched" does not in itself dictate "clenched" as the authorial reading, for Crane invariably wrote "clinched."[39]

Once the editor has made his judgments as to which variants are attributable to the author and which to someone else, he must consider the exact status of the latter group. Are all variants for which someone other than the author is responsible to be rejected outright, or is it conceivable that the author's intention may sometimes be fulfilled by other persons? It is not only conceivable but unquestionably true that others can and do sometimes correct an author's writing and in the process fulfill his intention. An author may write down one word but be thinking of a different one, or in reading proofs he may fail to notice a printer's error which creates a new word. When these erroneous words are plausible in their contexts, they may never be recognized by anyone as erroneous; but when one of them does not make sense, and when the correct word is obvious, anyone who makes the correction is carrying out the author's intention. Frequently an editor may believe that a particular word cannot have been intended but is not certain just what the intended word should be; only his critical assessment of the whole matter can determine whether it is preferable in that case to let the questionable word stand and call attention to the problem in a note or to substitute a word which catches the apparent intended sense (again, of course, with an explanation), even though that word may not be the exact one which the author had in mind. In the typescript of A Story Teller's Story, then, alterations in the hand of E. T. Booth cannot simply be dismissed; they must be inspected carefully, because Booth may have noticed places where the typescript reading certainly (or almost certainly) cannot have been Anderson's intended reading, and there is always the chance that an editor might otherwise fail to detect some of them.

An examination of Booth's revisions, however, leads to a more difficult question. Since Booth was the editor for the publisher, can one

argue that, because Anderson expected his book to be gone over in the publisher's offices, the changes made by Booth become a part of Anderson's intended wording? Or, to put the question in more general terms: can one argue that changes made (or thought to have been made) by the publisher and passed (or presumably passed) by the author in proof constitute changes intended by the author? This question is very different from asking about an editor's alteration of obvious errors. The correction of a reading which the author cannot have intended amounts to a restoration of what was in his mind but not on paper, or of what was in his now lost manuscript but not in print. It does not involve any change of the author's intention. But revisions, as opposed to corrections of outright errors, were not previously intended by the author; if the author then explicity endorses them, he is changing his intention. He is free to do so, of course, just as he may have shifted his intention several times before his manuscript ever left his hands. What is at issue, however, is whether he can delegate someone else to carry out his intention, or part of it. If he says that he expects changes, or certain kinds of changes, to be made in the publisher's offices, can the results be regarded as representing his intention, without shifting the definition of "intention"? One might argue, for instance, that Anderson—aware of some of the shortcomings, by conventional standards, of his spelling, punctuation, and sentence structure—did not "intend" for his writings to be published exactly as he wrote them but "intended" for them to be made to conform with conventional practice. But one might also argue, on the other side, that Anderson's writing as it came from him reflects his intention more accurately than it does after being standardized, and that any intention he may have held regarding publisher's alterations amounted only to his realistic understanding of what had to be done in order to get published (and thus was not part of his active intention in the text).[40]

The importance, for editorial practice, of settling this question in general terms is evident. When an editor faces a choice for copy-text between a fair-copy manuscript (or printer's copy) and a first impression, he needs to have—in the absence of convincing evidence—a general policy to fall back on, a policy based on the inherent probabilities in such situations. Of course, if the editor has convincing evidence— not merely the author's statements but detailed information about the author's methods of going over proofs—he can make his decision on that basis. But, as is more likely to be the case, if the evidence is not sufficient for making a competent decision, the editor must have further guidance. Greg's rationale, pointing out the usual deterioration of a text (particularly its accidentals) from one manuscript or edition to another, leads the editor back to the fair-copy manuscript or the earliest extant text which follows it. There has been some disagreement with this position, however, based on the view that the author's intention encompasses the activities which take place in the step from manuscript (or typescript) to print and that the intention is not "final" until the text conforms to the standards which will make it publishable. Philip Gaskell concludes that "in most cases the editor will choose as copy-text an early printed edition, not the manuscript"; the accidentals of the manuscript, he says, "the author would himself have been prepared—or might have preferred—to discard."[41] James Thorpe agrees:

In support of his view, Thorpe offers examples of a number of writers over the years who have expressed their indifference to matters of spelling and punctuation or have asked for help in making their spelling and punctuation conform to an acceptable standard.

