University of Virginia Library

I

It should not be surprising that all textual scholarship is related, for the same activities are involved, regardless of the diversity of the materials. One must decide whether to produce a diplomatic—that is, unaltered—text of a single document or a critical text, which is a new text that incorporates the results of editorial judgment regarding variant readings and errors. One must assemble the relevant or potentially relevant documents (handwritten, typed, or printed), then find out in what ways their texts differ by collating them, then attempt to determine the relationships among the texts, and finally, if the edition is to be critical, construct a new text by choosing among variant readings and by making conjectures where errors seem to be present in all texts. These stages are interrelated: the kind of thinking one brings to the task of determining relationships among texts, for example, will obviously have a bearing on the decisions made at the next stage. Although these two

stages are not entirely separable, textual discussions do, in practice, often emphasize one or the other; and I think it is fair to say that perhaps the principal distinction between the body of writing concerned with editing the classics and that dealing with editing modern literature is their differing emphases in this regard. Editors of the earlier material have access to no authorial manuscripts and must contend with copies an unknown number of steps removed (and often many centuries away) from those originals, and such copies sometimes exist in the hundreds, or even—as with the Greek New Testament—the thousands. The task of working out the relationships among the texts of these documents is indeed formidable, and it is natural that a great deal of the thought and writing about editing ancient texts[5] has concentrated on this stage of the editorial process. What has traditionally been called "textual criticism" —or, more recently, "textual analysis"—is this attempt to fix the relationship of the surviving documentary witnesses; and though many of the theories of textual criticism have entailed certain assumptions about how the editor's critical text should be constructed, the focus of attention has normally been not on the "editorial" phase (the actual selection or emendation of readings) but on the prior analysis of the texts that results in the assignment of relationships among them.[6]

Methodological writings about the editing of post-medieval literature, on the other hand, have reversed this emphasis. Although relationships among the texts from this period are by no means always clear-cut, the dimensions of the problem are often significantly different: manuscripts in the author's hand, copies made directly from them, printed editions set from such documents (and perhaps proofread by the author), and later editions during the author's lifetime (perhaps set from copies of the earlier editions annotated by the author) are the characteristic materials. Editorial theorists concerned with this period

have therefore not been required to give as much thought to the question of establishing relationships among texts and instead have concentrated their attention on the choice and treatment of a "copy-text." Choosing a copy-text is of course dependent on knowing the relationships among the texts; the central problem, however, is not the process of establishing those relationships but of defining the authorial intention that is to be reflected in the critical text, since often more than one document exists that is directly associated with the author. The term "textual criticism" can be used broadly to designate the evaluation of textual witnesses for writings of any period; but its traditional, and more restricted, application to the study of ancient manuscripts is appropriate, for it refers to the kind of analysis that has bulked largest in textual work on those manuscripts.

In classical textual criticism, these basic operations have generally been referred to as recensio and emendatio, and the distinction between the two points up another contrast with textual scholarship of later literature. Recensio refers to the process of establishing the archetype, or the latest common ancestor of all surviving manuscripts, insofar as it can be established from the evidence in those manuscripts, which are the only witnesses to the tradition. The particular decisions made about individual variant readings in the construction of this archetype depend, at least in part, on the relationships that have been postulated among the manuscripts; the practice of stemmatics—of constructing genealogical trees to show manuscript relationships—is therefore also sometimes called "recensionism." And whether or not one aspires to a system that eliminates judgment in the construction and use of the stemmata, the fact is that ultimately judgment will have been involved in the attempt to choose the wording of the archetype from among the variant readings. Swings in scholarly fashion toward, and away from, the use of critical judgment—along with the associated tendency to favor, or disapprove of, eclecticism—must be looked at later; but the point here is not whether a single text is principally adhered to in producing the new recension but the fact that the recension is defined as being limited to readings present in the witnesses (or obvious corrections of them). It is the next stage, emendatio, in which the editor can engage in conjecture to rectify what appear to be errors in all preserved texts.[7] Editorial discussion

dealing with post-medieval works, in contrast, generally takes the term "emendation" to refer to any alterations introduced by the editor into a particular documentary text (the one chosen as "copy-text"), whether the source of those alterations are other texts or the editor's own ingenuity. It is perhaps natural that this usage should have prevailed among editors of modern works, since they often have an author's manuscript or a text only one or two steps from it to use as a copy-text, and all their alterations may then be seen as corrections to that single documentary text; editors in the earlier manuscript tradition, on the other hand, normally have no such text to choose, and the process of arriving at what might be regarded as the counterpart is a major undertaking in itself, to be accomplished before one can begin to think about how that text departs from what the author must have intended to say. In any event, however the difference in usage came about, it should be clear that both groups of editors are talking about the same categories of editorial intervention—alterations based on readings present in one or more of the documents and alterations emerging from the editor's own conjecture.

