University of Virginia Library

The Concept of Ideal Copy
by
G. Thomas Tanselle

Whenever physical objects are produced in groups of supposedly identical exemplars, close scrutiny will inevitably disclose differences among them. Whether such differences are of any significance is a relative matter, depending on the degree of precision required for a given purpose. Some objects are made with such precision that elaborate equipment is required to detect differences, whereas other objects can serve their function well and be considered "identical" even if certain differences among them are easily visible to the naked eye. Copies of books that publishers present to the public as identical may actually vary from one another in a variety of ways, and a basic task of the analytical bibliographer is to determine what level of difference is of potential significance in a particular instance. Imperfections in the weave of the binding cloth or differences in the amount of inking of the type pages may prove to be insignificant, but the presence on some copies of a completely different cloth or the appearance in them of a difference in what is printed at a given point would deserve to be reported and analyzed by the bibliographer. A central truth that affects everything a bibliographer does is the fact that books are not meant to be unique items and are normally printed in runs of what purport to be duplicate copies. Manuscripts are by their nature unique, and anyone analyzing the physical details of a manuscript has only the one object to examine.[1] But anyone wishing to discuss the physical features of a particular edition or printing can never be content with a single copy, because one cannot know whether it corresponds with other copies. The totality of the evidence consists of all the copies; although one must frequently make generalizations without seeing every copy, such generalizations will be subject to revision as further copies are examined. In one way or another, all bibliographical

analysis has to contend with the fact that a group may not be adequately represented by any one of its members.

This point has of course long been recognized, at least since the time when the nineteenth-century incunabulists inaugurated the serious physical analysis of printed books. And it was repeatedly discussed just after the turn of the century by Pollard, Greg, and McKerrow in their pioneer work linking analytical bibliography and textual study. Pollard, for example, writing in 1907 on descriptive bibliography, recognized that one object of a description is to provide information that "may be used to ascertain whether other copies are complete and perfect."[2] At about the same time, he and Greg, in another basic article on descriptive bibliography, noted, "one of the first things that the bibliographical beginner learns is that the individuals which constitute an edition are frequently not identical,"[3] and they went on to explain how various combinations of corrected and uncorrected formes can turn up in copies of Elizabethan books. Greg, in his Malone Society Reprints (1907-35), and McKerrow, in his edition of Nashe (1904-10), reported some of the differences among copies of individual editions, and McKerrow devoted a whole section to discussing this problem in his influential "Notes on Bibliographical Evidence" of 1913,[4] which was expanded as An Introduction to Bibliography (1927). Although the term ideal copy was not introduced into these discussions as a way of defining the object of a bibliographical description, it is clear that the distinction between a description of a single copy and a standard description of an edition, based on the examination of multiple copies, was well understood. As early as 1934 Greg spoke of the "ideally perfect copy,"[5] and his work over the decades leading to the completion of his great Bibliography of the English Printed Drama to the Restoration (1939-59) shows his practice in describing "ideal" copies, even though it was not until 1959 that his introduction appeared with the formal statement

that each of his descriptions represents "a standard or ideal copy" (IV, cxlviii). Meanwhile Fredson Bowers had brought these ideas to a focus in the first sustained analysis of the concept of ideal copy, in a 1947 article[6] and then in his Principles of Bibliographical Description (1949).

In this as in other respects, Bowers's book clarified principles and procedures that had been evolving over the previous fifty years or more; the concept of ideal copy was not new, but he provided a workable formulation of it that has been standard ever since. That bibliographical descriptions must attempt to establish what the ideal copy in each case consists of would seem, at this late date, to be so widely accepted as to entail no further debate. And it would also seem to go without saying that the concept is equally crucial for the scholarly editor: if a descriptive bibliography of the work to be edited does not already exist, the editor must in effect produce one in order to be prepared for the job of editing. Yet, strangely, some bibliographies still appear that are based on the examination of only one or two copies of each item (in some cases following the erroneous assumption that more extensive investigation is unnecessary for twentieth-century books). And, similarly, some scholars still try to produce texts without engaging in much, if any, collation between copies of the same edition. The term ideal copy probably lends itself to misinterpretation, since it is not meant necessarily to imply a copy that is free from all error; but failure to comprehend the concept cannot be excused because of the infelicity of the term used to label it.

Perhaps the most flagrant recent misunderstanding of the concept occurs in an article by Lorene Pouncey, who is misled by the word "ideal" into thinking that ideal copy represents an unattainable goal, which must continue to elude us as our conceptions of the ideal go on evolving; she reaches the extraordinary conclusion that ideal copy is "an absurd notion of analytical bibliographers . . . recognizable as part of the human 'habit of perfection.'"[7] Although the extent of her confusion is obviously not shared by the bibliographical world at large, there do seem to be some aspects of the concept that genuinely require

clarification. What I shall say here should be regarded as a footnote to the basic discussion in Bowers's Principles (pp. 113-123, 404-406). But because Bowers treats the subject largely in the context of descriptive bibliography, there is some point in making more explicit its relevance to editing. Indeed, there is no better way of understanding the interdependence of descriptive bibliography and editing than by thinking about ideal copy. Whether we use that term is not important; but defining the concept and thinking through its implications are essential, because it goes to the heart of bibliographical scholarship. A renewed examination of this matter may also be opportune because of its bearing on the vexed question of the relation between library cataloguing and descriptive bibliography—a question that is now receiving considerable attention as a result of the work on the eighteenth-century short-title catalogue and the efforts to improve the standards for the machine-readable cataloguing of antiquarian books and special collections.

I

The concept of ideal copy is central to descriptive bibliography, because it is the element that distinguishes bibliographical description from cataloguing: whereas a catalogue entry, regardless of its level of detail, exists to record a particular copy, a bibliographical description aims to provide a standard against which individual copies can be measured.[8] Thus the catalogue entry is tied to the physical features of a given copy, even if some of the features are actually post-publication imperfections of that copy. The bibliographical description, on the other hand, must be purged of such deficiencies of individual copies; it must be based on an examination of as much of the evidence as possible (that is, as many copies as possible), since there is no way to tell from the scrutiny of one or two or three copies how they compare to the edition as a whole without checking a considerably larger sample of that edition. Sometimes a thorough catalogue of a collection does make reference to additional copies outside the collection, but in doing so it is going beyond the function of a catalogue; and such a catalogue is particularly useful, for the very reason that a descriptive bibliography—being based on more evidence than a single collection can provide—has a broader usefulness than a catalogue. The bibliographical description rises above the limitations of a single copy by reporting what emerges as standard after an examination of multiple copies.

This much is widely understood and indeed seems elementary. The

difficulty arises in attempting to define more precisely the rationale for making decisions about what is abnormal or defective in a given copy and what is not. Some defects seem obvious, as when a leaf present in many copies is missing from another copy. But rarely are such matters as obvious as they may seem at first. Is it possible that the leaf was supposed to be canceled and that the presence of the leaf in some copies results from an oversight? Is it possible that the leaf was to be replaced with a cancel leaf, not present in the copies thus far examined? The answers to such questions can in some instances be deduced from the content of the text on this leaf and the adjacent leaves. But the principal evidence, of course, must come from still more copies. And if no other copies are extant, conclusive answers may not be possible. As is always the case with inductive investigation, what may seem clear at one stage may be shown to be incorrect when more instances are observed. Furthermore, when there are numerous differences among copies, no single surviving copy may be free of what the bibliographer concludes are deficiencies. Thus the ideal copy presented in the bibliographer's description may be an abstraction, for no such actual copy may be known to exist. The description is a historical reconstruction, based on the evidence of the extant copies; as such it is a hypothesis dependent on the bibliographer's judgment. But to say that raises more questions than it settles, questions that must be thought about if one is to understand what one is doing in setting up a bibliographical description.

