University of Virginia Library

The Arrangement of Descriptive Bibliographies
by
G. Thomas Tanselle

On 25 October 1963 the Times Literary Supplement published a short article entitled "The Soho Recipe," taking the appearance of Richard Fifoot's bibliography of the Sitwells as an occasion for reflecting on the series of Soho Bibliographies (p. 876). That series, begun in 1951 by Rupert Hart-Davis,[1] has been influential in the twentieth-century development of author bibliographies, particularly in regard to the division of material into sections and the assignment of reference letters and numbers to individual items within those sections. If in some respects the Soho volumes were uniform enough to suggest that a "recipe" lay behind them, in other respects their practice varied, as the TLS article pointed out;[2] and the questions raised in that article stem from the fundamental issue of whether uniformity of plan and technique among bibliographies is desirable. The article recognizes that without fuller rules the practices of each Soho volume are bound to "depend as much upon the temperament of the bibliographer as upon the tractability of the material," and it understands at the same time that more detailed rules could be "oppressive." Yet it cannot resist wishing for more uniformity in the treatment of certain matters, asserting that the reader has a right to find "agreement, or explicit reasoned disagreement, when two or more bibliographers describe the same book."

There is, in fact, good reason not to believe that the reader of a bibliography has any such right, and the line of argument is implicit in what is said elsewhere in the article. Although the piece unfortunately

does not pursue these issues very far, it is essentially enlightened in its approach, for it sees that questions regarding the presentation of material in a bibliography cannot be divorced from a consideration of the particular nature of the sources involved and the individual aims of the bibliographer. Differing Soho treatments of publishing history and "the biographical element in bibliography," as the TLS writer correctly states, result not simply from the extent of the preserved documents in each case but also from "the degree of enthusiasm and understanding which the individual has brought to his task." The "understanding" is crucial: when one understands that bibliographical research is historical research and that bibliographies are historical studies, one perceives that biography and publishing history are not separable from bibliography. One should also then see how naive it is to suggest that different treatments of the "same book" ought to agree or be in "explicit reasoned disagreement." Different bibliographical accounts of a book, even of the same copies of a book, may have different emphases, details, and arrangements, in the same way that accounts of any other kind of historical event are likely to vary according to the historian who is ordering the material. And if specific facts and figures are at odds, there are explanations other than that an error has occurred: perhaps the copies examined do not match, or perhaps the examiners have observed different tolerances in measuring or in reporting measurements. Sometimes scholars, in bibliography as in any other field, may find it appropriate to comment on the differences between their own accounts and those of their predecessors; at other times they may feel that the defects of previous reports, and the advances made by their own, are obvious and that the point requires no explicit statement.

In some respects the TLS writer proceeds admirably from the recognition that bibliography is a branch of historical scholarship and calls attention to the fact that bibliographies vary according to the interests and points of view of their authors; at times, however—as in the expectation that two descriptions of a book will agree—the piece seems to reflect the narrow, but quite common, view that bibliographies are merely compilations, rather mechanically assembled. That a descriptive bibliography is actually a historical study has been understood since at least the time of Sadleir's Trollope bibliography (1928), which explicitly makes the case for an author bibliography as "a commentary on the book and publishing crafts" of a particular period. An author bibliography (as opposed to a simple guide to the identification of first editions) is a historical account of the production and publication of an author's works—and is therefore a partial biography of the author and a partial

history of publishing in the period.[3] Its value, as with any other presentation of the results of historical research, depends on the thought that has gone into the assessment of the evidence and on the coherence of the overall conception, which determines the degree of relevance of specific data and the relative space and detail to be accorded them. What constitutes responsible handling of evidence and effective presentation of conclusions is the same in bibliography as it is in any other historical investigation; in bibliography as in other fields different people may find different patterns in the material, and more than one account can be responsible. (Whether or not any of them has arrived at the "truth" will of course remain an open question.) Fredson Bowers was certainly right, in his Principles of Bibliographical Description, to speak of bibliographies as "written," not "compiled."[4]