It should further be evident that any approach or vocabulary suggesting that the latter are more conjectural than the former is delusory. Of course, a reading adopted from one of the documents may be a striking reading that the editor would not have thought of or dared introduce independently, but the decision to consider it as worthy of acceptance into the critical text is still an act of conjecture, always entailing the potential danger that the reading is accorded too much credence by the mere fact of its existence in one of the documents. To regard the choice among variants as "recension," defined as establishing "what must or may be regarded as transmitted" (Maas), and then to label further editorial alteration as "emendation" or "conjectural emendation," would seem to overemphasize the objectivity of the first and to imply a greater distance between the two than in fact exists. The recension, after all, is a conclusion resulting from scholarly judgment or conjecture—except, of course, when only one text survives or (theoretically) when all surviving texts are identical. Even when the archetype appears to be the text of one of the extant manuscripts, judgment regarding individual variants is still involved in reaching that decision. To think of

"what must or may be regarded as transmitted" as a single text, when variant texts survive, is to engage in conjecture; and some of the "conjectural emendations" that an editor thinks of may attain to a higher degree of certainty than some of the choices that are made among variant readings in the documents. I do not believe that editors in any field would disagree with this point, despite the implications of the language sometimes used. The fact that the terminology employed by editors of classical and of modern texts diverges somewhat is not important, so long as both groups of editors recognize that they are dealing with the same fundamental questions and so long as they are not misled by the superficial suggestiveness of some of the terms.

As the division of the editorial process into recensio and emendatio makes clear, editors of ancient texts are normally concerned with producing critical editions—editions, that is, containing texts that are different, as a result of the editors' intervention, from any of the documentary texts now existing. Editors of printed texts from the last five hundred years have also been engaged for the most part with this kind of edition: the extensive discussion in the wake of Greg's "Rationale," for instance, has concentrated on critical editions. Yet in the exchange of views that has increasingly been taking place in recent years between editors of modern literature and editors of statesmen's papers, some of the so-called "historical" editors have questioned the value of critical texts, or at least of texts that are "eclectic" in incorporating readings from two or more documents. It is easy to see why a historian editing letters and journals in the hand of a particular statesman would think primarily of a diplomatic edition, and similarly understandable that an editor of an ancient Greek text surviving in much later manuscripts would probably wish to construct a new text attempting to restore the author's words. But the difference between the two situations does not really rest on the different nature of the materials: there are different goals involved, the aim in the former instance being the reproduction of the content of a given document and in the latter being the reconstruction of what the author of a text intended to say. Both approaches are applicable to any material: documents containing ancient Greek texts, for instance, can obviously be treated as entities in their own right, with texts to be exactly reproduced, as manifestations of particular moments in the history of the pieces of writing involved; or they can be regarded as evidence to be used in reconstructing a text nearer its author's intentions than any of the surviving texts manages to come. Historians may more often find themselves producing diplomatic texts of particular documents (the contents of which were often not intended for publication), and scholars of literature (both ancient and modern)

may more often be engaged in constructing critical texts of works (ordinarily finished pieces of writing—whether "literary" or not—intended for public dissemination). But each group should recognize the value of both approaches and understand how they are related to one another.

This point would seem to be so elementary and obvious as not to need stating; but unfortunately some textual controversies have arisen through a failure to keep in mind the most basic distinctions and to appreciate the place each editorial undertaking occupies in the large framework that encompasses all textual work in all fields. One historical editor has gone so far recently as to make this statement: "To what uses literary critics may put bastard documents is for them to say, but the saying of the same will not likely change the historical discipline's rules of evidence and citation."[8] The narrowness and closed-mindedness of this position is astounding. In a more sophisticated form, however, this issue keeps turning up: the question of eclecticism has perennially been a point of controversy among editors of the classics as well as of modern works.[9] Some editors of modern literary works, who well understand the value of critical texts, have nevertheless argued against combining into a single text readings that reflect different stages of authorial revision. There is nothing wrong in principle, of course, with the position that authorially revised texts may at times be best handled by preparing separate critical editions of each version. But the mistake that sometimes follows is the belief that no variant from one version can be incorporated into another. That injunction would naturally be proper if one were producing a diplomatic edition of each version; but if a critical text of each version is the goal, then one must recognize that some of the variants among versions do not represent a particular stage of revision or rethinking but are precisely the kinds of corrections that the editor is already committed to inserting—without documentary authority. I make this point (which has been discussed more fully else-where)[10] in order to suggest, once again, that the distinction between

adopted variant readings and conjectural emendations needs to be thought about less mechanically than it often is and to show that editors would be well advised to keep abreast of textual debate in fields other than their own. Just as editors of statesmen's papers and of modern literature stand to benefit from knowing more about the editorial thinking underlying critical editions of ancient texts, so editors of the classics (and of statesmen's papers) will find that the discussions of authorial revisions, engaged in fully by editors of modern literature, raise questions relevant for them.

