University of Virginia Library

Copy-Text and Its Variants in Some Recent Chaucer Editions
by
Joseph A. Dane

The purpose of the present paper is to examine the use of both the term and concept "copy-text" with reference to some recent Chaucer editions. The paper will be in two parts: the first deals with the concept of copy-text in general, based on a conservative reading of Greg's definition; the second deals with the use of this and related terminology primarily in the recent Variorum Chaucer volumes, the Riverside Chaucer, and Blake's edition of the Canterbury Tales.

Greg's Notion of Copy-Text

In a recent book, Jerome J. McGann gives what seems to be a standard and unproblematic definition of copy-text: "In the post-Greg context, the term signifies what an editor chooses to take as the text of highest presumptive authority in the preparation of an eclectic, or critical, edition. . . . The copytext serves as the basis of the critical edition that is to be produced."[1] This definition is a clear one, but McGann associates the term with specific editorial procedures different from those assumed by Greg himself. As I shall discuss below, to invoke a copy-text in McGann's sense (with its reference to Greg) is to invoke potentially competing editorial theories. For Chaucerians, the problem is compounded by the assumption that medieval editors (scribes) and modern editors are analogous, and one Chaucer editor has used the term to mean what a medieval scribe (rather than a modern editor) might work from. N. F. Blake refers to the hypothetical exemplar for the Hengwrt manuscript of the Canterbury Tales as follows: "That all MSS are ultimately dependent upon Hg's copy-text will guide editorial practice; for it presupposes that there was only one copy-text."[2] That such uses of the term can be misleading is a point I shall be arguing in both sections of this paper. Here it is enough to note that what McGann and Blake refer to above as "copy-texts"

could be just as accurately and unambiguously referred to as "base text" and "exemplar" respectively.

Greg's definition differs from the understanding of the term both by his predecessors and by his followers. The copy-text is not necessarily (in McGann's words) "the text of highest presumptive authority." It is, rather, the version of a text the editor chooses to follow for "accidentals" as opposed to "substantives":

Substantives are lexical and grammatical elements; accidentals are what Greg calls "formal matters" (p. 385; the term "material matters" might paradoxically be more accurate). These include spelling and punctuation.[3] Thus the copy-text for Greg provides "guidance" in the editor's representation of accidentals in an edition (p. 384); it provides formal standards (e.g., spelling conventions) for the substantive changes an editor introduces ("editorial emendations should be made to conform to the habitual spelling of the copy-text," p. 386). But it also has a second function, not explicitly mentioned by Greg but certainly assumed, which is to serve as a "basis of collation."

In most practical instances of editing, the copy-text might well be accorded authority in substantive matters, and under certain editorial methods, it would necessarily have such authority. But an exemplar's status as copy-text has nothing to do with its potential authority on substantives, and on this Greg is explicit:

Greg's parenthetical "(generally)" is worth noting. So reluctant is he to accept the authority of any single exemplar, that he allows the copy-text itself to be corrected in the matter of accidentals or even disregarded: This implies that the copy-text can be an abstract rather than a material thing. For medievalists, this possibility would bear largely on questions of spelling and normalization, and would be of no more interest than Greg seems to give it. But the implication that the copy-text can be an abstraction realized only as an editorial construct has been more fully exploited in other areas (Gabler's Ulysses is an obvious example).[4]

Greg's distinctions have different value for the editing of texts from different periods. For an editor of classical texts, Greg's discussion is only partially applicable: classical editions are generally normalized. For most Greek texts, normalization to medieval standards is simply conventional; for classical Latin texts, the standard modern system of normalization is considered more representative of authorial spelling than what is found in any extant medieval exemplar. In either case, most accidentals are determined by the particular conventions of spelling the editor adopts. Once the editor has determined the system or rules governing accidentals, the only editorial decisions deal with substantive matters (lexical and grammatical), which, when combined with the system of normalization governing accidentals, will produce a normalized orthography and punctuation. Editorial decisions on punctuation (a period? or semi-colon?) must still be made, but such decisions regarding particular accidentals are to be made on the substantive level (grammar, lexicon) or even on a thematic or aesthetic level (theme, tone, etc.). What a classical editor might call a "copy-text" will thus not be selected for its presumed authority on accidentals. If one of its functions is to provide a basis of collation (or a set of preliminary line numbers) there might well be reason to choose as copy-text the textus receptus, however corrupt, or even a recent edition. But to call such a text a "copy-text" in Greg's sense would be misleading.[5]