Such arguments for preferring the first edition to the manuscript seem to me misguided. While it is true that most authors have the intention of getting published, such an intention is of a different order from their intention to have certain words and punctuation, resulting in a certain meaning, in their text. The intention of writing something publishable is what Hancher would call a "programmatic intention"; what the editor is properly concerned with, as we have seen, is the author's "active intention" manifested in the work. There is no reason why in some instances an author's active intention might not conflict with his programmatic intention. That an author may submit to various publishing-house alterations as a routine procedure in the process of publication does not amount to his changing his active intention about what his writing is to consist of. To say that he "expects" or is prepared to have certain changes made by the printer or publisher is not the same as to say that he prefers or wishes to have them made; to take his implicit approval of these changes on the proofs (or the printer's copy) as a sign that he is now more satisfied with his text is to ignore the many external factors (Melville's "Time, Strength, Cash, and Patience") which at this stage might prevent him from restoring readings that he actively desired. It is of little help to survey what writers in the past have said on the subject of publishers' alterations of their spelling and punctuation, even if there were a valid statistical basis for concluding, as Thorpe does, that most are "of the indifferent persuasion" (p. 151). Indifference is far from suggesting intention; and the motivation for the indifference would in each case have to be examined in order to know how to interpret the statements. But if the attitudes of writers toward this question cannot be fairly generalized about, the views and practices of printers and publishers can. Printers' manuals, after all, are normative and instructional statements, offering a far more trustworthy basis for generalization than individual authors' expressions of their own attitudes. Thorpe himself, after quoting from various manuals, recognizes that, for most of the period with which he is concerned, "it has been the printers (particularly the compositors and proofreaders) who have

mainly exercised this control over the text in the process of transmitting it" (p. 152). And Gaskell admits that "printers seldom gave authors much choice in the matter" (p. 339). If printers and publishers can be assumed as a general rule to have made alterations in the accidentals of the texts which passed through their hands, and if the attitudes of authors toward those changes have been complex and uncertain, it would seem that, in the absence of additional evidence, an author's manuscript could be taken as a safer guide than the printed text to his intentions regarding accidentals.[43]

Whether there is sufficient evidence in a given case to justify taking the first edition rather than the manuscript as copy-text is a matter of judgment. What the editor must attempt to assess is whether the author genuinely preferred the changes made by the publisher's reader or whether he merely acquiesced in them. The idea that an author can actively intend in his work a revision made by someone else depends in effect on the extent to which the two can be regarded as voluntary collaborators. Since collaboration implies shared responsibility, the "author's intention" in a collaborative effort results from a merging of the separate intentions of the individual authors; the final result is thus intended by each of the authors. A work need not be signed with more than one name, of course, for it to be a collaboration. Nor is it necessary for the authors involved to perform equal shares of the work; indeed, two people may collaborate only on certain aspects of a work, and their joint intention would apply only to the words or elements involved. The facsimile edition of the revised manuscript-typescript of The Waste Land offers a rare opportunity to observe some of the collaboration which can underlie a great work. In certain passages Pound's revisions (such as "demobbed" in line 139) or deletions (as in "Death by Water") actually constitute collaboration, though there are other places where Eliot rejects Pound's suggestions (as in the lines on Saint Mary Woolnoth, lines 67-68). That the work is to some extent collaborative is implied by Valerie Eliot's comment, in her description of "Editorial Policy," that "It has been difficult to decide who cancelled certain lines, especially when both Eliot and Pound have worked on them together."[44] A study of this facsimile does not suggest that an editor should incorporate into the text of the poem the lines which Pound rejected and Eliot did not restore; one can argue that at those points Eliot's intention merged

with Pound's intention, even though Pound's markings are what survive on paper. The fact that Pound went over the poem as a friend and not as a publisher's editor does not alter the essential point: in either case it is possible for someone other than the "author" to make alterations which are identical with the intention of the "author," when the relationship partakes of the spirit of collaboration.

The question, posed earlier, of whether it makes sense to believe that an author can ask someone else to carry out his intention in some respect may now be answered in the negative. By definition, an author's active intention cannot include projected activity and cannot include activity of which he is not in control. The ultimate example of delegated intention in writing would be for a person to ask someone else to write an entire work for him; if he then announced that it represented his "intention," he could only mean his intention to write a certain kind of work (his programmatic intention), for his active intention would not be involved. The same is true regardless of what portion or aspect of a work is at issue, as long as the element contributed by someone other than the author must be described with such expressions as "It is what the author expected to have done" or "It is what the author would have done if he had found time." However, if an author accepts what someone else has done not in a spirit of acquiescence but of active collaboration, the result does represent his active intention. Since the scholarly editor, in establishing a text, is concerned with an author's active intention in that text, he can accept into the text what he knows (or strongly believes) to be initially the work of someone else only when it can be regarded as having been accepted by the primary author as a true collaboration. This approach does not alter the crucial role which the editor's judgment plays in evaluating evidence, but it may provide a useful framework into which that evidence can be placed. It also suggests the relative infrequency with which publishers' alterations can be taken to supersede an author's known practice in a prepublication stage of his work.

III

After the editor has separated authorial from nonauthorial alterations and has decided how to treat the nonauthorial ones, he still faces the question of how to define "final" with respect to the authorial variants. Normally, of course, when there are two authorial readings at a given point and their sequence can be determined, the later one is taken to represent the author's "final intention." However, there are in general two kinds of situations in which this view of "final

intention" will prove unsatisfactory: (1) when the nature or extent of the revisions is such that the result seems, in effect, a new work, rather than a "final version" of an old work;[45] and (2) when the author allows several alternative readings to stand in his manuscript or vacillates among them in successive editions. In the first case, one may say that there is more than one "final" intention; in the second, that there is no final intention at all.