Further indication of these connections can be suggested by referring to three of the more recent manuals on textual criticism, published coincidentally at about the same time, James Willis's Latin Textual Criticism (1972), Martin L. West's Textual Criticism and Editorial Technique Applicable to Greek and Latin Texts (1973), and Vinton A. Dearing's Principles and Practice of Textual Analysis (1974).[11] These books offer several contrasts. Willis and West, classicists and editors, address their work, as the titles indicate, to other editors of classical texts; Dearing, a professor of English who is establishing the text for the California edition of Dryden and is also working on an edition of the Greek New Testament, intends for his book to be applicable to all editorial scholarship, indeed "to the transmission in any form of any idea or complex of ideas" (p. ix). Willis, whose writing is marred by unsuccessful sarcasm, is principally occupied with restating "the many ways in which scribes were accustomed to make mistakes" (p. ix), though he prefaces that account with a brief section on "Fundamentals"; West, who writes lucidly and concisely, would claim originality largely (though not entirely) for his way of stating certain complex questions and their conventional answers and for choosing passages to illustrate his points; Dearing, who writes at greater length and with some obscurity, covers what a manual must cover but uses the occasion to set forth his own proposal for the analysis of relationships among texts. Willis's book is the narrowest and least significant of the three, focusing on scribal errors and devoting considerable space to "trial passages," on which readers are invited to exercise their ingenuity by proposing emendations

(answers are provided). Dearing's experience with both ancient and modern texts (he is perhaps unique in working in both fields) and his view of textual analysis as "a completely general discipline of very wide specific applicability in the arts and social sciences" (p. 1) are encouraging signs, and one has good reason to expect his book to have broader significance than West's.[12] In my view, however, West's book is calmer, clearer, and more sensible and finally a better introduction for students from any discipline. Whether or not I am right, I hope that one point implicit in my opinion will be granted: that the interests of all who deal with texts are closely related and therefore that the sources of specific illustrations are of less moment than the basic statements and discussions of principles. A book that draws its examples from many periods and languages is not necessarily of more general applicability than one that takes all its illustrations from Greek and Latin texts; it may be, but the range of examples and even the immediate aims of the author are not the decisive tests.

One vital matter commented on in all three books—and one that editors of modern literature have a particular interest in—is the role of the analysis of physical evidence in textual decisions. What has come to be known as "analytical bibliography" is crucial to the editing of texts in printed books: in order to be in a position to understand textual anomalies in a printed text, one must first have extracted as much information as possible about the printing of the work from the evidence preserved in the printed sheets themselves. As a result of the efforts of McKerrow, Pollard, Greg, Bowers, and Hinman,[13] and of those that followed their lead, editors of printed texts must now deal with such matters as the identification of compositors' habits and of the order of formes through the press. Knowing as much as one can about what happened to a particular text in the printing shop or the publisher's office puts one in a better position to recognize those features of the text that did not come from the author (or at least were not present in the copy furnished to the printer); analytical bibliography has shown time and

again that much can be learned from physical evidence about the transmission of the text. The same principle obviously applies to manuscripts as well: in the case of manuscript texts not in their authors' hands, the scribe or copyist occupies the roles of publisher's editor, compositor, and pressman combined. Introductory manuals for editors of manuscripts have recognized this point to some extent in that they often contain fairly detailed comments classifying the kinds of errors that scribes were likely to make. Such "habits" are generalized ones, and less attention has been paid to uncovering the habits of particular scribes through physical evidence, including that which fixes the manuscript in time and place.

This whole question enters Dearing's book in the first sentence, where we are told that textual analysis "determines the genealogical relationships between different forms of the same message" but not "the relationships between the transmitters of the different forms"—or, as he puts it in the next paragraph, "the genealogy of the variant states of a text" but not "the genealogy of their records."[14] The distinction, indeed, Dearing regards as one of his central achievements: he believes that his book "carries out to the full" the differentiation set forth in the earlier version[15] between "the genealogy of manuscript and other books as physical objects and the genealogy of the ideas or complexes of ideas that these physical objects transmit" (p. ix). It is of course quite proper to begin with this basic point; editors of all materials from all periods must recognize that the chronology of texts does not necessarily match the chronology of their physical presentation. The point is perhaps not quite such a revelation as Dearing thinks. Nevertheless, it is always good to have fundamental distinctions set forth clearly at the outset of a discussion, and one would have no cause for complaint if Dearing had not carried the point to the opposite extreme, slighting the legitimate role of physical evidence in textual study. Writers in the past, he says, have "almost always" confused the physical document with the text it carries, and he admits that "it is extremely difficult to free oneself from the bibliographical spell"; but it is a "fundamental and important" matter, he insists, "to exclude bibliographical thinking from textual analysis"