The best point of departure is the matter of intention, which the concept of ideal copy is generally thought to entail: deciding that certain copies are anomalous is equated with deciding that they do not conform to what the producer of the edition intended. Greg, defining the aim of the descriptions in his Bibliography, says, "what I have in view is not any particular copy of a book, which may have its own imperfections or abnormalities, but a standard or ideal copy as it was designed and intended by its producer" (IV, cxlviii).[9] If only an imperfect copy is known, he adds, "I give the register of such a book in the fullest form that can be inferred from the existing evidence, with due warning of course of its possibly hypothetical character."[10] That the goal is the intended form of the book is reinforced by his explanation that his collation formulas represent "issue-registers" (the arrangement of the leaves as published) rather than "printing-registers" (the arrangement

as printed). He had been even more explicit in "A Formulary of Collation," where he equates "issue-formula" with the register of signatures defining "the make-up of the book . . . as it was meant to be issued" (p. 371). Thus when a printer uses the last leaf of the last gathering for the title, with the intention that the leaf will be placed first in the book as published, the bibliographer is to record the leaf in its intended place.[11] (Of course, for books that were sold in sheets and later bound to individual order, "binding-register" would sometimes be a more appropriate term than "issue-register.") Even if all copies examined by the bibliographer retain the leaf in its original position at the end, Greg was saying, the description would still place it at the beginning, for that is clearly its intended location, and the copies examined happen to have been anomalies.[12]

Bowers, in the standard definition of ideal copy, similarly makes the printer's intention the determining factor:

In the elaboration of the definition that follows, the producer's intention is again cited: The emphasis here is on the physical form of the book, on the order and completeness of the leaves[14] and not of their content. As Bowers points out, ideal copy does not imply "freedom from textual errors, misprints, variant uncorrected formes" and does not refer to "the quality of the text in any way as a criticism of the printer's final result." Nevertheless, textual considerations are not entirely irrelevant. A textual correction that results in a cancel leaf of course affects the physical structure of the book; but textual differences may exist that are not locatable through an alteration of the structure of the gatherings. Stop-press corrections may produce (and during the first two and a half centuries of printing routinely did produce) sheets containing various combinations of corrected and uncorrected formes, and books may be made up of various combinations of such sheets; no physical anomaly would call attention to the differences in text, which would be discoverable only through textual collation. Yet corrections or revisions of this kind are also encompassed in the concept of ideal copy. According to Bowers, while "ideal copy in its true sense of physical makeup is not affected," one would nevertheless "certainly choose the details of the corrected form to transcribe or note in one's description as the 'ideal' form, listing the other as a variant" (p. 114); the standards for ideal copy apply not only to the collation formula but also to the contents paragraph, "which should contain the contents in their most perfect state," including the correct form of variant catchwords, and the like. What Bowers is saying, in other words, is that the correctness of the text is an aspect of ideal copy to the extent that corrections were actually made within a given issue;[15] but errors that were not corrected do not prevent copies containing them from being ideal copies.

A principal difficulty in defining ideal copy therefore has to do with intention, for it is unreasonable to think that a printer's or publisher's intention is to produce errors, and yet the presence of errors—not corrected at any stage in the production of an issue of a book—is compatible with the concept of ideal copy. An analogy with editing—where another kind of intention figures—is instructive. Bowers, like Greg, points out that ideal copy may be "a purely hypothetical reconstruction" (p. 117).[16] One might therefore see a similarity between the descriptive bibliographer's usual concept of an ideal copy—a copy more nearly resembling the publisher's intention than any single extant copy may happen to do—and the scholarly editor's concept of a critical text—a text more accurately

reflecting the author's intentions than any single extant text may happen to do. There is an important difference, however. The critical editor's aim is to rectify errors in a text, whether or not they have ever been corrected in any previous edition; the editor's material is texts, which are arrangements of words and punctuation that happen to be given concrete embodiments in the form of printed editions. But the descriptive bibliographer's goal is to report the historical facts about an edition and not to perfect it or alter it beyond the latest point reached by those responsible for producing it; the bibliographer's material is the concrete embodiment of texts, not the texts themselves. Thus if all copies of an edition contain an obvious textual error, that error is a part of the ideal copy; but if some copies do not contain the error—either because a stop-press alteration rectified it during the course of printing or because the leaf or sheet containing the error was canceled and replaced with a corrected substitute after printing was completed—the description of the ideal copy would have to indicate the fact that the correction had taken place. The critical editor, on the other hand, is not limited by the fact that an error has not been corrected in documentary form, if its status as an error can be otherwise established. Both the bibliographer and the editor produce historical reconstructions; but the subject of the bibliographer's attention is a physical object, whereas the focus of the critical editor's attention is an abstraction. Critical editors must take the physical evidence into account, but it is of concern to them for what it reveals about the authority of particular readings, not for its own sake.[17]

One can see, therefore, that the introduction of the idea of intention into the definition of ideal copy is not without its difficulties. How it came about is easy enough to understand: when a leaf is missing, or a plate is inserted at the wrong point, or a canceled leaf remains alongside the cancel, the natural reaction is to say that these results were not intended—and that they are thus characteristics of aberrant or defective copies. But the recognition that something is unintended clearly provides no basis for saying that the intended form also exists (or once existed). If a given page in one copy of a book contains an obviously garbled text, one would not be inclined to jump to the conclusion that a corrected text exists in other copies, and one might even say that the matter is totally irrelevant to the determination of the ideal copy. But if another copy is located with the garbled text rectified by means of cancellation, ideal copy is certainly affected, because the description of the ideal copy would have to make reference to the cancel. Furthermore,

it is difficult to see how the situation would be any different if the correction had been made during the press run, though there would then be no cancel to draw one's attention to the textual alteration. In either case there is a physical difference among copies of a single issue: in one, the difference is of inked type-impressions only; in the other, it is that plus a change in the physical structure of a gathering. To make only one of these situations a factor in the determination of ideal copy, and the other one not, would be to insist on an artificial distinction and thus to trivialize the concept of ideal copy. As Bowers's discussion recognizes, stop-press alterations do affect ideal copy; defining "ideal" so as not to imply freedom from textual error does not mean that textual matters are to be ignored. If a printer corrects one error—either during the press run or by cancel—but in the process overlooks another on the same page, that remaining error is of course a part of the ideal copy; but where the alteration is in fact made, the altered form must be taken into account in thinking about the characteristics of the ideal copy. It is true that textual correctness is not relevant to ideal copy; but when textual differences actually exist between copies of an issue, determining which readings were the replacements (whether or not they were "corrections") is a necessary activity in describing the ideal copy for that issue. Authorial intention is beside the point; the question is not whether the printer's alteration brought the text into greater conformity with what the author intended but simply the determination of which reading the printer intended as the replacement. Intention is involved in ideal copy, then, only to the extent that alternatives to choose among exist in physical form in the issue as a whole. Which of two forms a printer intended as a replacement for the other is the bibliographer's business to ponder; but when only a single form was produced, the question of intention does not come up, even though some of the features of that form are not in fact what the printer intended.