Greg's article was speaking specifically to the problems associated with fifteenth- and sixteenth-century texts (p. 378). As in the case of classical texts, substantive matters can here be separated from accidental matters. But the editorial situation differs from that faced by the classical editor in two ways: (1) no standard system of punctuation and spelling exists, and (2) the earliest manuscript might well be contemporary (or nearly contemporary) with the author and thus could reflect authorial accidentals with some accuracy. Editors

of more modern texts face a different situation: the earliest edition is generally contemporary with the author and later editions may well be revised by the author. Thus the question of choosing a copy-text for the editor of nineteenth-century texts tends to involve substantive matters.[6] The editors of medieval texts draw on textual-critical theories and language from all these fields; but their situation is also different. They will not admit casually a modern system of normalization as do classical editors and as did the earlier editors for such series as the Société des Anciens Textes Français; but they are equally reluctant to accept the system offered by any single manuscript source unless that manuscript is also given credit for "presumptive authority" on substantive matters.

Base Text and Best Text

Before proceeding to some of the implications of Greg's theory and finally to the problems of medieval editing, "copy-text" needs to be distinguished from related editorial terminology, "best text," "base text," and such nontechnical terms as "basic text." The differences are not simply matters of definition. The terms "best text" and "base text" imply specific editorial procedures quite different from those implied by the terms "copy-text." The non-technical "basic text" owes its utility to the very absence of a restrictive definition.

Greg, as often noted, assumed a genealogical method of editing, and it was that system to which his terminology applied. Even though the editing of modern texts employs different methods, most of the interesting and productive theorizing on Greg has been based on situations where the language of the genealogical method still has some application. For example, the difference between early and later printed editions of a text (and consequently the choice of which to use as the basis of an edition) could be described as one of simple filiation, involving a single line of descent complicated by authorial variants. But medieval editing almost never confronts such a situation (most of the manuscript evidence post-dates the author), and the definition of "copy-text" in terms of an abstraction such as "authorial intentions" could apply to very few editorial situations. For medieval editions, the language of one editorial method is less easily transferable to another.

The types of editorial procedures implied by these terms are various, but in Chaucer editing, the three basic types of edition defined some eighty years ago by Eleanor Hammond can be used as a starting point: (1) the exact reproduction of single manuscript (Wright's 1848-51 edition); (2) eclectic (the editions of Tyrwhitt 1775-78 and Skeat 1894, 1899); (3) critical. By "critical," Hammond refers to a recension (or genealogical) edition; in 1908, there were

no full-length Chaucer editions of this type, but the most notorious later attempt at such an edition is the Manly-Rickert.[7]

The three terms "best text," "base text," and "copy-text" can be matched with these three types of editorial procedure. Hammond's first type of edition relies on a single exemplar; and this exemplar is often called a "best text." The word "best" may be ill-chosen, since a "best-text edition" could certainly be made of any manuscript, even a manifestly inferior one (for this reason, I shall refer to so-called "best-text editions" as "single-text editions" below). But if the term is used, it implies a specific editorial theory or procedure.[8] In medieval studies, a so-called "best text" is simply the exemplar followed conservatively in a single-text edition. The term "base text," by contrast, refers to the exemplar(s) on which an eclectic edition is based. Such a base text might also be called a "foundation text" or "basic text"; the advantage of these latter terms is that they do not seem to have developed technical meanings or implications. In practice, an eclectic method would call for the use of a particular exemplar even if only as the base in which to admit corrections from a number of other sources (e.g., Skeat's use of El for his Canterbury Tales); in early editing (and even in some recent editing) the base manuscript might be an earlier edition (e.g., Tyrwhitt's apparent use of black-letter editions).[9] A "critical" edition in Hammond's sense (a recension or genealogical edition) would not necessarily have a base manuscript, since all manuscripts might be of equal authority; but it must have a copy-text or at least something to serve the various functions of a copy-text. That one of these functions in most practical editorial situations is to provide a basis for collation is generally simply assumed; McGann is one of the few textual-critical theorists to make it explicit (Critique, p. 24). In addition, that critical edition must take its spelling conventions from somewhere, since the genealogical methods that lead to a substantive authorial reading do not lead to the author's conventions on accidentals (this, of course, is on the assumption that scribes and early publishers distinguished substantives from accidentals as we might, and further that they felt responsibility only to retain the former as authorial). According to Greg, a copy-text can be chosen "irrespective of descent" (and thus irrespective of its authority on substantive matters). The exemplar Greg selects as copy-text in his edition of the Antichrist Play from the Chester Cycle is the earliest extant, but not the highest in the stemmata; that is, authority