(p. 15). That the valid distinction he began with could have led to this wrongheaded conclusion is unfortunate; but the problem might have been predicted from some remarks made along the way to illustrate the basic point that texts are different from their transmitters. To show that "the same record" may preserve "two or more states of a text," he cites as one illustration a poem appearing twice in an anthology (p. 14). But how is "record" being defined here? What is the physical unit? One may ask the same questions when he then says that different records may "transmit the same state of a text when they are produced by a mechanically perfect reproductive process, such as Xerox copying, and whenever it is deemed vital to preserve the text without change, as in statute books, state documents such as the Constitution of the United States, religious documents such as the Book of Common Prayer, and careful scholarly reprints of all sorts" (pp. 14-15). Xerox copying is not "mechanically perfect," if only because the size of the image is not identical with the original. Beyond that, it is no criticism of the Xerox process to say that it is not "mechanically perfect," since no system can be, if what is meant is that the reproduction is identical to the original. The reproduction is a different physical object, and therefore it is not the same thing; and most, if apparently not quite all, users of Xerox reproductions are aware of the dangers of assuming that what they see in the reproduction is precisely what they would see in the original. Furthermore, is Dearing suggesting that whenever "it is deemed vital to preserve the text without change" such preservation is achieved? Are there never errors in the reprintings of statutes, or prayer books, or "careful scholarly reprints"? Does not the acceptance of aim for fact question the need for textual scholarship at all (or any other effort to establish truth)?

The serious bibliographical problem raised by these statements becomes even more evident with Dearing's next sentence: "The many identical copies produced by printing from the same setting of type, however, provided they are uniformly bound and readied for sale as a single lot, are usually counted as one record." Analytical bibliographers have been demonstrating for three-quarters of a century that surviving copies from the same setting of type (i.e., from the same edition) are not necessarily "identical" in their text—indeed, that they are frequently (or, in some periods, usually) not identical. (Whether or not they are "uniformly bound" or "readied for sale as a single lot" has nothing to do with their text.) Differences can come about either intentionally or inadvertently, through stop-press corrections and alterations between printings or through accidents that damage the type (perhaps necessitating some resetting) and deterioration of type or plates through wear. The essential point is that different copies of an edition are different physical

objects and are therefore separate pieces of evidence; it is unscholarly to assume, without investigation, that they are identical, and in fact such an assumption would very often be wrong, for books of any period, even the twentieth century. If one were to regard all copies of an edition as a single "record," one would have to define "record" in a special way, for there would frequently be textual variants among particular copies of the "record." It is difficult, for example, to say what "the" text of the Shakespeare First Folio is; one might say that it is the text in Charlton Hinman's Norton Facsimile (1968)—but that work assembles from various copies of the Folio the pages representing corrected formes. No one surviving copy contains all the corrected pages, and constructing such a copy in facsimile is a task requiring scholarly judgment. Some people who work only with manuscripts (and some who work with printed books as well) think of copies of printed editions as identical, in contrast to manuscripts, each of which is expected to be different. Undeniably manuscripts and printed books are produced in fundamentally different ways; but the fact that copies of an edition are mass-produced and intended to be identical does not mean that they are actually identical. Indeed, they cannot be identical, since no two physical objects are identical in every respect; and textual differences are among the kinds of variations that occur. Printed books resemble manuscripts more than many people seem to think. These are elementary points, and Dearing (who has done a great deal of work with seventeenth- and eighteenth-century English books) certainly understands them; why he fails to take them into account here is inexplicable.[16] The fact that their absence is more likely to be noticed by students of printed books than by students of the manuscript tradition is an indication of the distance that exists between the two groups—an unfortunate distance, since these points clearly have their implications for manuscript study as well and form one more illustration of the common issues facing all textual scholars.

We thus come back to Dearing's assertion that "bibliographical thinking" should be excluded from "textual analysis." It is no doubt true that some textual critics have been confused in their thinking and have not differentiated between a document and the text it contains; but it is an overstatement to say that "textual critics in the past almost always confused the two genealogies when they did not devote their attention exclusively to the genealogy of records" (p. 15). In any case,

the solution to the problem is not to banish the allegedly overemphasized bibliographical approach, since it unquestionably plays a crucial role in the whole process. To be fair to Dearing's argument, one must remember that he distinguishes "textual analysis" from "textual criticism": the latter is the larger term, covering all the stages of textual work, whereas the former is one particular operation, concerned with working out the "genealogy of the states of a text" (p. 2) and reconstructing their latest common ancestor. It is from "textual analysis" that bibliographical thinking is to be excluded. Nevertheless, one can insist that even here bibliographical analysis is important without being guilty of equating texts with records, for the texts are tied to the records, and an understanding of the physical evidence is necessary for an informed interpretation of the textual evidence. Dating a document (manuscript or printed book), for example, is significant even if one recognizes that the state of the text is not necessarily of the same date. Dearing makes much of what he sees as different uses of manuscript dates for textual analysts and for bibliographers: he neatly pairs the successful copyist, who "produces a record that postdates the state of the text it records," with the successful editor, who "produces a state of the text which anedates his exemplar" (p. 39). But to contrast "successful" copyists and editors is to place the emphasis on what they intended to do. In actuality copyists do not always reproduce with fidelity the texts in front of them, and though the records they create certainly postdate the records they use, their texts may also postdate those of the earlier records. Similarly, editors do not always succeed (it must be assumed but cannot be proved) in reconstructing earlier forms of the texts, and the texts they do produce may be said to postdate the other extant states.