If some of the differences among copies were created by the activity of individual owners or were caused in some other way after the time when the copies involved had left the hands of their producer, one could say that such differences were not intended by the producer. It could well happen, however, that some of these differences might actually be a fulfillment of the producer's intentions. Obviously if a leaf is lost from a book through rough treatment over the years, this loss is not a cancellation intended by the printer; but if some owner has a misplaced leaf moved to its correct location, this action carries out what the printer intended. Yet it is equally irrelevant to the bibliographer's concerns, for it is not a part of the production history of the book. Intention of the printer or publisher is therefore not really the most satisfactory

way of approaching the question of ideal copy. The crucial point—at least the first crucial point—is to distinguish those differences among copies attributable to the producer of the book from those not so attributable. The fundamental purpose of establishing such a concept as ideal copy, after all, is to segregate characteristics of a book that occurred during its production from those that occurred at some later time. There are further considerations that the concept leads to, but its basic function is to enable one to identify which features of a given copy of a book were present when the book was released or put on sale and which ones originated after that point. The intention of the printer or publisher of a book has nothing to do with the matter: unintended features may be present in copies as they are released, and intended features may be brought about later.[18] But the bibliographer, as a historian of the production of the book, must be interested in the former and is not concerned with the latter.[19]

What the bibliographer must decide, to put it another way, is which differences among copies constitute states and which result from the history of individual copies following their release from the publisher or printer.[20] In bibliographical terminology, a state refers to any part of a copy of a book that differs, at the time of such release, from the corresponding part of other copies, within the same issue or impression, as a result of an error or the correction of an error, or for any other reason

that causes the publisher not to call attention to the difference as a discrete publishing effort.[21] A state is one of the forms of a sheet or a casing available to a purchaser at the time of the release of individual copies; but it is by no means necessarily an intended form. States do often, it is true, originate in a printer's or a publisher's effort to correct an error discovered in previously printed, but not yet released, copies, and such corrected states are indeed intentional. But states can equally well derive from errors that occur during the process of production, making the later states less correct than the earlier. An omitted or a duplicated gathering, or a page with severe type damage along one margin, is clearly not an intended feature of a book; but a copy that reached the market in this condition could not be dismissed by the bibliographer, because this feature is a part of the production history of the copy. The number of copies involved cannot be the test, any more than intention can. Even if only one copy lacks a given gathering when put on sale, or shows type damage, that is still one of the states in which the book emerged from production; and if twenty copies of a book have by coincidence had identical leaves removed from them after their release, that curious situation has nothing to do with the production of the book, and those copies would not represent a state. Of course, the evidence of a single copy (or a very small number) is often difficult to interpret, and one might conclude on such evidence that a particular feature was not present when the copy was released, and yet might conclude differently if more copies exhibiting this feature were located; if, however, the nature of the evidence allows one to know, even from a single copy, that a feature was indeed present at the time the copy was released, one must take it into account somehow, even if one wishes to stress its insignificance.[22] The point is that the number of copies exhibiting a particular feature cannot by itself determine whether that feature qualifies for inclusion in a standard description. The primary decision for the bibliographer, therefore, is whether or not a given feature of a book is a product of its printing and publishing history. The decision is not a simple one: because intention and statistics are not determining factors, an obviously unintended feature observed in a single copy cannot automatically be ruled out as not constituting a state.

The concept of the ideal copy of an issue—though the term is usually expressed in the singular—must accommodate all the states of the issue. A single copy cannot be expected to contain all states, of course, but all the states must be a part of the description of an issue. The notion that a description of an ideal copy must be set up as a description of a single copy no doubt results from what is perceived as the practical necessity of choosing certain features—a particular title page, collation formula, list of contents, and so forth—for primary representation in the description. The only real necessity, however, when two versions of the same thing exist, is to describe one first, and there is no reason why the one presented first need be thought of as "primary" in any qualitative sense. In fact, the historical purpose of the description is best served by reporting—for any given feature that exists in two states—the first state first, followed by the second.[23] It is true that in some cases this

arrangement would mean that the less correct state would be reported first—but not always, for a first state is not always less "correct" than a second, since a state can result from accident as well as intention. And in any event, the presence of a less correct state of a given sheet in a copy of a book does not make that copy any less "ideal," because the copy still represents one of the forms in which the book was offered to the public. A state by definition relates only to one part of a book, such as a sheet, and not to a book as a whole; when several sheets are produced in variant states, it is possible for a particular copy of a book to have earlier states of some sheets and later states of others. Thus the description of variant states must be a part of a description of an issue or impression, and the "ideal" being described encompasses certain sets of alternatives. Although alterations to title pages often produce new issues because they identify a discrete publishing effort, some variations in title pages may produce only states; in such cases a description would have to provide two title-page descriptions, one labeled "first state" and one "second state." When variant states affect the collation formula or the contents paragraph, alternative accounts of the affected element can be reported; or when stop-press alterations result in variant formes, the various possibilities can be outlined. The important point to recognize is that the standard description, or ideal copy, of an issue must allow for the variations identified as states.[24] A different issue will be represented by a different ideal copy; but for any given issue, ideal copy is likely to embrace a set of alternatives.[25]

The most difficult problem, therefore, is not the mechanical one of setting up a description so as to accommodate different states but the prior question of how to decide which variations are in fact states and which are alterations resulting from the vicissitudes that a copy has endured after the time of its release. The latter must of course be recorded in the section of the description identifying the individual copies examined; that section constitutes the primary documentation on which the description is based, and peculiarities of particular copies must be mentioned there, so that readers can know what evidence the bibliographer surveyed. But the variations that can be classified as states—those occurring during the process of production—become part of the body of the description itself. Indeed, the purpose of the concept of ideal copy is to cause the bibliographer to think about this distinction; to put the matter concisely, ideal copy is concerned with those differences within an issue that can be regarded as states. It follows, to repeat two points not widely understood, that textual differences are to be included among the physical variations with which ideal copy is concerned; and that the producer's intention is relevant only when variations actually existed among the copies as they were issued, and then only to the extent that it may be of use in thinking about the order of those variations.

All these points inevitably lead to the central question: how one can possess the requisite physical evidence for knowing whether differences exist among copies of an issue, and then for deciding which of them occurred during the production of those copies and which occurred at later times in their histories. Naturally, one begins by looking at numerous copies, when they exist. Even though the number of copies exhibiting a given characteristic is not necessarily a criterion for classifying it as a state, one can in fact frequently decide that a certain characteristic appears too many times to be attributed to the independent action of individual owners. And of course, the inspection of multiple copies is necessary simply to locate what the points of variation are. But the process raises difficult questions. How can one know when all variations have been discovered? How many copies constitute a sufficient sample? May it not happen that all surviving copies, even when there are many of them, do not form a large enough proportion of the total

edition to serve as a reliable sample? If it is not feasible to see every copy of an issue (and generally it is not, either because they do not all survive or because the issue was too large to make tracking them all down a practicable undertaking), can the attempt to generalize about the issue as a whole be justified?[26] The first answer is to say that these questions can be asked about any inductive investigation; that, as one becomes expert through the process of investigation, one is in a position to make an informed judgment about the extent of the evidence required or the sufficiency of the obtainable evidence; and that careful generalization, based on cautious and responsible use of the located evidence, is valuable, even if it is later shown to be erroneous in certain respects. These points do of course apply to bibliographical research; but because descriptive bibliography is not always recognized as the branch of historical scholarship that it is, some further comment on these questions may help to clarify the nature of bibliographical descriptions.