The editorial problem in both cases usually reduces itself to quantitative terms: when the authorial variants are few in number, it makes little practical difference if an editor selects one group of readings as "final" and incorporates them into his text, since the reader will be able without much difficulty to analyze the variants for himself and come to his own conclusions about the way in which these variants alter the total effect of the work; but when the number of variants is great, the system of presenting one final text with variant readings in notes is less satisfactory, and the only practical solution is to produce more than one text (perhaps arranged in parallel columns), each representing a different "intention."[46] That the recognition of more than one valid text of a given work is often forced on an editor by the practical exigencies of recording variant readings should not obscure the fact that the theoretical problem of determining "final" authorial intention has no necessary connection with the quantity of variants.

Turning to the first of the two categories—in which an author's revisions produce, in a manner of speaking, a new work—one can identify several patterns. The most clear-cut involves those situations in which the author's revisions reflect motives which make it impossible for an editor to accept the later version of a work as truly representing the author's intention, even though, in temporal terms, this version is "final." If, for instance, an author deletes passages for the purpose of producing a condensation or simplifies the language to make the work appropriate for younger readers, the special motives in each case prevent the resulting revisions from being definitive. The revised version, in such cases, does not represent a refinement of the work as previously "completed" but a new work conceived for different purposes;

if the new version has merit, it is as an independent work to be edited separately. This is not to deny that the author might make in the process some revisions which an editor would adopt as emendations in his copy-text, but in order to qualify for adoption they would have to be revisions unconnected with the aim of condensation or simplification.[47] In other words, two types of revision must be distinguished: that which aims at altering the purpose, direction, or character of a work, thus attempting to make a different sort of work out of it; and that which aims at intensifying, refining, or improving the work as then conceived (whether or not it succeeds in doing so), thus altering the work in degree but not in kind. If one may think of a work in terms of a spatial metaphor, the first might be labeled "vertical revision," because it moves the work to a different plane, and the second "horizontal revision," because it involves alterations within the same plane. Both produce local changes in active intention; but revisions of the first type appear to be in fulfillment of an altered programmatic intention or to reflect an altered active intention in the work as a whole, whereas those of the second do not.

A similar situation occurs when an author makes revisions, not because he wishes to, but because he is asked or compelled to. Herman Melville, after the publication of Typee (1846), was asked by his American publisher, Wiley & Putnam, to soften his criticism of the missionaries in the South Seas for a revised edition, and in July 1846 he complied by deleting about thirty-six pages of material and revising other passages. These changes alter the tone of the book and are not in keeping with the spirit of the original version. There is no question that Melville is responsible for the changes, and in this sense they are "final"; but they represent not so much his intention as his acquiescence. Under these circumstances, an editor is justified in rejecting the revisions and adopting the original readings as best reflecting the author's "final intention"; in fact, to accept the readings which are final in chronological terms would distort that intention. But again the two types of revision must be separated, for Melville made some revisions in July 1846 which had no connection with the expurgation of political, religious, and sexual references—and these an editor would

adopt.[48] In other words, the "vertical" revisions are rejected, and the "horizontal" revisions are accepted.[49] Just as accidental and substantive variants are, in Greg's rationale, to be treated separately, so, too, known authorial revisions must be divided into categories for editorial decision according to the motives or conceptions they reflect.

A further related problem—the weight to be attached to an author's statement about his revisions—can also be illustrated by the Typee case. After removing the passages on the missionaries from the American edition, Melville wrote to John Murray, his English publisher, "Such passages are altogether foreign to the adventure, & altho' they may possess a temporary interest now, to some, yet so far as the wide & permanent popularity of the work is conserned [sic], their exclusion will certainly be beneficial." One could argue that Melville is simply making the best of the situation, that he is rationalizing the changes and trying to convince himself that they are for the better; on the other side, one could say that here is a strong statement from the author about his "intention" and that the author's wishes, so stated, must be respected. However, in the same way that an author may make revisions which do not reflect his ultimate wishes about his work, he may also make statements which, for various reasons, are less than completely candid. In the end, one cannot automatically accept such statements at face value; as in any historical research, statements can only be interpreted by placing them in their context, by reconstructing as fully as possible the course of events which led up to them. The publisher, in the case of Typee, and not the author, initiated the revisions, and there is no evidence, internal or external, to suggest that they are the kinds of changes Melville would have made without pressure from someone else; even his statement implies that the revised work is in a sense a different work, stemming from a different set of programmatic intentions—aimed at producing a permanently popular work, not dated by discussions of current issues. After these considerations, an editor need not feel that Melville's statement makes the case for rejecting the expurgations any less strong.[50] Such statements

by authors should always be carefully evaluated, like any other evidence, but they cannot be binding on an editor. Only the circumstances of each case can dictate the weight to be accorded to these statements, just as the author's actual revisions cannot be indiscriminately adopted without reference to the entire historical situation surrounding them.