Without losing sight of the idea that the genealogy of texts is a different concept from the genealogy of documents, there is a real sense in which one may still claim that a text does date from the time it is inscribed or set in type. The changes introduced by a scribe or compositor, whether out of habitual practice or out of inadvertence, produce a new text;[17] and understanding as much as possible about the production of that text—the habits of the individual scribe, the characteristics of the period, and so on—helps one to know how certain readings occurred. If one rules out this knowledge, one makes textual analysis a rather fruitless exercise, for one may postulate relationships that are shown by

physical evidence to be incorrect. Purely as abstract statements about agreements and divergences among certain messages, they are not incorrect, and this is Dearing's point; but so long as textual analysts are concerned—as they ultimately must be—with direction of descent, with genealogical relationships, they cannot ignore any physical evidence that eliminates certain relationships from further consideration as factual possibilities. Dearing himself does discuss general categories of scribal error (pp. 44-54) in his treatment of directional variation. Recognizing the influence of the physical process of transmission on what the text says (that is, "bibliographical thinking") cannot therefore be divorced from the analysis of the relationships among texts and need not involve a confusion of texts and records. Dearing asserts that textual analysts describe "not what was but what is and therefore what all can agree upon" (p. 19). They describe "what is," however, only if their aim is to record variant readings, and even then it is by no means certain that all would agree on what the reading of each text is at every point. But if the textual analyst's work includes reconstructing "the latest state from which all the extant states have descended," a state that is "in most respects the closest we can approach to the author's original intention" (p. 2), the textual analyst does deal with "what was," just as the bibliographer does. Both are engaged in historical reconstruction, and their tasks are intimately linked. When Dearing speaks of "bibliographical thinking" he may be referring more to descriptive bibliography (the history of the physical forms in which a work has appeared) than to analytical bibliography (the analysis of the physical evidence present in those physical forms), but the latter is of course a tool used in the former; in any case Dearing's approach, in the course of emphasizing texts as "messages," neglects the inextricability of the "transmitter" and the "message" and therefore the role of physical evidence in the interpretation of textual evidence that the textual analyst must perform.[18]

In contrast, West's manual sets forth emphatically, if briefly, the role that the analysis of "external" evidence must play. He points out that the process of examining texts has refined "our understanding of the languages, metres, and styles of the Greeks and Romans," which in turn provides a background for examining further texts; we learn about "such matters as the proclivities of scribes" and "the processes governing the spread of texts at different periods" (p. 8) and need that knowledge (which, as he correctly says, is of interest in its own right) in evaluating particular texts. At other points he refers to the use of paleographical

evidence and watermarks in dating manuscripts (p. 30), along with information about "what is known of the general historical conditions that governed the transmission of classical texts at different times" and "more particular facts such as the movements of individual known scribes" (p. 31); and he speaks of eliminating from the text "those features which we know, from our general knowledge of the history of books and writing, to have been introduced since the time of the author" (p. 54). He recognizes, in other words, the relation of physical evidence to textual study, even if he does not go beyond the traditional statements of its role. Like most writers of textual manuals, he explains the categories of alteration that scribes are likely to be responsible for (pp. 18-29);[19] as usual, these characteristics are discussed as generalized possibilities, and little attention is given to procedures for establishing the habits of individual scribes or assessing the textual implications of physical features that may reveal information about the method of production of individual manuscripts. One misses the kind of detail now standard in the compositorial and physical analysis of printed books; more work is needed for manuscript studies along the lines of Farquhar's and Van Sickle's investigations of the physical characteristics of codices and book rolls and Colwell's focus on the characteristics of particular scribes.[20] But even if, to students of printed books, West's manual seems to slight the explicit treatment of techniques for analyzing physical evidence, they would nevertheless find his essential position congenial. His repeated, if general, comments on the necessity for investigating the processes of the production of manuscripts form a sounder basis on which to proceed than Dearing's well-grounded but inappropriately applied segregation of bibliographical and textual concerns.