All this is a necessary preamble to a discussion of the arrangement of material in a descriptive bibliography. Without seeing bibliography in the context of historical research in general, one might infer from such a discussion that definite rules for the arrangement of bibliographies could be prescribed. The fact that some parts of bibliographies are set forth in a formulaic way and not in straightforward expository prose has no doubt contributed to the notion that a bibliography is a compilation, subject to mechanical rules. What would it mean to ask how a history of a battle or a biography of a statesman should be arranged? Methodological reflections on the writing of such works can be useful, one but cannot set up a specific order in which the material must be presented.[5] This obvious point has not always seemed so obvious in relation to bibliography. But once a bibliography is recognized as a history, one must further recognize that the arrangement of its contents has to grow out of the nature of the material and out of the bibliographer's own approach and aims. There are always people, in all fields, who wish

to be told what to do and who will proceed unthinkingly to follow instructions. Whether bibliography has more than its share of such people is hard to say; certainly anyone who has been associated with the publishing activities of a bibliographical society knows that requests come in from would-be bibliographers who want to know what arrangement of material the society prescribes for the bibliographies it publishes. Would a biographer ever write to a publishing firm asking whether its biographies were to be chronological or thematic in arrangement? Probably not, but in any case the point is clear: scholars of intelligence, who understand what they are doing, will work out in each instance the arrangement of material that serves their purposes.[6] Persons who need to be told how to organize their material have not given sufficient thought to their subject to produce useful results.[7]

These, then, are reasons for not writing on the subject of the arrangement of bibliographies. In spite of them, I propose to do so, now that I have tried to make clear the context for my remarks. Discussions of methodology can be salutary in any field, and I intend here simply to set forth some of the considerations involved in thinking about the problem of arrangement in bibliography. My comments are meant to encourage such thinking, not to imply that a single scheme of arrangement can, or should, be declared standard for all situations.[8] The present

moment seems an appropriate one for examining these matters, because in recent years some bibliographies have appeared that focus new attention on the question of arrangement—particularly as a result of increased examination of twentieth-century authors[9] and the attendant problems created by newer methods of plating. The first order of business is to look into the arrangement of the descriptions of the various editions, impressions, and issues of a single work. Then there is the question of how those composite entries fit into the bibliography as a whole—how, that is, the various kinds of material (books, contributions to books, contributions to periodicals, and so on) can be arranged in relation to one another. Finally, I should like to comment on the numbering of the entries, for reference numbering inevitably reflects the arrangement given to the material.

I

One of the most basic problems of arrangement in a descriptive bibliography results from the obligation to identify the relationships among editions, impressions, and issues of individual works. The problem of arranging entries for a single work is not entirely distinct from that of the arrangement of the whole bibliography: if one decided, for example, that chronology were to take precedence over all other considerations, impressions would be scattered under the years in which they appeared and would not be gathered together as parts of a single edition. One would still specify the edition relationships in words, of course, but the physical arrangement would be determined by chronology—which is to say that classification is a different matter from arrangement (even though the clearest arrangements are likely to be those that reflect classification). Most bibliographers do, for obvious and good reasons, adopt a chronological arrangement within each section of a bibliography.[10] But chronology does not dominate, if it prevails only within each section. Furthermore, chronology is violated within a section if all editions and printings of a given work are grouped together. Readers are legitimately interested in the full history of a single work, but there is also a defensible interest in what happened in each year of an author's career—the mixture of first printings of new works and later printings of earlier works.[11] Since there is reason to have the material both ways, the

approach chosen as the basic one should be complemented by an index that provides access from the other approach. One can argue that normally the most sensible basic arrangement is one that brings together all printings of a single work, for each entry can then build efficiently on those that went before and can conveniently be used in conjunction with them. More is gained by having each impression placed in the context of the printing and publishing history of the work than is lost by having the impressions of different works in the same year dispersed and retrievable only through a chronological index. This issue has rarely posed a problem, and bibliographers of authors have generally followed this plan, taking up an author's separately published works in the order of their publication and completing the publication history of each work before passing on to the next work.[12] However, they have not as routinely provided a chronological index, which—I think it hard to deny—ought to be a regular accompaniment to this scheme of arrangement.