This treatment of author's statements has certain further implications. The essential issue, stated baldly, is whether an editor can presume to reverse an author's decision. Even if Melville did not want to revise Typee, the fact remains that he did so and even asserted decisively that the result was an improvement. Is it not the author's prerogative to determine the ultimate form of his work? Suppose that Melville meant what he said and that, even though he would not have made the revisions without external influence, he was sincerely convinced that he had done the right thing. Most editors would disagree with him, but they would also say that it is not an editor's place to determine what the author should have done. If the author has a lapse in taste, the argument goes, that lapse is a historical fact which scholarly research cannot undertake to repair. There is no answer to this argument, of course, so long as the revision was definitely an attempt to improve the work in terms of its original conception (horizontal revision). But when the revision shifts that conception and thereby produces a different work (vertical revision), the editor may only confuse matters by presenting the revised version as the basic text: if he finds the original version a more faithful representation of the author's vision, he is not abdicating his scholarly responsibility in favor of an undisciplined subjectivity if he edits that version on its own terms as a separate work (and goes on to include the variant readings of the revised edition in notes). It is one thing for an editor to impose his taste upon an author's work by choosing among variant readings solely on the basis of their appeal to him; it is quite another for him to put that taste to the service of historical understanding by allowing it to guide him in distinguishing among the levels of authorial revision and discriminating among the various artistic conceptions they represent.

The most familiar situation in which more than one "final" intention can be said to exist occurs when an author, at a later stage in his career, extensively revises a work completed years before—not because

he is compelled to, nor because he wants to condense it, expand it, or adjust it to a different audience, but because he feels he can improve it artistically. The classic case of such revision is the New York Edition (1907-09) of Henry James. It seems to be generally agreed that an attempt to record in textual notes the variant readings between the original and revised versions of the novels and stories included in that edition would be of questionable utility, since the revisions are so pervasive that they create substantially new works. Both versions of a given work deserve to be read in their own right, and an essay generalizing upon and categorizing the differences between them may be more useful than a list of variants appended to one of the texts. Many essays of this kind have appeared, and a fairly recent one, on "Pandora," sums up the situation: "the net result is neither striking improvement nor fatal tampering. The story is better in some ways, worse in others. But it is different—one cannot assert that the changes really add up to nothing."[51] The revised version, because it is essentially "different," manifests a "final intention" which does not supersede the "final intention" of the earlier version. Merely because the revision came at the end of a long career, when James's artistry and insight were presumably more mature, it cannot invalidate the intrinsic merits of the original version. The two are discrete works.

If this point of view has been readily accepted in extreme cases of revision, it has scarcely been considered at all in instances of slight revision. But why should the quantity of alterations affect one's theoretical position? If one treats a heavily revised text as an independent work simply because the difficulties of handling the variants in any other way are overwhelming and then edits a less heavily revised work as a single text with notes because it is possible to do so, the theoretical basis of the whole operation is questionable. The idea that a revised version can be considered a separate work is sometimes said to rest on the concept of organic form—the view that form and content in a work of art are so integrated that any alteration produces a new entity. Of course, it is not necessary to adduce this concept in order to make the point: changing a word in any utterance results technically in a different utterance. Although the change of one word in a novel

makes less practical difference than the change of one word in a brief lyric poem, strictly speaking each version (both of the novel and of the poem) is a "separate work." Maintaining this position would not quite put an end to all scholarly editing, because editors would still have the task of detecting nonauthorial readings (emendations of publishers' readers, compositors' errors, and the like) and purifying the text of them;[52] but they could not choose among authorial variants, for they would have to consider each group of them, for each successive impression or edition, as resulting in a new work to be edited separately.[53] Clearly such a situation would be intolerable from a practical point of view; in the majority of instances editorial choice among authorial variants does not deprive readers of the opportunity for reconstructing other forms of the text on the basis of the material presented in the apparatus, and a list of variants has the positive advantage of drawing together the evidence from various versions into a form where it is conveniently comparable.

If, in practice, editors are not going to regard each version as necessarily a separate work, then some rationale is required for distinguishing those instances of revision which are to be edited as separate works from those which are not. A quantitative dividing line is not logical: it would be impossible to set up a particular number of revisions, or words involved in revisions, as the test for defining a separate work in this sense.[54] What is more meaningful than the extent of the revisions is their nature. One author might make 3000 changes in his selection of adjectives and adverbs, for instance—and perhaps improve his book stylistically—without altering his original conception of the work at all; another might make only ten revisions in key passages and change the whole direction of the book. Whether or not two versions of a book are treated by an editor as independent works should depend on a qualitative, not quantitative, distinction. If revisions do not spring from the same conception of an organic whole as the original version manifested (what I have called vertical revisions), then they produce a new work, even though the actual number of new readings is small; if revisions are attempts to develop and improve the original conception (what I have called horizontal revisions), then they do not produce a separate work for practical purposes, regardless of the number of changes involved.[55] Generally, large numbers of alterations do follow from a changed conception or programmatic intention, but the point is that there is no necessary connection between the two.