on substantives (genealogical priority) is independent of its authority on accidentals (here a matter of chronological priority).[10]

Greg's theory of copy-text deals specifically with genealogical editions. Thus, an edition that relies on a copy-text in Greg's sense does not necessarily give what McGann calls "highest presumptive authority" to a single manuscript or exemplar. An edition that does so rely on a single exemplar may characterize it more usefully as a "base text"; if this exemplar has even greater authority (overriding or preemptive authority), the editors are producing a variant of a single-text edition and can then legitimately refer to this exemplar as a "best-text" (see, however, n8 and discussion above).

Although there can be no justification for calling a base text a copy-text, there are still advantages for retaining the term, even in editions that do not use the genealogical methods of Greg. Under any editorial method, an exemplar can be copy-text in Greg's sense if it serves as an authority for accidentals and (as a practical matter) a basis for collation. The use of the term should force an editor to describe editorial procedures and in particular to articulate the nature of the authority possessed by an exemplar or manuscript. Surely a copy-text can serve as base text, and in a single-text edition the copy-text is generally best, base, and copy-text.[11] But the choice of a base text does not mean that the question of copy-text is closed; in addition, the choice of an exemplar as a basis for collation and an authority for accidentals (a copy-text) does not mean that a base text or best text must even exist.

There are further implications to Greg's theory; under a perhaps overly literal interpretation, the copy-text could be a text of some other text. (I am going to reject such a use shortly, for obvious reasons, but the theoretical possibility of it should be reckoned with.) For early modern texts, this theoretical possibility poses few practical difficulties. If a fifteenth-century text existed only, say, in an eighteenth-century print, it is difficult to imagine why an editor would wish to produce an original spelling edition or how that edition could be justified. But the edition would certainly be possible. To produce it, an editor might rely either on a selected system of normalization as do classicists or might choose in lieu of such a system of normalization another text, one that would be ignored in all substantive matters. In classical editions, a major function of the copy-text is served by the text (a dictionary or perhaps better a school-grammar) that contains the spelling and punctuation conventions the edition follows. As for the basis of collation, any earlier edition (or translation) can serve as well, whatever its authority; even a list of line numbers

could conceivably serve this function.[12] In these cases, there again is no reason to invoke Greg's copy-text, since all the explicit functions of a copytext are served by texts with no authority in substantive matters. Greg's repeated denials that the copy-text has authority on substantive matters somewhat paradoxically implies that it has at least potential authority in such matters. And it is this potential authority that modern textual-critical theorists have strengthened by referring to "presumptive" authority.[13]

In many cases, the term copy-text might well be avoided, and its potential functions defined and dealt with separately. There is clearly no reason for a classicist to speak of a copy-text in relation to normalization. Nor is there any advantage to using the term if all it refers to is a "basis of collation"; the more explicit term is preferable. Moreover, a system of normalization or basis for manuscript collation that does not contain a version of the text to be edited cannot usefully be called a copy-text even if it serves the same function.[14]

Greg's article was a reaction against what he called "the tyranny of the copy-text" (p. 382). It was an attempt in part to reduce the functions of the copy-text, taking away from it the substantive authority that modern textual critics have begun to restore. A return to Greg's definition of the term might define the copy-text out of existence in many editorial situations. For medievalists, this might not be a bad thing. In practice, the difference between copy text, base text, and best text involves the relative authority granted a certain exemplar; the difference could be considered one of degree. But keeping the theoretical distinctions in view would lead to a more accurate assessment of that authority. Furthermore, a more conservative use of the term would avoid the confusion between the imagined tasks of a medieval "editorial office" and the real tasks of a modern one. Since we do not know in most cases the precise procedures or theories a medieval editor followed,

there seems little point in describing them with technical vocabulary developed to apply to the twentieth-century editor.