Where a difficult problem can enter is at the next step, after one has decided to group together the entries for the editions and printings of a single work—because the best order for those entries is not always readily apparent. There are many simple cases, of course, in which the order implied by the classification into editions, impressions, and issues can be followed without question: the successive impressions (subdivided into issues where necessary) of the first edition are followed by those of the second edition, and so on. But complications frequently arise that tempt one to add further organizing principles, superseding those inherent in the basic terms of classification. Although such situations occur fairly often in connection with books of the past two centuries (because of the growth of transatlantic republication of writings in English and the use of plates), earlier books can pose these problems as well. Even the chronological ordering of editions can be brought into question when, for example, a work appeared in a succession of editions on both sides of the Atlantic (and perhaps in other English-speaking areas). Should the editions be entered chronologically under each country, thus placing geography ahead of chronology as an organizing principle? Should this principle be followed when duplicate plates from

one country are sent to another country, thus placing geography ahead of the classification into editions? Should it be followed when sheets with cancel titles are exported to another country, thus placing geography ahead of the classification into impressions? Within a single country, when duplicate plates are used during the same period of time by two publishers (as might happen if one publisher had the paperback rights or the rights to include the work in a cheap series), should impressions be grouped under each publisher, thus subordinating the chronology of impressions to the classification into plate families? Is the situation similar when a single publisher designates certain printings as part of a series? If plate families do not consistently coincide with particular publishers (or series), do the plate families or the publishers (or series) take precedence?

These questions suggest that the problem of arrangement is not entirely separable from that of classification. (Arrangement need not follow classification, but an arrangement implied by the classification cannot be hastily dismissed, either.) And some people have regarded such questions as posing a serious challenge to the validity of the traditional system of classification, at least for books of the past two centuries. I do not think it necessary here to reopen the whole question of classification, which has in any case been extensively discussed already.[13] But I do think it appropriate, at the cost of some repetition, to affirm that 19th- and 20th-century printing technology, in which plating, photography, and computerization have reduced the need for resetting, has not made the edition-impression-issue classification any less relevant or helpful. The fact that many works printed in those years never reached a second edition because various means were available for reproducing the original typesetting does not render the concept of edition superfluous: it refers to the basic fact that all copies deriving from a single act of typographic composition, however diverse they may appear to be (even to the point of exhibiting different typefaces), are inextricably linked by that common origin. That works in sufficient demand to require more than a single press run were likely to appear in earlier centuries in a succession of editions (not subdivided into impressions) and in later centuries in a succession of impressions (of a single edition) suggests that the scheme of classification—far from being defective—is serving (as it should) to reveal fundamental differences in the physical relationships among copies in different periods.

A question of more serious import concerns the relative emphases on printing and on publishing in the classification. There are times

when printing history would entail one arrangement and publishing history another, and bibliographers are faced with deciding which arrangement should take precedence. When they look at the standard classification to determine whether it is built on one of these approaches rather than the other, they find that it involves both and are then sometimes critical of it and may decide to revise it, separating terms relating to printing from those relating to publishing.[14] Without doubt the edition-impression-issue hierarchy does involve both considerations, but whether that fact points to a flaw is far less certain. What the bibliographer is describing is published (or distributed) pieces,[15] since the activity of publishing postdates that of printing. Yet grouping copies according to publisher or to publisher's imprint dates (or copyright or copyright-page dates) leaves something to be desired, since impressions are sometimes not differentiated in these ways by the publisher but are always a fact not simply of printing history but of publishing history as well.[16] Because publication is an activity that involves the distribution of physical objects, the study of publication cannot be divorced from an understanding of the production history of those objects. It is not surprising, therefore, that the basic terms edition and impression refer principally to printing but that publishing enters more explicitly into the division of impressions into issues: a cancel title with a new date, for instance, does not affect the printing history of the work, for the sheets are of the same impression as those with the earlier title; but it reflects a publishing (that is, a marketing) decision, and the copies with the new title form a discrete publishing unit. Publishing decisions frequently leave their mark on the physical books; and even if the bibliographer's aim were limited to the classification of physical books, publishing history could not be divorced from printing history in the classification.