Suppose, for instance, that one has examined four copies of a book, all of which have a dedication leaf at the end of the last gathering, presumably intended to be cut out and inserted among the other preliminaries in the first gathering. Does the seeming obviousness of the intention justify constructing a collation formula for the ideal copy indicating the insertion of this leaf into the first gathering? How could one be certain where in the gathering to indicate its insertion? If the leaf bore a special signature or a page number or some other guide to its intended location, would the case for placing it in the preliminaries be strengthened? If one looked at twenty-five, or fifty, copies of the book and still found all to have the dedication leaf at the end, would the question be affected? Would one's thinking about the matter be different depending on whether the book was issued in a publisher's casing or whether it was from an earlier period and was bound according to the wishes of individual owners? These questions illustrate why one has to consider whether the purposes of bibliographical description ever require that a printer's or publisher's intention take precedence over the historical fact of what actually happened. If (for the sake of argument) this hypothetical book was from the period of publisher's binding and if the entire issue was bound with the dedication leaf at the end, the bibliographer would have no justification for altering its position in describing the ideal copy, even if in fact this position was not the intended one. If the

book was from the period before publisher's binding (or from a later period but issued in sheets), one might more relevantly raise the question of intention, because the producer of the sheets would not necessarily have had control over the binding of any copies. Even so, if all copies were bound with the dedication leaf at the end, there would be little reason for presenting a standard description that differed from them—unless clear evidence existed to show that all the binders had disregarded the printer's binding specifications. In practice, of course, one can scarcely ever be in this omniscient position with regard to the issue as a whole. But the value of thinking about an extreme situation is that it may make clearer the framework within which more realistic situations need to be viewed. A standard description of a book is a historical generalization, rising above the idiosyncrasies of individual copies but resisting the temptation to substitute what should have been for what was. Perhaps there would be less need to make this point if "ideal" had not been the adjective in the generally accepted term.

In actual situations, much of the difficulty in distinguishing what the producer was responsible for from what later owners were responsible for arises from the fact that one does not have access to all the evidence. Even if one examines all surviving copies of a book, the portion of the original issue that one has not seen may still be sizable; making generalizations about the issue as a whole can therefore be an uncertain business. Acts of historical reconstruction must very often be based on something less than the full evidence and are thus dependent on the judgment of the historian in weighing such evidence as there is. Different historians may of course come to different conclusions, and more than one account may legitimately exist at the same time. Descriptive bibliographers must make numerous decisions in the process of interpreting the evidence of individual copies and constructing from it a generalized account. There are a great many variables that they must consider, such as the printing, binding, and publishing customs of the period, any external evidence bearing on the situation, and the number and nature of the surviving copies. So much inevitably depends on the learning and judgment of the bibliographer that it would be futile to attempt to set up detailed rules regarding ideal copy.

What is important, instead, is to have a clear idea of the rationale that should underlie specific decisions. Surely such a rationale must start from the premise that, while the production of a standard description is a creative activity, none of the points that go into it can be invented. Everything else follows from that. If a leaf is obviously missing, or an obvious typographical error is present, in all examined (or known) copies, one cannot invent a corrected state. But if tangible evidence

exists, one may then be legitimately faced with alternatives. It is then that one begins to weigh the various characteristics of the specific instance—the printing practice of the time, the number of copies seen, and so on. If a dedication leaf consistently appears at the end of the known copies of a book, one may still decide to note it in the description at the conventional location, if one has reason to believe that all the known copies are idiosyncratic—if few exist, for example, and if they were bound by individual owners. Under different circumstances, one might decide that the standard description should show the leaf in its position at the end—if, for example, a great many copies, all in publisher's binding, have been seen. In neither case is one inventing evidence: the dedication leaf exists. And in the first case one is not falsifying the evidence, for a standard description does not purport to be identical with any given copy; by noting the dedication leaf at its conventional place, one is expressing an informed and reasoned opinion that the evidence supports the choice of this position as the place where the dedication leaf would have appeared if the printer had had control over the binding. Printer's intention thus can play a role for books not issued in a binding or casing, but only when compelling evidence exists that the order of the leaves as bound is not the printer's intended order. Even so, it is important to have the description show the order as printed as well as the intended order of issue. And in other respects the printer's intention is irrelevant, for intention does not necessarily coincide with performance. Projecting what was true for the issue as a whole is not the same as projecting what was intended; both activities are of interest, but the bibliographer must keep the distinction clear and understand which governs the description of a so-called "ideal" copy. (In making this distinction one must remember, in other words, that leaves intentionally printed at one point for insertion at a different point constitute a special case: in books issued unbound their presence at the original location at the time of issue is not an error, as it would be in books issued in publisher's casing, and thus a bibliographer's decision to record both the proper position as well as the original position of such leaves is not an instance of correcting an error but simply a way of clarifying the nature of the issue.)

There are times, of course, when intention and actuality coincide; but if bibliographers are to have a firm understanding of what they are about, they must be prepared to deal with situations where the two do not coincide. All decisions regarding ideal copy must be reasoned ones; none can be automatic or mechanical. Less misunderstanding about ideal copy would probably have arisen if the status of descriptive bibliography as history had been better understood. Once one leaves the

security of describing individual copies and attempts to produce a standard description that will accommodate not only all examined copies but those not examined as well, one is engaging in the creative scholarship of historical interpretation. Responsible historical accounts do not depart from the facts, but they are more than assemblages of discrete facts; they bring the facts together in such a way as to reveal a meaning or order in them—not the meaning and order, because the facts may well, and probably will, support other interpretations. Such an account is thus necessarily hypothetical, but for all its lack of certainty it marks an important advance in understanding; even several conflicting (but responsible) accounts are more informative than the undigested facts. A description of a standard or ideal copy of a book, in other words, is "truer" than the description of any one copy, even though it rests to a greater degree on subjective judgment. Good descriptive bibliographies are normally more useful than catalogues because they are based on a qualified individual's reasoned assessment of the evidence relating to the edition or issue as an entity. When one focuses on the fact that a bibliographical description is a piece of historical scholarship, one can readily see that the standard copy or copies it describes cannot necessarily be equated with a copy that fulfills its producer's—much less its author's—intentions. A standard description reports the "ideal" only in the sense that it records the features (or variations in them) thought to characterize individual copies as they emerged from the printer or publisher. Any other construction of the word, or of the concept of ideal copy, would conflict with the historical aims that "description" suggests and would confuse rather than clarify the orderly delineation of past events.

II

Ideal copy, as this account of it is intended to make clear, cannot be divorced from textual considerations. The fact that descriptive bibliography is concerned with physical details has caused some misunderstanding on this point and has led some people to think that ideal copy is unrelated to textual matters and therefore to editing. Ideal copy does, of course, refer to physical features of a book, but the inked type-impressions that transmit the text are physical features, and textual differences among copies are physical differences. Textual correctness is obviously something else, and descriptive bibliography is not concerned with it; but textual differences among copies cannot be ignored in descriptive bibliography, or in the concept of ideal copy. The idea that analytical bibliography—which underlies both descriptive bibliography

and editing—does not require a knowledge of the meaning of the texts printed in the books under analysis has frequently been held in exaggerated form, largely as the result of some overstatements on Greg's part (made in the course of trying to explain the aims of what was then a relatively new field).[27] It is true, of course, that the analytical bibliographer is not concerned with the meaning of texts in the way that an editor is; but a knowledge of the language of the texts is always helpful to an analytical bibliographer, and sometimes essential (as in recognizing variant spellings in compositor analysis). Analytical and descriptive bibliographers would often leave their work undone if they did not take note of textual differences in their accounts of physical differences among copies.