Recent Chaucer Editions I (Single-Text Editions)

One of the more striking aspects of recent Chaucer editions is the privileging of the Hengwrt manuscript (Hg) for the Canterbury Tales. Among the editors to have done this are Pratt, Blake, those involved in the Oklahoma Variorum project, and even Donaldson in his earlier normalized edition. The justification for the reliance on Hg is generally claimed to lie in the Manly-Rickert edition of 1940.[15]

The problems of using Manly-Rickert (a genealogical edition) in support of a single-text edition based on Hg have been pointed out before.[16] Manly-Rickert's prefaces are often baffling, and the varying stemmata constructed never show Hg in a position of supreme authority for O' (the supposed common ancestor of all manuscripts that Manly-Rickert attempt to reconstruct). The conflation of competing and often antithetical editorial theories has led to confusion, both in the methods themselves and in the language used to describe them (e.g., best text, base text, copy text, basis of collation).[17]

Such conflation seems to be acknowledged in the Editor's Preface of the 1979 facsimile—the first volume produced by the Variorum Project:

According to this, a best-text method is used to provide a text to serve as the basis of commentary; but the supposedly "neutral" text that results is supposedly one that cannot be improved, that is, the best-text method yields the best edition.[19]

The Variorum Chaucer has dual purposes, and these lead to contradictions (both in tone and in substance) in the prefaces. In the General Editors' Preface (I quote here the version printed in Ross's Miller's Tale), the editors say that their purpose is "only to provide a text upon which the commentary should depend" (p. xv). But the conflicting claims of the 1979 Preface (to produce the best possible edition) are scattered through each volume. In his own introduction, Ross (perhaps following Pearsall p. 97) states: "The text of The Miller's Tale in this edition is in one way more ambitious than is the monumental work of MR. . . . The Variorum Edition may thus present The Miller's Tale as Chaucer wrote it, as nearly as our present knowledge and resources permit" (p. 61).[20] These inflated claims and attendant rhetoric are occasionally repeated in reviews. According to one reviewer, a recent Variorum editor gives "all the evidence necessary for establishment of a text which would probably be as near to the original as present knowledge and scholarship could make possible."[21] Pratt's earlier edition makes similar claims: "the present text represents as accurately as possible Manly's 'latest common original of all extant manuscripts' (O'), with the correction of all recognizable errors in the transmission to O' of Chaucer's own text (O). . . . In attempting to recreate the text as Chaucer wrote it . . ." (p. 561).

The arguments of Pratt and the Variorum Editors seem to assert that Manly and Rickert's reconstruction of the latest common ancestor (O') of all MSS is itself not in question. All that remains to do is to correct the "manifest errors" in that reconstructed ancestor and we are as close to Chaucer's text (O) "as it is possible to get."

But how can Hg be used for what precedes Manly-Rickert's O' when O' is it itself constructed in part on the basis of Hg? The argument for this depends on a serious misrepresentation of Manly-Rickert's methods; Pearsall's statement is an example:

The manifest circularity of the first part of this statement is not at issue here. What concerns me is only the failure to distinguish a "basis for collation" from "base text." Pearsall's reasoning, in a single-text edition, conflates the language of two competing methods. Manly-Rickert used Skeat's "Student Edition" as a "basis of collation" for their recension edition; Skeat's edition is itself based on El (it is eclectic). But Pearsall implies that Manly-Rickert took Skeat as their "base text," emending it in the direction of Hg; that is, he argues as if they were producing a different type of edition.

Manly-Rickert, in the section entitled "Manner of Collating" to which Pearsall refers, discuss only the method of collating manuscripts and the mechanical means of recording variants; as a "basis for collation," they used Skeat's Student's Edition (2:5). Its function was only to collate manuscripts and to aid in the construction of lemmata. The readings in that edition are of course irrelevant and unrecorded. As a text, it has no more authority than a translation, which could have served the same function. To ignore this is to assume that Manly and Rickert, whatever their failings as editors, after examining and describing all the Canterbury Tales manuscripts, did not recognize the difference between a manuscript authority and a modern edition. I am not certain what Pearsall means by his statement that the Manly-Rickert edition "moves consistently from the text used as the basis for collation . . . toward Hg." But if all this means is that in cases where Manly-Rickert differ from Skeat they tend toward Hg, I see nothing surprising in that. Had they used Hg as a basis for collation, similar results might have obtained. In cases where they differed from Hg, they might well "tend" toward something else, perhaps El, perhaps even Skeat.