In making decisions about authorial intention, an editor may be inclined to take into account a related factor, the timing of the changes. When an author, late in his life, makes revisions in an early work, one could argue that the result will almost surely constitute an effectively different work, since it is unlikely that the author will have the same conception of his work in mind as he had during the process of its original composition. James's revisions are a case in point, but the argument can be applied to other instances in which neither the extent of the changes nor the shift in intent is so pronounced. When

Arthur Stedman edited Typee in 1892, he claimed to have made alterations "by written direction of the author" (who died in 1891). The only presently known evidence of any authorial direction is a note in Mrs. Melville's hand listing four changes which her husband requested.[56] Even assuming that this note accurately represents Melville's final wishes, how much weight is to be given to a few isolated changes suggested nearly fifty years after the original composition of the work? Two of the changes represent the same kind of expurgation which Melville was required to make for the earlier revised edition and are not consistent with the spirit of the work. These revisions are different from James's not merely in quantity but in the fact that they are not part of a sustained and coherent reshaping of an early work. Instead, they are simply instances of sporadic tinkering; such tinkering, when performed during or soon after the composition of a work, can be expected to fit the general tone and spirit of the whole, but when it occurs much later the results may well seem out of place. A systematic job of revision, even if it does not result in many changes, may have a coherence of its own, but isolated changes frequently clash with the larger context. Nevertheless, it is obviously possible for authors to make consistent sporadic revisions late in life, and the timing of revisions is therefore not in itself the key. Just as a quantitative measure of revision will not serve to distinguish what versions are to be edited as separate works, so a time limit is similarly unrealistic: one cannot say that all revisions made within a week, or a month, or a year of the original composition are to be accepted as part of the same conception, while those made after that time either result in different works or represent random thoughts not consistent with any coherent plan. What is important, once again, is the nature of the changes, and no mechanical rule—about their extent or their timing—can produce meaningful distinctions among them with respect to underlying conceptions or motives.

The role which these considerations play in editing and the critical nature of the decisions they imply are well illustrated in Bowers's edition of Stephen Crane's Maggie.[57] Crane's book was first printed privately in 1893; three years later, in order to secure publication by D. Appleton & Co., he agreed to make revisions, particularly the elimination of profanity. But, as Bowers points out, Crane's alterations were not limited to bowdlerizing: "It is clear from many examples

that he took the opportunity to make stylistic revisions as well as literary improvements" (p. lxviii). As in the case of Typee, an editor will reject the enforced expurgations and will accept the stylistic revisions made at the same time. But Maggie offers in addition a difficult intermediate category: the removal of various sordid details, culminating in the cancellation of a 96-word paragraph describing a "huge fat man," which had appeared in the 1893 edition at a strategic point, just before Maggie's death at the end of Chapter 17. The first critical question is obviously to decide whether these alterations were among those which Crane made under duress or whether he made them independently, judging them to be artistic improvements. Answering that question, as Bowers recognizes, involves literary judgment, and he provides a long interpretive discussion (pp. lxxvii-xci) of the implications of the removal of that paragraph, concluding with the view that Crane did delete it for artistic reasons. Once that decision is made, there is a second critical question to be faced: does this revision (and the scattered lesser ones similar to it) produce an essentially different work? Bowers clearly states the possibility: He then concludes that "Maggie does not bear comparison with these examples," arguing that Crane "was operating from a strong literary conviction about the integrity of a text once written and published" (p. xcv). It is possible, then, by incorporating the 1896 revisions which Crane "made for his own purposes and satisfaction" (p. xcvii), to produce a single "'ideal' text of Maggie as a literary fact, not a limited 'ideal' text either of the 1893 or of the 1896 edition" (p. xcv). Obviously this conclusion is reached through critical analysis, and another editor might analyze the situation differently and come to the opposing view—that there are two distinct Maggies which it would be improper to merge.[58] No incontrovertible answer is to be expected to a question like this, dependent on judgment; but every editor in his own work must recognize that the question exists and reach an answer to it.[59]

With some authors the possibility of multiple "final" intentions is further intensified. Instead of making one systematic revision of a work at some point later in life, they revise their work continually throughout their careers. An extreme example of this method, referred to by Bowers in the passage quoted above, is Whitman's Leaves of Grass, which was extensively revised eight times between 1855 and 1892. When an author works in this way, successive editions constitute a printed record of a developing mind. The fact that Whitman said of his final "deathbed" edition (1891-92) that any future edition should be "a copy and facsimile, indeed, of the text of these . . . pages" does not mean that critics and scholars must reject all earlier editions as works in their own right.[60] Even if Whitman came to think of the earlier editions as preliminary drafts for his final version, each of those editions was published and at the time of its publication represented a final version that he was willing to present to the public and thus his final intention as of that moment.[61] If one decides that the revisions at each stage are the kind which spring from an altered conception of the whole, one can argue that each edition of Leaves is a separate work with its own final intention. (The situation would differ from that of Henry James only in the greater number of separate works, resulting from the greater number of stages of revision.) In that case, Whitman's last text is not—as an intended work—any more "final" than his earlier texts; it merely comes later. To one taking this position, Whitman's own judgment should of course be no embarrassment; it is only a critical pronouncement about his work, not an element within the work.