The relation of the textual interests of descriptive bibliographers and those of editors can be somewhat clarified by considering the use that editors of both critical and noncritical editions might make of bibliographical descriptions. Noncritical editions provide the more obvious starting point, because a noncritical edition is one that aims to reproduce exactly a particular manuscript or printed text—without, that is, calling upon the editor's critical judgment to provide emendations. For a noncritical edition of a manuscript, there is no question about what copy to follow, since a manuscript is unique; there may of course be a question about which of the various manuscripts of a given work is to be reproduced, but once a particular manuscript is settled upon as worthy of reproduction, there is no further choice to be made among copies. But with printed material the situation is different. Even after one has decided which edition—or impression or issue—of a work is the one to be reproduced, one still must face the problem of which copy should be used, since one obviously cannot assume—for books of any period—that all copies are the same. Whether the noncritical edition is to be a photographic reproduction or a new setting of type, it will contain that form of the text present in the copy selected, since no emendations are to be made, and the copy should therefore be chosen carefully. Of course, the copy used must be specified in the noncritical edition, and some editors may argue that the choice of copy is not crucial, so long as readers know exactly where they stand—so long as they are told that the noncritical edition follows the text of a particular copy and that it is thus limited to the documentary evidence present in that

copy. But surely in most cases it would be foolish to reproduce a copy that had been altered over the years if another copy existed that was closer to the form (or one of the forms) in which the book originally appeared. Comparison of extant copies, then, is a necessity even for editors of noncritical editions, so that they will be in a position to determine what features of particular copies have resulted from alteration occurring since the time of the publisher's release of those copies. In addition, they may find that copies of a given issue differ among themselves in containing variant states of the text, and they have the problem of choosing—even among copies that have not been altered by individual owners—the copy that is most desirable from a textual point of view. Some editors of noncritical editions have long recognized the usefulness of pointing out textual differences between the copy reproduced and other copies. Even when an editor feels that it is not feasible to provide a full list of variants, some statement is called for—beyond the identification of the particular copy—generalizing about the issue as a whole and what relation the reproduced copy bears to it. Just because editors of noncritical editions do not have to make critical choices among individual variant readings does not absolve them of the responsibility for knowing what variants exist within the edition they are concerned with; choosing a copy for reproduction is itself a critical choice, and it should be as informed a choice as possible. When a descriptive bibliography has sorted out the various states comprising the ideal copy (or, it might be better to say, the ideal copies) of an issue, the editor of a noncritical edition can use this information with great profit. But when that bibliographical work has not been performed, the responsible editor of a noncritical edition has no alternative but to undertake the task. Editing, even of noncritical editions, cannot be divorced from descriptive bibliography and from the concept of ideal copy.

Charlton Hinman's Norton Facsimile (1968) of the Shakespeare First Folio provides the natural focus for a discussion of this point. The Folger collection of eighty copies of the Folio made feasible Hinman's detailed study of the printing of the volume[28] and put him in possession of a thorough knowledge of the various mixtures of corrected and uncorrected formes present in different copies. Because each of the copies contains a mixture of earlier and later states of the text of those pages known to exist in variant states, a photographic reproduction of any one copy—though it would be useful in showing one of the forms in which the book appeared—would be limited to the chance combination of sheets present in that copy. What Hinman conceived was the idea of

photographing the latest state of the text of each page, from whatever copies these states were found in, and bringing these photographs together; the resulting facsimile is not a facsimile of any extant copy but a collection of facsimiles of individual pages. In the end Hinman found it necessary to utilize thirty copies as the sources for photographs, and he carefully records the copy used for each page. The Norton Facsimile is a noncritical edition because Hinman does not introduce editorial emendations into the text as it appears in the Folio; but because there is no such thing as the text of the Folio, he has clearly exercised his informed editorial judgment in deciding which of the Folio texts of each page is to appear in his facsimile. This approach is the logical result of a search for the most appropriate copy to reproduce.

Whether or not the result should be thought of as representing the ideal copy is a question that has engendered some disagreement. Hinman himself, after describing the basis for his choice of pages, calls it the first facsimile that undertakes to reproduce "such an 'ideal' copy of the First Folio" (p. xxiv). Those who object to this application of the term are apparently under the misapprehension that the concept of ideal copy has nothing to do with textual variants. Actually, the basis for possible criticism of the statement lies elsewhere. In the first place, if this collection of pages were claimed to be the ideal copy, one could object that the book appeared with various combinations of sheets and that the bibliographical concept of ideal copy must accommodate the divergent forms that in fact appeared; this combination of pages might indeed be an ideal copy, or one of the forms of ideal copy, but cannot be the ideal copy. Second, whether it is an ideal copy depends on whether it would have been a physically possible form for the printer to have issued. Obviously in preparing a facsimile text one cannot pick and choose among readings on the same page, combining a reading from one copy of the page with a reading found only in a different copy at another point on the page, if there is no evidence that these two readings ever existed simultaneously on the same page. But as long as it was physically possible for sheets printed with first formes in the latest state of alteration to be perfected with second formes in the latest state and for successive sheets in such a state to be assembled into individual copies of the completed book, one can legitimately posit that combination of "corrected" formes as falling within the range of copies encompassed by ideal copy. The fact that such a copy does not actually exist is not in itself a deterrent to the classification, since the extant copies are a random assemblage, and the combinations of sheets in them are also produced by chance. But the degree of likelihood that such a copy ever existed, if one is in a position to weigh that likelihood, may indeed be a

deterrent. If one had been able to see every copy as issued, and none contained all the corrected formes, one could not think of such a copy as an ideal copy, even though it was technically possible, for ideal copy is a historical concept applicable to the copies of a book as issued. But when one has access to less than the full evidence, what is possible becomes less easy to rule out.

In any case, the question whether the Norton Facsimile can be considered to represent an ideal copy is of more use as an exercise in thinking about the implications of the concept than for any practical purpose it might serve. A thorough bibliographical description, after all, would record the various states of each forme (and, in the documentation, the combinations present in examined copies), but it would not endorse particular combinations of those states, except where there is evidence that certain states are linked together. And Hinman's facsimile—though it may indeed, as he claims, be a "reproduction of what the printers of the original edition would themselves have considered an ideal copy" (p. xxii)—need not be labeled an "ideal copy" except to the extent that it presents one of the possibilities that ideal copy for this book apparently embraces. What it does, to be more precise, is to single out from the various forms in which the book appeared or conceivably appeared the one that makes most sense, in Hinman's view, to reproduce for purposes of textual study. The historical question of whether a copy could have appeared, or did appear, in this form is of less moment for such purposes than the value for reference of having this particular assemblage of material. One should not forget, however, that editors and other students of the text cannot limit themselves to "corrected" formes, which are often less correct—at least as far as accidentals are concerned—than "uncorrected" formes.[29] They must be fully apprised of all variant states of the text and must therefore consult, or prepare for themselves, the information that a bibliographical description provides.

What Hinman did with photographs, some collectors have done with originals: to insert material from one copy into another copy. Of course, Hinman could move single pages, whereas those who are recombining originals cannot move less than a single leaf (containing a type page from an inner and from an outer forme). The practice of

"making up" copies by inserting leaves to replace those that are missing has been variously discussed, and with considerable feeling. One prominent instance concerns the Scheide copy of the Gutenberg Bible, in which several missing leaves were replaced after it came into the Scheide collection. Writing about this copy in The Scheide Library (1947), Julian Boyd says that there are two schools of thought on the matter. Some would hold that the copy should have remained in exactly the form in which it was received; others, he believes, would take the point of view This passage, apparently meant to defend the practice of making-up, does not give it as good a defense as might be constructed and in the process exhibits a number of confusions—confusions that a careful consideration of the concept of ideal copy would have set straight. It is true of course that a great many variables determine the particular combination of sheets brought together for each copy of a book and that individual copies cannot be expected to consist throughout of sheets that were first off the press, or second off the press, or third off the press, and so on. (In any case, the order "off the press" can refer only to the printing of second formes; the order in which the white paper was printed is likely to be very different from the order of perfecting.) But this fact cannot be taken to suggest that all mixing of sheets has the same status regardless of the point in the history of a copy when it takes place. The point where the "trusteeship" of a copy begins, to put the question in those terms, is the moment when the copy is released from the control of its producer. Even though the sheets of that copy may be mixed (or even deficient), the combination and arrangement of sheets present in it are evidence of one form in which the book emerged from the process of its production; any physical alteration made in the copy after that point, even if it brings the copy nearer to the form intended by the printer, violates the integrity of the historical evidence.