West's inclusion of the results of paleographical analysis and watermark study among the types of "external" evidence points to the way in which terminology reflects point of view. To the analytical bibliographer such evidence would be thought of as internal, because it is part of the physical evidence of the document, as opposed to relevant information that comes from outside the document, such as that from publishers' archives or from one's knowledge of the book-making practices of the period (itself built up from internal evidence from other documents). West can place both watermark evidence and one's general knowledge of the period together as external evidence because they are both external to the text, even though one is not external to the document transmitting the text. Both usages—that of the analytical bibliographer and that of the textual critic—are proper: they simply result from different approaches to the material. Such differences in terminology should prove no obstacle to mutual comprehension so long as the operations being referred to are thoroughly understood and so long as the line between the approaches is not imagined to be firmer than it actually is. Some techniques of analytical bibliography—compositorial analysis, for instance—involve evidence drawn from the text itself; determining the habits of a compositor, or a scribe, depends on a close examination of practices within the text. One can say, and some have said, that this kind of examination takes the text only as additional physical evidence, regarding it simply as ink on paper. It is true that those inked shapes constituting the text are physical evidence; it is also true, however, that the analytical bibliographer must understand what the text says, in order to know which characteristics are worth studying as possibly attributable to compositorial or scribal practice. Determining whether this evidence is internal or external is not a very productive problem; what is important is to guard against equating "external" with "objective" and "internal" with "subjective." The terms unfortunately come to have these connotations in many discussions of the textual criticism of manuscripts. Sometimes "internal evidence" is used to refer to the kind of evidence adduced by an editor to support a conjectural emendation (largely evidence from context, which, in varying degrees, involves interpretation and is therefore subjective), and "external evidence" is taken to mean the relationship among manuscripts, which in turn leads (without the necessity of literary judgment) to the adoption of certain readings rather than others. West is too sensible to make this mistake: he describes "the more exact information derived from internal evidence" as "the interrelationships of the copies as inferred from comparison of their readings"

(p. 31), recognizing that the establishment of a genealogy involves inferences and that the result is normally conjecture rather than established fact.

Nevertheless, his statement is not as clear as it might be. One assumes that "copies" refers to texts, for if it referred to the manuscripts themselves the statement would be guilty of the confusion, about which Dearing warns, between records and texts. Even so, there is a problem, for the sentence seems to make "the interrelationships of the copies" wholly dependent on a "comparison of their readings" and leaves one wondering how the "external" evidence previously described fits in. "The inquiry," West says, "proceeds on two fronts, from external and from internal evidence" (p. 30). The external evidence of provenance, paleography, "general historical conditions," and the like then becomes the "historical backcloth" (p. 31) against which to "project the more exact information derived from internal evidence." Stated in this way, it is hard to comprehend precisely what the function of the "backcloth" is in the whole process. West does understand that physical evidence plays a role in interpreting the readings present in a text, but his category of "external evidence" is here presented largely as having to do with the relationships of documents rather than of texts. What is lacking is explicit recognition that the comparison of readings, leading to inferences about the relationships of the texts, must involve analysis of physical evidence as well as literary analysis: examination of the physical features of a document is relevant not only to dating the document but also to evaluating the readings in the text contained in that document. Whether one is dealing with printed books or with manuscripts, understanding the physical evidence may set limits on the literary speculation that can be engaged in. I do not wish to dwell on what is only an infelicity in West's exposition, but it provides an occasion for underscoring a significant point. Distinguishing "external" and "internal" as they refer to evidence is finally not so important as recognizing the interrelatedness of all evidence. Neither kind of evidence has a monopoly on demonstrable conclusions; because generalizations based on inductive evidence are inevitably provisional, some historical "facts" may be more conjectural than emendations based on an editor's judgment. And if one wishes to think of the physical evidence of a document as external to the text, then one must think of degrees of externality, for such evidence is not in the same realm as the larger historical framework into which one hopes to place both the document and the text. If West, instead of asserting that "the interrelationships of the copies as inferred from comparison of their readings" should be projected onto a "historical backcloth," had said—turning his statement around—that the comparison

of readings, taken together with physical evidence and the historical background, results in the tentative establishment of the relationships among the texts, his comment would not have required discussion. That he could state the matter as loosely—in effect, carelessly—as he did must reflect, at least to some extent, a general lack of awareness, among textual critics of manuscript material, of the contribution analytical bibliography has made to the editing of printed texts.

Some of the techniques of bibliographical analysis, such as some of those used to distinguish compositors, rely on characteristics of spelling and punctuation; and data about compositors' habits in these respects are important to the editor whose aim is to restore the author's spelling and punctuation as well as wording.[21] Most scholarly editors of postmedieval writing have such an aim: since spelling, punctuation, capitalization, and so on must be considered an integral part of texts, affecting their meaning, and since most of the extant documents containing these writings are relatively near to the authors' manuscripts (or include those manuscripts), editors dealing with these centuries have given a great deal of thought to the problem of authoritative spelling, punctuation, and capitalization (or "accidentals," as these features are sometimes called, in distinction to "substantives," or the words themselves). It is a common notion that the treatment of accidentals is one of the major respects in which the editing of ancient writing differs from the textual work on more recent material; because the spelling, punctuation, and system of abbreviations in the surviving texts of ancient works generally reflect the customs of scribes who lived long after the authors of the works, many people assume that even if these matters are of bibliographical significance they are not of textual importance and that editors of such texts would therefore find nothing relevant in the extensive discussion of recent decades concerning the accidentals in literature of the last five hundred years. At first glance there would seem to be good reason for this position, when one considers, among other points, that, although punctuation was in use from at least the fourth century B.C., the extent of its use is a matter of considerable debate; that in ancient times texts were normally written as a continuous series of letters, without spaces to separate words; and that in both Greek and Latin there was flexibility in spelling in certain periods.[22] Thus both the possibility of restoring to a classical text the spelling and punctuation of its author

and the desirability of restoring certain features in use in the author's time (such as the lack of word-division) are seriously to be questioned. This conventional position is expressed by West in his manual when he says, "The critic is at liberty . . . to repunctuate, even if he has taken a vow never to depart from the paradosis" (p. 55).[23] Later he repeats, "Careful thought should be given to punctuation, which can be a great help or hindrance to following the author's train of ideas, and which is of course entirely a matter for the editor's discretion" (p. 69).