These issues are raised in an acute form in connection with the poetry of W. H. Auden. Joseph Warren Beach, in The Making of the Auden Canon (1957), describes in great detail the way in which Auden continually omitted or revised passages to bring his poems into conformity with his current ideological preoccupations. In 1945, for example, he gave the poems from "In Time of War," in Beach's words, "a more distinctively religious cast than they had when first written in 1938 and published in 1939" (p. 10). Throughout his career, according to Beach, Auden displayed a faculty for "domesticating, within the frame of mind that at any moment possesses him, work conceived in some quite different frame of mind" (p. 15). In preparing a collected edition, he not only revised poems to make them "reasonably acceptable to him at a time when he was concerned that his work should be as edifying spiritually as it was imaginatively arresting" (p. 242); he also arranged the poems in alphabetical order so that their connections with previous volumes or particular stages in his career would be obscured. The situation is reminiscent of Whitman's preparation of an authorized final edition, but in Auden's case the emphasis is more clearly on ideological content than on artistic form. As with Whitman, Auden's final text can be regarded as only another text, reflecting a different conception of his work.

Beach's analysis presents, in effect, the apparatus for a critical edition in essay form—a method which, for this kind of author, is perhaps more useful than a list of variant readings, since such a list tends to suggest that the versions compared are essentially the same work. At the end of his book Beach takes up—in one of the few discussions of the subject in print—some of the editorial implications of multiple authorial intention. One can concur with his feeling of dissatisfaction about Auden's collected text and yet find a curious logic in his conclusion:

In these comments Beach is explaining what he wishes Auden had done but in the process confuses the roles of author and scholarly editor. The particular revision and arrangement which Auden decided upon for his collected poems represent, in themselves, one of the "successive periods" in the "course of an interesting mind." The fact that he did not arrange his poems in chronological order or leave them unaltered does not prevent the editor from making the earlier texts available. Many of Auden's decisions may not please an editor, but, whatever they are, they constitute the only material the editor has to work with. One can criticize an author's lapses, but one cannot expect him to treat his own work as if he were a scholarly editor.

Pursuing the nature of Auden's "identity," Beach sensibly asks, "And how . . . can we question the right of an author to be his own judge as to the intent of a piece of writing, or to make it over so as to give it a new direction?" (p. 251). This, after all, is the central issue. But the answer again proves troublesome: "What I have suggested is that such a making over of a work of literary art is not to be accomplished by cutting out a few offensive passages, or by merely hanging the work in a different gallery in different company; and that it is vain to suppose that now it means something essentially different from what it did." In other words, as he goes on to say, a work of art should have "a wholeness, or integrality, that underlies all the diverse and even conflicting elements" (p. 253). It is precisely because of its "integrality," however, that any adjustments made in a work of art may turn it into a different, if no less integral, work. To say that an author's last version of a work means the same thing as his earliest is to abandon all criticism; but to find that a late version fails to supersede an early one is not to deny the author's right to do with his work as he pleases. In the end, whenever there are authorial revisions, an editor is not fulfilling his responsibility to the work of literature if he does not assess the nature of those revisions, in order to determine whether he is really dealing with only a single work.

Before glancing at the argument that this approach to editing gives the editor too much freedom to be eclectic, let us turn to the second major category in which "final intention" is problematical—those instances in which there is literally no final intention, either because the author never prepared his manuscript for publication or because he wavered in his revisions for successive printings.[62] Perhaps the most

common instances of this situation occur in the editing of letters. Although letters, generally full of abbreviations and elliptical remarks, can be described as manuscripts not prepared for publication, they have one peculiar feature: they were not (in most cases) intended for publication. Whatever form the manuscript is in, therefore, if the letter was sent, represents "final intention"; the posting of a letter is equivalent to the publication of a literary work, for each activity serves as the means by which a particular kind of communication is directed to its audience. When letters are published, do they automatically become a different genre, subject to different conventions, or is their intention distorted if they are not reproduced exactly as they arrived in the recipient's hands? If the author prepares his own letters for publication, he will almost certainly alter them (at least with respect to accidentals, but possibly also to substantives), and he will probably expect them to be subjected to the same processes of copy-editing and house-styling as any other work. But when letters are published posthumously, does the fact that the author would have expected them to be adjusted to conform with the conventions of published writing justify an editor's attempt to perform those adjustments? Clearly it does not, because, as we have seen, what an author expects is different from what he actively intends. In any case, an editor cannot possibly put himself in the frame of mind of a publisher's house-stylist of some previous period, and the changes he would introduce, however knowledgeable he may be, could carry no authority. Additionally, the abbreviations and other unconventional features of a letter may be its most revealing characteristics; if they are removed or normalized, the substance of the letter and the nuances conveyed to the recipient may be obscured, if not substantially altered. Naturally, some adjustments are inevitable, since complete fidelity to the original would mean photographic reproduction on the same quality of paper. But alterations—even in such matters as the misspelling of words—should be made with extreme caution if the effect of a letter as a private document is to be retained. This procedure comes closer to the author's intention, as revealed in the finished text of each letter, than following any directions the author may have pronounced when he was thinking of his letters more as literary property than as private expression.[63]