To say that the "alteration of the condition" of a book by binders is accepted "unquestioningly" is to confuse the issue in several ways. First of all, it is one thing for a binder, in the process of binding a book that was not published in an edition casing, to rearrange certain leaves within a copy, as specifically called for by the printer (e.g., moving a half-title from the end to the beginning). It is quite another for that binder to undertake to remedy what are presumed to be defects in the copy by inserting leaves that were not originally part of the copy. If binders have often thought of this latter procedure as a routine part of their business, and if their clients have sometimes accepted the practice "unquestioningly," these attitudes do not justify the practice or provide grounds for later owners to continue it. Making up copies, whether by early binders or by later collectors, remains a tampering with the evidence. Furthermore, the situation is not simply one of continuing to mix sheets in the way that they were originally mixed; what is inserted later is hardly ever a sheet (a whole piece of paper as folded according to the format) but a leaf, thus mixing together a leaf from one sheet with the remaining leaf or leaves from another sheet. Students of the text naturally need to ascertain the order of any stop-press alterations and must therefore work out just what was on the press at any one time; but one can never know whether the text of inserted leaves is identical to that of the missing leaves, and inserted leaves obviously cannot stand as satisfactory substitutes for the originals. Of course, one should normally be able to tell that such leaves are disjunct, and one would then look at them with particular skepticism; but still one needs to know whether they are true cancels or later insertions, both for textual purposes and as a general matter of understanding the nature of the preserved evidence. Finally, there is the point about the "interchangeability" of leaves printed from movable type. The invention of movable type clearly has nothing to do with the matter, since other ways of preparing printing surfaces (such as that used for block books) were also intended to result in identical copies. The important point, however, is that the intention to produce identical copies cannot be accepted by the historian as equivalent to the deed; one cannot assume that all copies of a leaf were in fact identical and that missing evidence can simply be replaced by its supposed duplicate.

For all these reasons bibliographers and editors always try to be alert to the possibility of alterations made during the process of binding or other alterations made by individual owners. Books not issued in publishers' bindings or casings, as well as rebound or recased copies of

those so issued, must always be examined with unusual severity; and in some cases one remains so unsatisfied as to rule such copies out completely as sources of evidence. Certainly one should have no faith in rebound copies of nineteenth- and twentieth-century books if copies in publishers' casings are plentiful (though one should still examine them on the chance that they might contain otherwise unknown textual variants). There is no way of rationalizing the making up of copies to give them the same status as books that have not been tampered with. A much better, and far simpler, defense of the Scheide Bible would have been to say that, since certain leaves were missing from it, no further damage was being done to the copy by replacing those leaves, so long as the known facts about each inserted leaf were clearly indicated in order to prevent any misinterpretation of the evidence. This may not be a very strong defense, but at least it says that the insertion of leaves, if done properly, causes no serious harm, even if it provides no great benefits, either. In making such a point, one of course assumes that the leaves used for replacement are already disjunct and that another copy of the book, however imperfect, is not being destroyed to provide a supply of leaves. The loss of evidence that may result from dismembering even an imperfect copy is not offset by any important gain when leaves from it are made to serve a cosmetic function in another copy. John Carter, discussing made-up copies in his ABC for Book Collectors (5th ed., corr., 1974), warns his readers against copies containing leaves from wholly different editions (resulting, he says, from "faking-up"). But whereas such "bibliographical felony," as he describes it, is naïve and easily detectable, the subtler activity of inserting leaves from the same edition is no less a felony and is equally productive of fakes. The argument made for the Scheide Bible is by no means unique, but it provides a convenient occasion for showing why editors of noncritical editions must be concerned with the concept of ideal copy and why their work and that of descriptive bibliographers are inseparable.

What is true of editors of noncritical editions is also true of editors of critical editions; and it is worth considering briefly why this is so, since one might at first think that establishing the forms of ideal copy would be less important to the critical editor, whose aim is not to produce a facsimile or an exact rendering of a single documentary text. The goal of the critical editor is to construct a text that conforms as closely as possible with some particular standard—normally, in scholarly editing, the author's intention. It may be that none of the documentary texts meets this standard, and the critical text constructed by the editor may be different from any that has gone before. Critical editors are not limited, in other words, to readings that have appeared in print in a

given edition;[30] they may combine readings from different editions and make corrections on their own authority (corrections that had not previously been made in any edition), so long as in their informed judgment the available evidence justifies such emendations as bringing a text more into line with its author's intention. In order to be in a position to use this editorial freedom responsibly, they must have a thorough knowledge of all relevant evidence, the primary part of which is the physical evidence.[31] Even when they are certain that a particular reading requires to be corrected in a given way, their thinking must still be grounded on the firm base of knowing all the readings that have appeared at that point in authorized editions. If they do not make a search among copies of the same impression or issue—with or without the benefit of a descriptive bibliography—they will not know whether the correction they are proposing was in fact already present in one state of one of the sheets of the book. The difference between the descriptive bibliographer and the critical editor can be illustrated by a situation in which a printer makes a stop-press correction but through a mischance simultaneously creates an error in the following line. The editor will choose the two correct readings, one from each state; the bibliographer will record the two states as alternative forms for ideal copy but cannot contemplate the two correct readings together, since there is no physical evidence that the two ever existed together on the same page. Despite their differences in the use of the information, both editor and bibliographer must make the same search for evidence: both must in effect define ideal copy, because both must be aware of all states of an issue.

That states result from inadvertent as well as intended alterations and that later states are not always more correct than earlier ones is demonstrated over and over in editorial work. For example, progressive

deterioration of type or variations in inking during the course of an impression may create what appear to be textual errors in books of all periods. Many editors of The Tempest have printed one line as "So rare a wondred Father, and a wise," believing—even after checking a number of copies of the Folio—that the last word is indeed "wise" (printed of course with a long s). Jeanne Addison Roberts, however, after examining many more copies of the Folio, has discovered that the word originally read "wife" and that the crossbar of the f shows various stages of progressive damage in most copies, making it look like a long s.[32] The lesson for editors is clear. Similarly, there are many instances in the original printings of Melville's work where letters or marks of punctuation failed to print; the spaces for them are present, but nothing appears in the spaces in any copies thus far examined. In "The Piazza," no punctuation follows the phrase "this weariness and wakefulness together," and the following word, "Brother," appears to start a new sentence; presumably a period is missing, though conceivably the mark was an exclamation point (the passage is in direct discourse). In Moby-Dick there is an unusually large space following the word "almanack" at the end of the first clause of a sentence in Chapter 99; an editor will probably insert a semicolon there because the English edition (set from proofs of the American) has a semicolon. But a comma would also be appropriate, and there is no way of knowing what mark the compositor set unless a copy turns up with the mark printed. Even when there is no question what is missing—as when, in Mardi, the phrase "crossing his wooden eg" appears—the editor who has located no copy containing the missing letter will have to list the correct form as an emendation. In the English edition of White-Jacket the word "very" in the phrase "very industriously" (in Chapter 54) is not present, though there is a space for it. Since the American edition is the copy-text, an editor does not of course need to report "very" as an emendation, but its absence in the English edition must be recorded in the list of variants between the two editions. If, however, a copy could be found in which "very" prints—and that is certainly possible, since the book was printed from type, not plates—the situation would be reported differently.[33] These few examples are enough to suggest the way in which the editor of a critical edition is dependent on the facts uncovered in a search for the range of states that define ideal copy for a given issue. Critical editors can in any case make the emendations they consider necessary, but their range of options may be altered by the states located in an extensive examination of multiple copies—and what is regarded as an emendation or a variant may also be affected. There is no escaping the conclusion that the concept of ideal copy is as vital to the editor as to the bibliographer. Even though the concerns of editors are with texts as intellectual constructions, they must recognize that as full a knowledge as possible of the vehicles that physically convey those texts is essential to their work.