A basis for collation is something used to collate manuscripts and produce lemmata, not a "base text" for an edition. As the Variorum Editors recognize, "The decision about what is a lemma is, of course, purely arbitrary" (Ross, p. 52). To choose a version of the text to be edited is wise from an economic standpoint only, since it would be tedious to set forth every manuscript reading as a variant of an arbitrary lemma. But two sets of lemmata must be distinguished. The preliminary lemmata produced while collating MSS (defined as variants of a "basis of collation") are not the same as those listed in the notes to an edited text (defined as variants of the edited text).[22]

Manly and Rickert use a genealogical method, and as such, they have no base text at all. Pearsall acknowledges this, but then describes Manly-Rickert's "basis for collation" as a "copy-text": "It is noteworthy, therefore, that MR, though they use no base manuscript (the copy-text is SK), draw frequently toward Hg and away from El in their choice of readings" (p. 122).[23]

To speak of a copy-text for Manly-Rickert is misleading and unnecessary, even though certain texts can be identified as serving functions associated with a copy-text. The Student Skeat operates as a basis of collation only (it does not even provide line numbers). For matters of spelling, the function of copy-text is served by a system presumably based on a comparison of Hg and El:

In reference to the Hg-based texts themselves (the Variorum and the editions of Blake and Pratt), the notion of "copy-text" should be merely redundant (thus unnecessary), since "copy-text" is simply subsumed under the notion of "base text" and occasionally "best text." The usual way the Variorum Editors speak of Hg is as a "base text" (the Variorum is "based on/upon Hg."[25] But the term "copy-text" is sometimes used as a variation: "On the other hand, Hg omits two couplets, both of which are included in the present

edition, though enclosed in brackets to indicate that they are not in the copy-text" (Ross, p. 54). Moreover, "copy-text" is also used to mean "the exemplar for a specific extant MS": "[Hg and El] were written from different exemplars at different times. . . . El's copytext had two extra couplets, which may have been Chaucer's . . ." (ibid.).

There is no question that Hg is the "copy-text" for the Variorum Edition, but to speak of it as such is merely to invoke textual-critical language that applies to a different editorial situation. Hg's function as copy-text is trivial, since it is also the base text and for these editors the best text. The Variorum is a simple variant of a single-text edition; Hg is "conservatively emended" from a number of manuscripts, selected on the basis of Manly-Rickert's groupings.[26] It thus has the potential for incorporating not only the virtues of the genealogical, eclectic, and single-text methods but their failings as well.

The arguments of N. F. Blake for the privileging of Hg are similar in many respects to those of the Variorum Editors. But Blake's editing theory gives greater authority to Hg, and provides as well a dynamic model of manuscript exemplars that complicates the entire enterprise of producing a static (i.e., printable) edition.[27] Blake's edition is, like the Variorum, a single-text edition, although Blake refers to Hg as "base MS": "in the light of our present knowledge it is safest to edit the poem . . . using Hg as the base manuscript and excluding anything not found in it" ("On Editing," p. 111). Blake acknowledges the convenience of an assumption of strict linear descent of MSS, an assumption that would turn any manuscript into an absolute authority for all posterior readings: "If we accept that there is a manuscript tradition which goes back to one manuscript, Hg., then there are three possible ways to edit the poem" (p. 105). Blake's purpose is to discard the notion of "authorial variants" and thus to simplify the editorial process; the assumption of lineal descent of all MSS from a single manuscript is a convenient polemical position. But the assumption Blake seems to have made is less radical. Blake assumes the descent of all MSS from an exemplar copied by Hg: "That all manuscripts are ultimately dependent upon Hg's copy-text will guide editorial practice, for it presupposes that there was only one copy-text" (p. 112); "later scribes used Hg's exemplar rather than Hg" (p. 113). This assumption, of course, challenges the absolute authority of Hg, since it acknowledges other lines of descent from O' (i.e., radial descent, rather than linear descent). If this is the case, Hg has no more authority a priori than any other MS., a difficulty Blake tries to overcome by allowing that other MSS may "suggest . . . how Hg may be emended or corrected" (p. 119), leaving open the question of whether they can do so with any authority.[28]