A similar situation exists in connection with journals, notebooks, and other personal papers, except that for these classes of material

there is not even that degree of finality accorded to letters by the act of posting. If the writer made no final selection among alternative words or phrases, an editor has no basis—nor justification—for doing so; to prepare a "clear text" which reads smoothly is to change the essential nature of the document. Such works, though they may turn out to be literature, form a special genre in which the necessity for final choice (forced upon an author in the case of published works by the act of publication) does not apply. From a practical point of view, some of these works gain little from the preservation of their formal texture, and it may be that the group of readers who will be turning to a particular document may find the loss of such fidelity a price worth paying for a conveniently readable text. The nature of the document and the uses to which it may be put will in each instance determine the degree of compromise which can be tolerated. In some cases a full transcription may be accompanied by a separate "reading text."[64] But the theoretical point remains: altering private papers to conform to conventional standards of publication makes different works of them and thus is bound to distort their meaning.

When writers leave unfinished, or unprepared for publication, literary works of other genres—those which are normally circulated in published form—the problem is somewhat different. In these cases the rejected readings, false starts, and uncanceled variants are of interest in showing the writer's manner of working and stylistic development, just as they are when found in the surviving manuscripts of a published work; but they do not reflect the essential nature of the work itself, as they do in a letter or a journal. An editor who completes the author's job by preparing such works for conventional publication (correcting errors, choosing among uncanceled variants, and the like) is not obscuring the final effect or meaning of the work but rather clarifying it. When a poem, left in manuscript, is posthumously published in the form of an exact transcript, it is being treated like a historical

document; when it is published in a clear reading text, it is being treated like a work of literary art. Both forms may have their uses, but only the second can represent (or attempt to represent) the author's intention.[65]

The poems of Emily Dickinson present a special situation: they are clearly poems (not journal entries or letters), but they were not intended for publication. They contain both eccentric punctuation (often impossible to reproduce in type) and uncanceled alternative readings. If an editor decides to publish as exact transcriptions as possible of these poems (or even photographs of the manuscripts), he is doing what normally is most sensible for works not intended for publication. But in this case he would be doing less than full justice to the material, which belongs to a genre conventionally circulated in some kind of published form and with decisions among alternative readings already made. The fact that Emily Dickinson did not "intend" publication does not alter the basic nature of the material and automatically convert into notebook jottings what would have been called poetry if published. Her distrust of publication does not obligate an editor to leave her poems unpublished (or to edit them as if they were private papers) any more than an author's "deathbed" edition obligates an editor to regard previous editions as superseded. In either case the work has an existence distinct from the wishes (expressed or implied) of its creator, and "intention" regarding publication is different from the active intention embodied in the work. Whether or not Emily Dickinson's manuscripts were specifically "intended" for publication is really beside the point; the important matter is that they are manuscripts of poems not prepared for publication. Although an editor will rightly feel an obligation to present as fully as possible the evidence available in those manuscripts (as documents in the history of American literature), he should feel equally obliged to make decisions among the author's alternative readings and produce a clear text of the poems (as literature).[66] An editor who

thus "completes" unfinished poems is not being presumptuous but is simply facing his responsibility. One editor's choice among alternatives may of course differ from another's, but the excellence of any critical edition—whether based on unprepared manuscripts or not—is directly related to the critical powers of its editor.

IV

Some of the implications of the Dickinson problem for editorial theory are discussed by R. W. Franklin at the end of his important book The Editing of Emily Dickinson: A Reconsideration (1967). Franklin correctly asserts that "from the variant fair copies of a single poem we should choose its best" (p. 133), and he objects to any nonliterary or mechanical basis of selection among alternatives in unfinished manuscripts as resulting merely in "a worksheet without all the work" (p. 134). Since an uncompleted manuscript obviously lacks finality, he concludes, "The principle of editing that a text exactly represent the author's intention is inadequate." He therefore calls for "a new editorial procedure for material unprepared by the author for publication"—a procedure which would be "a compromise between the demands of authorial intention and the demands of the poems" (pp. 142-143). One might carry the argument a step further, however, and note that since authorial intention is ultimately ascertainable only through the poems, no compromise is necessary except in the sense that two kinds of edition, rather than a single one, may be desirable: a complete transcription, faithful to the demands of the document, and a reading text (or more than one), faithful to the demands of the work of art.

Generalizing upon the specific situation, Franklin points to the "conflicting bases of criticism and editing"—conflicting because the modern critical position upholds "criticism divorced from authorial intention." Is there an inconsistency, he is asking, within the discipline of literary study, if a text, presented to the literary critic for analysis in the light of one set of principles, is prepared for him by the editor under a different set of principles? An author's final intention, he believes,

Readers of literature, he says, are not accustomed to dealing with multiple wordings in a final text nor with composite authorship. Yet one Dickinson "poem," "Those fair—fictitious People," has twenty-six variants that fit eleven places, amounting to 7680 possible poems; and other poems, as traditionally printed and studied, would have to be called "Dickinson-Todd-Higginson's," since editors were responsible for some of the words. In the end, Franklin observes, "the fact that we are not organized to talk about an altered poem as a poem shows how little the subject of our pursuit is poetry" (p. 141).