Whether differences among copies of an issue are the result of accident or design, whether they improve or worsen the product, they qualify—so long as they were present in copies at the time of issue—for recording as states by the bibliographer and for examination as relevant textual evidence by the editor. Bibliographers, and theorists of descriptive bibliography as well, have been reluctant in the past to insist on the presentation of textual evidence in descriptive bibliographies;[34] it is tempting to say that an edition is the place for such information and to argue that further deterrents should not be brought forward to interfere with the production of descriptive bibliographies. But it is equally difficult not to agree with the view that textual differences are physical differences and that states of an issue or impression cannot be satisfactorily sorted out without taking textual variants into account. The notion that descriptive bibliographies should limit themselves to details necessary for identification, instead of providing fuller historical records, has now been discredited; but in any case, it is impossible to know what is necessary for identification—indeed, how many states there may be for identification—without being as fully acquainted as possible with all the differences among copies, including textual differences. A number of bibliographies during the last generation have incorporated the results of textual collation.[35] But where the interdependence of description and textual study has been particularly felt is in the large series of editions published in connection with, or following the guidelines of, the Center for Editions of American Authors and its successor, the Center

for Scholarly Editions.[36] Bibliographies of varying quality existed for the authors that have been edited, but none was really satisfactory in providing the kind of history of publication that editors need. These editors therefore undertook much of the work that goes into producing a descriptive bibliography as part of their editorial research, and as a result the editorial matter in most of these editions now amounts to the most thorough bibliographical treatment we have of the works in question. Formal bibliographies have been planned in some cases to accompany these editions in due course, and such an arrangement—a bibliography and an edition emerging from the same research—is the best one imaginable. An edition must, in any case, establish the detailed printing and publishing history of a work and must in the process distinguish among impressions, issues, and states. And it should be likewise recognized that a descriptive bibliography cannot adequately record the printing and publishing history of a work without utilizing information secured through textual collation.

III

These considerations can perhaps best be summarized in the form of a new definition:

A further statement can make more explicit the relation of binding to ideal copy, though what is said here is implied in the foregoing definition: This approach to ideal copy is somewhat different from the views of it that have been expressed previously, though I think it is in line with the spirit of the best of those earlier discussions.

In particular, this approach departs from the usual concept of an ideal copy (in the singular) by suggesting that the standard to be described is not a single preferred form but a range of alternatives that encompasses all the states of each sheet as they were issued. It does not seem logical to distinguish those differences that affect the physical structure of the folded sheets (and thus the collation formula) from those differences that affect only the text printed on the sheets, and then to limit the concept of ideal copy to one of those categories. Both kinds of differences are physical, and both may result either from conscious intention or from accident—or from both at once, as when new errors are created in the process of making corrections. And it does not therefore seem reasonable to limit ideal copy to a single state or single collection of states: there is no reason to elevate either a first state or a second state to a position as a basic standard, since neither is necessarily closer to the "intended" standard of the producer. What a bibliography can more valuably concern itself with is constructing an accurate historical account of the various forms of an issue that were put on the market; the fact that one form of an issue may happen to be more "correct," either in the arrangement of its leaves or in the quality of its text, does not make it the standard for the issue, since copies exhibiting other states may also be authentic exemplars of the issue as a whole. The editor of a facsimile edition must select a single state of any given page to appear in the body of the facsimile (and may well choose the "uncorrected" state in many cases as being closer to the author's wishes); but the bibliographer has no such necessity for choice and indeed may offer a misleading historical picture if certain states are subordinated to a single "ideal copy." The bibliographer's task requires careful judgment and historical scholarship, just as the editor's does. But that judgment and scholarship must first of all be directed toward distinguishing between what has happened to copies of books since their release and what happened to them before that time. This essentially simple view of ideal copy—or whatever we choose to call the concept—emphasizes the distinction that must be basic to the historical study of books. And while the concept may be simple, the decisions to be made in determining the status of certain features of individual copies of books are frequently far from simple.

Thinking about ideal copy as a concept inevitably leads one to consider the nature of library cataloguing and its relation to descriptive bibliography. The fact that there is sometimes a confusion in cataloguing

as to the aim of the process—whether works (intellectual constructions) or books (physical objects) are being recorded—has previously been commented upon.[37] Insofar as catalogue entries are taken to concern physical books, a related confusion concerns whether each entry applies to a particular copy of a book or more generally to the issue or impression of which that copy is a part. Ostensibly, of course, a catalogue records the copies in a given collection and does not pretend to be a bibliography (which draws on evidence from many copies to present a generalized account). But matters are often not so clear-cut. There are instances in which a cataloguer consults some sources external to the book being catalogued (other copies of the book in the same collection, perhaps, or catalogues of other collections, or a standard reference work) but does not undertake the extensive research that underlies a bibliographical description. This situation causes no problem so long as the specific copy is accurately recorded, in whatever level of detail has been established, and any observations on the relation of that copy to others are labeled for what they are and documented. Where the trouble occurs is when a cataloguer blurs the distinction between the individual copy and the so-called ideal copy, blending details from both and not keeping either approach clearly in focus.

It would seem that the Library of Congress has unwittingly been the culprit in fostering this confused view. The public distribution of LC printed catalogue cards, which began in 1901, has had its effect both on cataloguers at LC and on some cataloguers elsewhere: it has caused the former to subordinate the presumed defects of LC copies and the latter to believe, following the example set by the widely circulated LC cards, that the function of catalogue cards is to describe ideal copies. LC cards, it is true, mention what are taken to be imperfections in LC copies, but frequently in the form of notes to the basic entries, thus reversing the proper emphasis and placing in primary position entries that do not in fact correspond exactly to the copies they represent. Some people have argued that this approach does little harm in connection with twentieth-century books—and it is largely for twentieth-century books that LC cards have served other libraries' cataloguing needs. But it is a mistake to assume that differences are uncommon among copies of twentieth-century books: they are in fact quite common. Contrary to what many people still seem to believe, machine-printed books demand just as close scrutiny as hand-printed books and often present more perplexing problems. In any case, the distinction between a description based on a single copy and one that draws conclusions based on the evidence

from many copies is essential and must apply to books of all periods.