The issue is not as simple, however, as these statements, taken out of context, would suggest, and West's own discussion raises some of the considerations that link so-called accidentals with meaning and with the author. West recognizes that "in theory an accent or a breathing in a medieval copy of a post-Hellenistic writer might go back to the author's autograph"; but he goes on to say that in most cases "all such features of the tradition will represent some later person's interpretation of a text consisting of virtually nothing but a continuous sequence of letters" (pp. 54-55) and that the textual critic is also "at liberty" to "reinterpret" the text in this respect. The question of how, or whether, to divide a continuous text into separated words is one that editors of modern works do not have to face, and it has therefore not been included in those editors' discussions of accidentals. In one sense word-division does fall into the group of features sometimes classed as "accidentals," for it is a matter of spacing and not of what letters are present. But obviously there may be ambiguous spots in an undivided text, where the letters can be formed into more than one set of words that make sense in the context, and matters of wording are usually called "substantive." The point is not what label ought to be used but the fact that the distinction between substantives and accidentals involves form, not meaning.[24] West further underlines the connection between marks and meaning when he says that, in the case of a nonsense word, "accents etc. may be valuable clues to what lies behind it, since they must have been supplied when the text was in a more intelligible state" (p. 55). Words that are not nonsense, however, may still be wrong (as West recognizes elsewhere), and accents or punctuation may provide clues anywhere in the text, not just where the text fails to make sense. Similarly, scribal

abbreviations in manuscripts are not always unambiguous; and though West is right to caution that "abbreviations are not actually misread as often as some ingenious emenders think" (p. 28), editors should remember that abbreviations do sometimes affect substance as well as form. The same points are also applicable to spelling. As West realizes, spelling variants "can be of use (though not by themselves) in working out the details of a stemma, and they are not uninstructive in themselves" (p. 66). This view is not always accepted, even by persons who emphasize the breadth of significant evidence; thus Willis claims that "any variant other than the purely orthographical may be significant, however trivial it may appear" (p. 36). For Dearing spellings, and the other accidentals as well, enter into bibliographical, but not textual, analysis: "the textual analyst," he says, classifies variations as "substantive, quasi-substantive, and accidental and ignores the latter class" (p. 34); later he says that an editor who concludes that scribes or compositors were following copy "even in accidentals" can "make a bibliographical analysis in which he includes accidental variations with the rest of his evidence" (p. 154). The idea that accidentals partake more of the nature of physical than of textual evidence fails to give them their due as elements of meaning in a text; indeed, the interpolation of a class of "quasi-substantives" between substantives and accidentals shows that accidentals are being thought of here as not involving meaning. However, they are indisputably part of the text and may affect the meaning of the text at any time; whether or not they do in a given instance[25] has no bearing on their classification as accidentals. I am not suggesting that scribal punctuation, abbreviations, and spelling necessarily ought to be preserved in a critical text, but they should certainly be taken seriously as part of the textual evidence present in manuscripts; they may point to how the text was understood at a particular time and may therefore—like the words themselves—be a link to a tradition and help to establish the author's meaning.

The attention one pays to accidentals, in other words, goes beyond the question of whether they reflect authorial or scribal practice. It is no doubt true that the accidentals in surviving manuscripts of classical texts exhibit more alterations by scribes than do the words, and equally true that editors have more basis for attempting to establish authors' wording than punctuation and spelling. Nevertheless, scribes do alter words as well as accidentals; the distinction is one of degree, and the texts one has to deal with contain both words and accidentals. If one decides in a given case that the accidentals of the manuscript tradition have no authority,