Two issues are involved in these considerations, and they are basic to all kinds of editing, whether the copy-text is an uncompleted manuscript or a printed edition: (1) What does "intention" signify, and when is it final? (2) Does it matter whether the author's wording is recovered, particularly when emendations by others are improvements? These questions ask for definitions of the three words "final authorial intention" and for justification of them as an expression of the goal of editing. I hope that what I have said up to now has provided some answers to them and will serve as background for the following brief replies, specifically directed at Franklin's conclusions.

The second of the questions is easier to answer than the first. No one presumably would deny that any alteration in a literary work could be regarded as producing a different work and that the new work could be made the subject of critical analysis. Neither would anyone deny that nonauthorial revisions could produce a work superior to the original and more rewarding for study.[67] Nevertheless, if an editor sets out to edit the works of a particular writer, he has undertaken

a task of historical research, and his goal must necessarily be the recovery of the words which the author actually wrote. That the bulk of scholarly editorial work has been of this sort does not imply that all critics will find this kind of edition appropriate for their purposes or that no other approach to editing is legitimate. A critic may choose to discuss a series of poems on death, say, rather than a series of poems by Milton, Shelley, Tennyson, and Dickinson; so long as he operates outside of a historical framework and makes no references to the authors or their times, he need not be concerned with whether he has the precise words of a particular author but only with whether he has the "best" version of each poem from an aesthetic point of view. Similarly, an editor could edit a collection of poems on death, letting his own aesthetic judgments guide him in improving upon any previously known version of each poem; the editor would become a self-invited collaborator of the original author, and the editorial process would be creatively, rather than historically, oriented. This kind of editing occurs regularly in publishing houses, and many books normally attributed to a single author are already the work of more than one person by the time of their first publication (one thinks immediately of the editorial labors of Maxwell Perkins at Scribner's).[68] The crucial point is that once a critic refers to two poems because they are by the same author, he has introduced a consideration extrinsic to the poems, and he must thereafter be concerned with the words which the author wrote. Studying poems by particular authors or representative of particular historical periods, therefore, requires a knowledge of what the authors themselves wrote; studying poems by theme or type, without regard for biographical and historical contexts, requires only poems, and the number of hands through which a poem has passed to reach its present state of excellence is irrelevant. Academic departments are usually organized to study the historical development of literature, and it is not surprising that scholars in those departments produce editions which attempt to recover authorial wording. That they do so, however, is not indicative of a split between editorial and critical theory.

The other question—the meaning of "intention"—is too complicated a philosophical issue to be settled here; but we can at least agree that authorial intention in literature cannot simply be equated with an explicit statement by the author explaining his motives, purposes, aims, wishes, or meaning, for intention must surely exist even if no such statements were made or are extant, and any available statements may be inadequate or misleading. The only direct evidence one has for what was in the author's mind is not what he says was there but what one finds in his work. An editor, only through his analysis and understanding of the meaning of the work in the light of his knowledge of the author and the times, will be in a position to use authorial active intention as a basis for editorial choice. That is to say, of the meanings which the editor sees in the work, he will determine, through a weighing of all the information at his command, the one which he regards as most likely to have been the author's; and that determination will influence his decisions regarding variant readings. Recognizing "finality" of intention, in turn, depends on his ability to distinguish revisions which develop an intention in the same direction from those which push it in another direction: the former represent final intentions, the latter new intentions. Whether the editor rejects such "new" intentions or edits a separate text embodying them will vary with the particular situation. But so long as he is producing an edition of an author's writings, he must choose among the author's uncanceled

variants or published revisions in the light of his total understanding of the work and its author.

If it is objected that this conception of the editorial process gives an editor excessive freedom and substitutes subjectivism for rigorous discipline, two answers may be made. In the first place, a scholarly editor (as opposed to a creative one) is still pledged to print only the author's words. He may select readings on the basis of his own literary judgment only when the alternatives are authorial variants; when he chooses an authorial reading previous to the author's last one (or what he judges to be the last one), his justification is that the reading is "final" in terms of his view of the work as an organic whole and that the later reading either creates a new work or is an isolated alteration at odds with the spirit of the work. Beyond that, one may observe that critical perception is necessarily crucial to any act of historical reconstruction, any evaluation of evidence, and thus any edition labeled "critical"; therefore, as Greg says, "it would be disastrous to curb the liberty of competent editors in the hope of preventing fools from behaving after their kind." Perhaps the principal source of difficulty lies in thinking of the editorial and the critical functions as essentially distinct. When one recognizes that justice can be done to an author only by doing justice to his text, one also understands that the editor and the critic must be inseparable.

Notes