The emphasis in discussions of library cataloguing has of course now moved from card catalogues to computers, but this blurred distinction between individual and ideal copy, inherited from card-catalogue days, is still with us, and in an insidious form. Those who talk about computer-assisted national bibliographic networks often refer to, or at least assume the future existence of, a data base containing entries that can be regarded somehow as standard, without considering how entries supplied by individual libraries become converted into a more authoritative statement utilizing the evidence from the various copies reported. The networks that presently exist, like OCLC and RLIN,[38] are used to a large extent by their subscribing libraries as a means for producing catalogue cards: the libraries check to see whether a given item has already been entered into the system; if it has, they make any alterations they wish in the entry and then indicate that they are ordering cards, and if it has not been previously entered, they enter it and order the cards. The real importance of the system has nothing to do with cards, and the central bibliographical problem will remain even when cards are no longer a part of the thinking of those who consult the record. Even though individual libraries can change the recorded entries for their own purposes—or to fit their own copies—the question of what description is to appear on the screen of the terminal is still a matter of considerable importance. Is it simply the first copy to be entered in the system? Or is the network to offer in some way what it regards as the "best" entry for a given item? Can any library or individual that calls up an entry infer that what appears on the screen represents a standard against which any given copy is to be measured? At present OCLC generally provides only a single entry for an item, whereas RLIN allows users to see multiple entries when they exist. The latter approach is the only one that is defensible from a scholarly point of view, and in the long run it is the approach that will have to be made available in a consolidated international network. When entries are based on individual copies—if, that is, they really are, as they purport to be—there is no mechanical or easy way of selecting from them a single "basic" or "standard" entry. Nor is there any reason why we should expect, or desire, a single entry to be chosen for us. It would be unfortunate if, in the early stages of a new medium, we were to allow the confusions

previously nourished by LC printed cards to dominate our thinking. Entries, regardless of their level of detail, that refer to an issue or impression as a whole can only be the result of an act of historical scholarship—which involves judgment and interpretation—and cannot be based on the examination of a single copy or on a superficial choice among the reports of several copies.

Individuals or institutions (i.e., individuals representing institutions) may wish to construct such historical accounts, and there is nothing wrong with their entering the results into a data base, so long as what they enter is clearly labeled. Indeed, such scholarship should be encouraged and welcomed, just as printed bibliographies are welcomed when they mark an advance in knowledge through the play of an informed intelligence over the assembled pieces of evidence. This kind of description should have the same status whether it appears on a cathode ray tube or on a printed page; it is an act of integration, and it can be well done or badly done, and it may stand as the authority for a considerable time or be superseded in short order by superior work. The quality of particular pieces of research cannot be legislated, nor can scholarly authority be upheld by decree or administrative fiat. Thus any attempt to construct an electronic bibliographic network in which one description is elevated over others is ill-conceived; and there is similarly no legitimate way to admit into the system only a single attempt to describe ideal copy, any more than one can limit the number of printed bibliographies of a writer or subject, since each involves the interpretation of evidence, and more than one responsible interpretation may exist.

The most satisfactory way of handling the situation is to require that every entry placed in the system be labeled as to whether it is an attempt to record a particular copy or an attempt at a more generalized description based on the examination of multiple copies; it goes without saying that the source of each entry should also be indicated. Then if one wishes, in consulting the system, to see what entries for the socalled ideal copy of a given book exist, one can call up those entries; or if one prefers to begin with, or place principal reliance on, the entry supplied for its copy by a particular library, one can proceed in that fashion. The user would be free to exercise judgment in selecting among the entries offered. In some cases one might wish to see the entry prepared by the library specializing in a given field (for an eighteenth-century American imprint one would want to see how the American Antiquarian Society had handled it); in other cases one might decide to look over all the entries for specific copies in the system, or perhaps only at those entries that attempt to generalize, depending on how reliable

one judges them and their stated sources to be. The crucial point is that users (to say nothing of cataloguers) will know where they stand when every entry is labeled according to what it sets out to do (that is, whether it refers to one copy or to a whole issue) and when all such entries are available for inspection. As the system grows in this way, it will increasingly be able to demonstrate the advantages that computer assistance can offer to bibliographers. Provision will have to be made, of course, for recording many categories of physical data about books,[39] and there will have to be a general understanding that interest in books as physical objects is not limited to early or "rare" books but embraces all books.[40] The ways in which physical data can then be manipulated—to learn the printing characteristics of a particular period, place, or printer, for example—will be far superior to what has been possible in the past. But this vast structure of information will be meaningless if it does not rest on the scrupulous recording of details exactly as they appear in particular copies, with any more generalized descriptions clearly labeled as such.

The same points naturally apply also to work on union catalogues or so-called short-title catalogues, which are the counterparts in printed or card form of the new computer networks. Such works contain entries for specific copies, but they also frequently contain consolidated entries that are meant to stand for a number of supposedly identical entries. The latter kind of entry comprises somewhat more evidence than

an entry based on a single copy, but the extent to which it can be regarded as an entry for an issue as a whole depends on the depth and breadth of the research undertaken. In using the National Union Catalog: Pre-1956 Imprints, for example, one is bound to wish, in certain instances, that the individual entries supplied by different libraries for a given book had been retained, rather than grouping the symbols for those libraries under a single entry. Supposedly, of course, this kind of consolidation was not undertaken except where there were no significant differences among the individual cards; but in some cases one would like to see just how the consolidated entry was arrived at and what the individual entries consisted of. Katharine Pantzer's exemplary revision of the Pollard and Redgrave STC relies at some points on considerably more detailed bibliographical investigation, with the result that discriminations among copies carry much greater authority. Yet the number of titles involved precludes equally detailed treatment of all entries; and the copies grouped together under certain entries might not be so grouped in a full-scale descriptive bibliography based on the textual collation of all known copies. No one expects a short-title catalogue, even one with the high standards Miss Pantzer sets, to be a substitute for descriptive bibliographies of individual writers or books. What one recognizes is that a union catalogue or a short-title catalogue (of the kind that goes beyond the confines of a single collection) occupies a middle ground between a catalogue of single copies and a true bibliography. Falling thus in between two relatively well-defined types of work, it must be careful to define its exact position: the standards for deciding when two copies are to be accorded two separate entries rather than a single entry must be made clear if the user is to know how to interpret a given entry. The eighteenth-century short-title catalogue now in progress faces such an enormous task that the quantity of detail examined for each item cannot be as great as might be feasible for a smaller project. Short-title catalogues can operate—and serve an important function—at various levels of bibliographical detail. But the level at which a given catalogue is to be pitched is a central decision in establishing the catalogue and a central fact to be announced in it. In confronting an entry, the user needs to know what level of discrimination has produced it: one needs to know, in other words, what kinds of differences among copies have not been taken into account, and what kinds therefore may exist among the copies cited within a single entry.

One cannot generally think of the entries in such catalogues as ideal copies, because frequently insufficient detail is provided to distinguish among states; on the other hand, the entries often represent more than one copy and rise above what are deemed to be the peculiarities of those

copies. They are ideal copies of a kind, then, but not up to the standard implied by ideal copy in descriptive bibliography. As an intermediate category, they help to show up the problems with the phrase "ideal copy," because it is awkward to try to conceive of a quasi-ideal or a semi-ideal. A convenient term used in discussions of library cataloguing to refer to features of particular copies is "copy-specific"; when one generalizes on the basis of copy-specific details to produce a description that encompasses the individual copies, the result should logically be something on the order of "issue-specific" rather than "copy-ideal." Such descriptions are in fact generalizations or conclusions about the issue as a whole, based on the evidence of specific details. And the differences among short-title-catalogue entries and bibliographical descriptions rest on two variables that affect the nature of the generalization attempted in a given instance: the level of detail examined in each copy, and the number of copies surveyed. Short-title-catalogue entries by definition are limited in detail, and they may or may not be limited in the number of copies examined; but their function is not, as with bibliographical descriptions, to provide detailed printing and publishing histories. Nevertheless, they are like bibliographical descriptions in going beyond the evidence of single copies: the difference between them is of degree rather than kind. Whether the generic term for what they contain is "generalized description," "standard description," "issue description," or "ideal-copy description" is not of paramount concern. What is important is maintaining a clear distinction between that kind of description and "single-copy" or "specific-copy" description. The distinction between the single unit and the group of which it is a part is basic to all bibliographical and textual research; it must inevitably be pervasive in any endeavor concerned with the physical aspects of books.

Notes