But let us look here at Blake's notion of copy-text, by which he means

the exemplar for Hg, "Chaucer's working copy" ("On Editing," p. 115).[29] Blake claims that this hypothetical exemplar was constantly revised in an "editorial office" (p. 115). As does Pearsall, Blake uses "copy-text" to mean both an editor's copy-text as well as an historical exemplar for an extant version of a text. This notion of copy-text is part of further terminological slippage: "But if the exemplar (i.e., Chaucer's own fragments) was being constantly emended in the editorial office, the good text would gradually disappear under a host of corrections" (p. 116). Whereas Manly-Rickert studied the extremely complex relations among real manuscripts in search of a singular hypothetical origin (O'), Blake reverses this logic, dismissing all complexities among real manuscripts as meaningless, and hypothesizing instead an equally complex (and less well-documented) history that produces the single extant manuscript Hg. Thus, the problems associated with the editorial copy-text (problems the very nature of a single-text edition should solve) are reintroduced on a hypothetical, historical level, where such terms as "exemplar," "Chaucer's own fragments," "copy-text," and "good text" exist in some uneasy equation.[30] Blake's apparent reliance on a single authority, an assumption that should simplify matters, in fact hypothesizes a situation (a medieval editorial office) in which no single authority seems to exist or can be articulated much less recovered. Thus Blake can dismiss as "the rather uncertain art of literary criticism" ("On Editing," p. 103) all attempts to recover editorially more than a group of Chaucerian fragments.

Chaucer Editions II (Eclectic Editions, Riverside Edition, Windeatt's Troilus)

Other recent Chaucer editions have been eclectic, and the language describing them is varied. Fisher, in what is primarily a student edition, uses the language from a number of textual-critical schools:

Fisher is clearly producing an eclectic edition using a base manuscript. There is no reason to refer to copy-text at all, either to describe his own or other editors' procedures, although the distinction Fisher implies here is that a copy-text in his sense (the status accorded Hg by other editors) has more "presumptive authority" than a base text.

Of more concern to me is the Riverside edition—an edition that like the Variorum attempts to serve a number of different purposes, some of which may be incompatible. The title page claims it is The Riverside Edition, Third Edition, "based on" the second edition of Robinson.[32] That bibliographical ambivalence is a reflection of the uncertainty and often contradictory nature of editorial procedures. Robinson has always been perfectly serviceable as a student edition, and the Riverside attempts (successfully) to maintain that serviceability. Yet Robinson's editorial procedures have been so often questioned that a more radical revision would certainly be required to maintain its status as a scholarly text (or reference text). The Riverside editors have not decided whether to depart absolutely from Robinson, and the result is that Robinson often functions as copy-text and perhaps as base text. The edition that by its very existence should supersede the authority of Robinson's earlier editions has paradoxically transformed Robinson's earlier text into a textual authority.

The Riverside edition and its individual editors have responded rationally to the problem of updating a standard edition. Individual editors are not forced to adopt a single system, nor to proclaim an unlikely unanimity on editorial procedures. The relative clarity of the descriptions of editorial procedures may well be a consequence. In general, the editors avoid the issue of copy-text, and speak in a non-technical vocabulary. An exception is the preface by Hanna and Lawler on Boece, where a very accurate indication of the opposing functions of base text and copy-text is implied: "The work has been previously edited, always with C1 or C2 as base. . . . In this edition we follow C1 as copy-text. We chose this manuscript because it is complete, tolerably consistent in its spellings, and one of three manuscripts most faithful to O'" (p. 1151). Following Greg's distinctions, C1 is chosen not because of its presumptive authority on substantives, but rather because of its accidentals (its spelling system) and its completeness. Elsewhere, the Riverside editors tend to avoid the term, even when they are warranted in using it: John Reidy, editing the Astrolabe, attempts to "establish an archetype" (p. 1194), and selects a MS (B11) specifically on the basis of its spelling conventions: Reidy does not refer to this as a copy-text, although it is so precisely

in Greg's sense. John H. Fyler's edition of House of Fame is based on different editorial methods, which, however justifiable, are clearly described: "I have made only a few changes in Robinson's text. . . . The many departures from the base text, F, . . ." (p. 1139).