the decision results not from theory but from the circumstances of the individual situation. To agree with Greg that people involved in the transmission of a text have generally felt freer to alter accidentals than to alter substantives is not to say that accidentals in scribal copies are necessarily unauthoritative, though it does provide a basis, when the evidence warrants, for treating accidentals differently from substantives. The editor is finally responsible for establishing both substantives and accidentals, and to assume that scribal accidentals are too far removed from the author's practice to be worth preserving is to ignore the connections between accidentals and meaning. One may not in the end accept those accidentals, but the question of what accidentals to include in a critical text must be faced. Authorial accidentals in ancient texts may be more conjectural than in modern texts, but the attempt to approximate them is not necessarily to be rejected in favor of standardized spelling and modernized punctuation. West's discussion takes this point into account: Presumably this "regularizing" would be to the practice contemporary with the author, insofar as it is known. West also takes up briefly the question of variations in spelling in Greek and Latin: in contrast to early Greek, he says, "In Latin there is not the problem of different alphabetic systems, but notions of the correct way to spell things were more fluid until the first century of the Empire, and here again (though with less justification) the convention has been established of presenting authors at least of the late Republic in the orthography of a somewhat later period" (pp. 69-70). This matter is not pursued, but a doubt about the convention has been registered. The point he proceeds to make about still later texts in which "it is often impossible to distinguish between the barbarisms of copyists and those of the original" is illogical but instructive: "In this situation, rather than impose a consistent system which can only be chosen rather arbitrarily, it is better to follow the paradosis, not under the delusion that it is at all reliable, but as the most convenient way of exhibiting it" (p. 70). Of course, in a critical, as opposed to a diplomatic, text, exhibiting the paradosis within the text is not in itself a virtue, and the accidentals of the paradosis would properly be retained only if there is a chance of their reflecting authorial practice more accurately than another system. But West's concern for establishing accidentals contemporary with the author and possibly retaining the accidentals of the manuscript tradition is uncommon in this field; though his discussion is undeveloped, it implies a greater link than some have imagined with the approach to accidentals followed by many editors of post-medieval work. Indeed, one must ask whether for a work of any period there is ever a justification, from a scholarly point of view, of any aim regarding accidentals other than the reconstruction of the author's own practice; however imperfectly that aim may be realized in many instances, it is the only aim consistent with the view that accidentals are integral to a text and that modernization therefore has no place in scholarly editing. Without underestimating the differences between ancient and modern texts, one can see a common issue here; West's brief remarks would strike an editor of modern literature as bordering on familiar territory.[26]

These considerations have direct implications for the apparatus. Since accidentals can affect meaning and scribal practices in accidentals can constitute important textual evidence, a complete recording of such details, as well as of substantive variants, would seem to be desirable. Selectivity, of one kind or another, has been the rule, however, in apparatuses for early texts, though disagreement has existed about the principles of selection. Willis approvingly claims that it is "a matter of common consent that purely orthographical variants should be excluded" (p. 35). His statement is disproved by West, who finds it "advisable to record orthographical variants fairly systematically, at least for portions of the text," and who further implies their significance by holding that, if one decides not to record "certain orthographical trivialities," "the fact should be stated" (p. 66).[27] Of course, printing costs

sometimes dictate selectivity. Even some of the CEAA/CSE editions of nineteenth-century authors, which always contain complete lists of editorial alterations in the copy-texts, exclude accidentals from the lists of variants among other collated texts. But it is one thing to be selective as a result of believing that completeness is unimportant or even undesirable and quite another to recognize selectivity as a matter of expediency only—for in the former case one is likely to think that subjective notions of significance are sufficient for determining what to include, whereas in the latter one will wish to define the categories included or excluded as objectively as possible. Willis advocates (in addition to the elimination of spelling variants) the exclusion of certain manuscripts, but on grounds that grossly misunderstand the purposes of an apparatus: "An apparatus criticus," he maintains, "which tries to use too many manuscripts is liable not only to be obscure and hard to use (the problem of finding enough sigla is not the least), but to be inaccurate, since the task of collating accurately some thirty or forty manuscripts is enormous for one man, and to find and organize reliable helpers is scarcely less difficult than doing it all oneself" (p. 43). It seems hardly necessary to reply to these objections, for the difficulty of using a complex apparatus is of no importance to a person who values having the evidence recorded (the designation of sigla is surely a trivial matter), and the attainment of accuracy is always a challenge, regardless of the scope of the coverage (one wonders how scholarship would ever progress if this challenge were allowed to inhibit activity). Willis's statement that "About ten manuscripts should be enough to set up the text of a Latin author" (p. 42) may be accurate on the average (I am not questioning Willis's knowledge of his field), but it seems unwise to prescribe how much evidence will be significant. In justifying a selective approach, Willis enunciates the principle that it is preferable "to give all the readings of some than some of the readings of all" (p. 44). This particular point can be (and has been)[28] legitimately objected to, but the motivation behind it—to find an objective basis for selectivity—is one to be taken seriously. Dearing, thinking along the same lines, says, "The critic should define the interesting variations as precisely as possible so as to be consistent and accurate himself and to facilitate consistency and accuracy in the work of his staff or associates" (p. 147). There is no doubt that the number of manuscripts available in some cases, especially in biblical studies, is so large as to make impractical the goal of recording everything in all of them (or even the examination of all of them). But whenever one finds it necessary to be selective—either in the texts to be covered or in the details from those texts to be recorded—one should remember that no amount of rationalization will conceal the fact that the resulting apparatus is a compromise and is less satisfactory than a complete record. In such cases one owes the reader the courtesy of an unambiguous definition of what is included (not just some such statement as "the most significant readings"), so that the reader will be in the position of knowing precisely what kinds of evidence must be looked for elsewhere.[29] These questions concerning the construction of an apparatus thus raise issues relevant to all textual scholarship, regardless of the period involved or the nature of the surviving textual witnesses—as indeed, I have tried to suggest, do many other questions relating to other aspects of the editorial process.