More complex is the text of the Canterbury Tales; here, the eclecticism of the Riverside shows to advantage. In comparison to Boece, the Canterbury Tales is hardly edited at all; but Hanna's notes make no claim to the contrary:

In this straightforward, non-technical paragraph (itself in contrast with the trenchant description of the earlier history of editions in the same section), Hanna acknowledges that Robinson functions as base text. More important, he proves that it is still possible to produce a serviceable edition without reliance on a sudden and remarkable editorial consensus.[33]

In the textual notes to Troilus, written by Stephen A. Barney, Robinson seems to serve a different function: that of copy-text. Barney begins with a statement of editorial consensus: "Windeatt largely agrees with Robinson, Pratt [Pratt previously made "much of the analysis of the variants and many decisions about authentic readings"], and me about the appropriate methods of establishing the text, and for that reason Windeatt's text and this one differ little in substantial matters" (p. 1161). The base text is Cp: "The text here presented, like Robinson's (and Donaldson's, Baugh's, and Windeatt's) is based on Cp. When Cp is rejected or deficient, this edition prints the readings of Cl or J, in that order" (p. 1162). But Barney's edition also makes use of a copy-text: "The present edition is based on microfilm and other photographic copies of all the authorities, supplemented by reference to printed editions and discussions of the text, primarily Root and Windeatt. The goal has been to adopt the forms of Robinson's text, which is sensible and intelligent, while reconsidering "from scratch" the readings . . ." (p. 1161).[34] Barney's

first claim makes an apparent distinction between "forms" and "readings." This is, I think, equivalent to Greg's distinction between accidentals and substantives (punctuation is not at issue here). What this statement implies, then, is that the copy-text (in Greg's sense) for this edition is Robinson's second edition. Such claims elevate Robinson to the status of independent authority on Chaucer's use of accidentals. Where Robinson himself spoke of his spelling system as one of normalization, that vocabulary has now disappeared.[35]

Earlier, I noted that there was simply no point in calling a "system of normalization" a copy-text, since such a system did not have to exist as a version of the text to be edited (see Bowers, "Regularization and Normalization," and above n. 14). The reason for that is obvious. Robinson's system of normalization is not simply that found in his text of Troilus but one that he constructed from his experience with Chaucer's manuscripts and his knowledge of standard descriptions of Middle English grammar. There is no reason for Robinson to speak of this as a copy-text, since among Chaucer texts it is represented only by his version. Barney disguises that system by allowing it to intrude into the text as if it were represented in a medieval copy-text. And for that reason, it would not only be legitimate for Barney to speak of Robinson as copy-text, but also advisable, since such terminology would warn readers of the extent to which Robinson's text serves as authority.

Windeatt's Troilus has a much different look, due in part to format (the printing of Boccaccio's Filostrato in a facing column, the double column of notes), and in part to Windeatt's decision to represent initial capital F graphically as ff.[36] But Windeatt also wishes to present a different type of edition:

Manuscripts, thus, are not to be construed necessarily as evidence of authorial intentions, but rather as evidence of audience responses. This allows Windeatt to direct his discussion away from the question of the relative authority of manuscripts and to speak of manuscript relations as "various scribal traditions of copying the poem" (p. 37). It also allows him to define a copy-text: Windeatt's naming Cp a copy-text rather than a base text (even though it arguably serves such a function) seems in line with his effort to reduce the authority of any single manuscript (representing authorial intentions) in favor of the extant manuscripts (representing the text's reception). That is, the edition is an attempt in some way to present an audience-based edition. The wisdom of this may be questioned, but it does allow Windeatt to limit the authority of his copy-text to formal matters.[37]

Yet in practical terms, Windeatt's edition is little affected by his theory. Like Blake, Windeatt simplifies editorial procedures by discounting authorial revision (in this case, the theory of three authorial versions of Troilus). Coherent authorial intentions can then be determined by manuscript relations (p. 41), and some manuscripts better reflect those intentions than others; the relative authority of manuscripts is of course implicit in his description of Cp and Cl (pp. 68-69). Windeatt's "copy-text" finally has as much authority over the substantives of the text as Barney's "base text."

Kane's Piers Plowman

Without question, the most significant recent edition of a Middle English text is the Kane-Donaldson Piers Plowman [38]—a work that will probably influence future Chaucer editors as much as any of the specific Chaucer editions discussed above. I cite this in conclusion in hopes that its editorial language will prove as influential as its substance and tone. Kane's entire enterprise is directed against the possibility of a recension edition; thus, the terminology of Greg designed particularly for such an edition is not easily applicable. The language adopted by Kane, however, is instructive. In the edition of the A Version:

T is as true a copy-text as could be possible in a non-recension edition, here serving to supply a system of regularization for accidentals. It is also a base text (although Kane does not so describe it), since the kind of edition Kane is engaged in must be called eclectic. Kane wishes to place himself in a direct line with Greg and Housman, and his primary target is the "tyranny of the copy-text"—the editorial procedure that would substitute a physical authority for editorial experience. I assume this is why he refers to T as a copy-text or (using non-technical terminology) as a "basic text." Since all changes from T are shown in square brackets, the degree to which T is the "highest presumptive authority" will reflect the editor's willingness to include such brackets in the text.

In the B Version, the textual-critical language becomes even more explicit, as does the reference to Greg:

Kane-Donaldson choose W as their "basic manuscript" (p. 216): "For one thing W's consistent spelling and systematic grammar afford a clear model for the many readings that have to be introduced into the text by emendation" (p. 215). The citation of Greg is significant, since the function of a copy-text becomes more limited as the editor's own intervention increases. The Kane-Donaldson edition is still an eclectic or base-text edition, and this is reflected in the language above. But the presumption of authority in substantive matters is secondary. Again, describing such a base manuscript in the language of Greg puts the editor under fewer constraints to follow it. And the difference between the B Version and A Version editions is in one sense a recognition of those implications.

Kane uses terminology only when its history has some import: thus his use of the term "copy-text." Elsewhere, non-technical terminology suffices (thus "basic-text" instead of the technical terms "best text" or "base text"). Furthermore, he uses technical terminology to reduce external authority, not to elevate it.[39]

Conclusion

McGann's definition of copy-text seems to be a reasonable extension of Greg's notion, and certainly is useful in practice:

The rejoinder to this argument is that it provides an excellent definition of a "base text" (implying an eclectic edition), not a "copy-text" (implying a recension edition or what was once meant by the phrase "critical edition"). Furthermore, the argument assumes that editors must choose a text as their highest authority. Yet recension editions still exist, and Kane has proved that even a base-text edition can exist without attributing undue authority to the basic text itself.

Chaucerians, however, seem to be moving toward single-text editions, and Greg's inadvertent defense of such editions can certainly be taken at face value: "what many editors have done is to produce, not editions of their authors' works at all, but only editions of particular authorities for those works, a course that may be perfectly legitimate in itself, but was not the one they were professedly pursuing" (p. 384). Editions of Hg and editions of Chaucer are two different things, and there are certainly reasons to prefer the former (economics, editorial consistency in a project involving many editors, etc.). And editors might do well to portray legitimate, economically based decisions for what they are, rather than to obscure them with textual-critical jargon. This is the approach successfully taken by the Toronto Medieval Latin Texts series. Furthermore, editorial projects are not necessarily doomed because they have multiple (and possibly conflicting) purposes. Poiron's cheap student edition of the Roman de la Rose follows a single manuscript (allowing the reader to reconstruct it) while adding in brackets the lines of the textus receptus not included in it. In so doing, Poiron can incorporate earlier editions rather than condemn them.[40]

An obvious conclusion here would be for Middle English editors to drop the notion of copy-text altogether unless they are willing to define it precisely (as, say, Greetham, "Normalisation") and to speak directly to the problem of what the accidentals provided by such a text are supposed to represent

or suggest: those of the author? or simply those of one of the author's near-contemporaries? Unless an editor is interested in grappling with such questions, I see little reason to invoke Greg's term. A statement such as "the base text for the edition is X (corrected), with forms normalized according to the edition of Robinson" makes perfect sense and can be easily justified. An edition will not (or should not) be condemned simply because it is selective in the issues it deals with. The use of the term "copy-text" for Middle English editions that are completely different from the type of edition on which Greg based his theory leads generally to confusion and to an obscuring of the often legitimate editorial procedures employed.

Notes