Even though the mixture of printing and publishing considerations cannot be regarded as a defect in the system—it is, rather, a necessary element—there remain situations in which an emphasis on printing in the ordering of entries results in a different arrangement from the one that would emerge from an emphasis on publishing. Although this dilemma derives ultimately from the fact that the classification involves

both printing and publishing, one must recognize that the practical question of what to do in such instances is usually concerned not with the classification but only with the physical arrangement on the printed pages of the bibliography. Thus if one were to decide, in a case where an American publisher issued English sheets with an American imprint (on a cancel or an integral title), to place the American issue at the head of a line of American issues and impressions carrying the same publisher's imprint, one would not be quarreling with the classification of the American issue as an issue—as a part of an impression—but would simply be emphasizing, through physical arrangement, the publishing history over the printing history. It is crucial to remember that these questions do not require the choice of one approach at the expense of the other: the main entry can be placed where it would fall in the publishing (or printing) history, and a cross reference can stand at the appropriate spot to reflect the other approach (or a separate chart or tree can delineate the second approach).[17] Still, one has to decide on one place or the other for the principal statement, and the decision is not without significance; it should be made thoughtfully in terms of efficiency of presentation, convenience for use, the overall conception of the work, and the particular nature of the materials involved. That more than one physical arrangement, reflecting somewhat different classification schemes, can be compatible with the larger edition-impression-issue framework is to be welcomed, not regretted; such flexibility within an accepted convention allows one to accomplish the basic task of classification in a way that can readily be understood by others and at the same time permits one to present the account in a way that emerges naturally from the material.[18] Only those who hope to avoid thinking (and who believe that bibliography can be a mechanical process) would wish to be told that a predetermined arrangement is obligatory.

Nevertheless, the importance of having an established level of classification that falls between edition and impression is undeniable, for it would supply a standard means for placing together (both in thought and on the printed page) any impressions that constitute a separate

group within the entire series of impressions. Fredson Bowers recognized this point years ago, in his Principles of Bibliographical Description, when he discussed the concept of "subsidiary edition" (or, for short, "subedition" or "sub-edition")—a concept that has been insufficiently examined and used since then. A subedition, as Bowers conceived it, would be formed by impressions with a different publisher's imprint, or those with an added indication of series, or those that might popularly be called "revised edition" or "enlarged edition" (that is, those in which the text is somewhat altered, but without enough resetting to be considered a new edition, or somewhat augmented, as with a preface or a new chapter). These kinds of changes result from publishing decisions: they generally represent the kinds of alteration that would produce issues if they affected only parts of impressions rather than whole impressions. The concept of subedition therefore makes formal provision for incorporating publishing considerations into the edition-impression classification, which is based primarily on printing considerations. One could, of course, enter all the successive impressions of an edition in a single sequence, even if all the impressions from a particular publisher, or in a particular series, did not fall together—as would happen, for instance, if the original publisher continued to place new impressions on sale while a second publisher produced impressions for a cheap series. Even though one could then supply a separate outline or stemma to show the groupings of impressions according to publisher, series, or the like, the basic arrangement would be awkward, since it would not explicitly recognize that printings from a single publisher (or in a single series) normally are more closely related with one another than with printings from another publisher (or in another series). Acknowledging that impressions fall into groupings by virtue of the publishing process leads to the belief that publishing history should be allowed to supersede the strict chronological ordering of impressions and that the subedition classification should be regularly employed. Whereas most options regarding physical arrangement are best left open for the bibliographer to determine in each situation, the value of allowing certain sequences of impressions to show up clearly as distinct groups lends support to the idea of attaching to them a term—like subedition—that indicates a classification intermediate between edition and impression and thus implies a particular arrangement.

Any suggestion of a fixed arrangement, however, is not without its problems. Since subedition refers to publishing history, its use involves a potential conflict with printing history, if the groupings of impressions by sets of plates does not coincide with the grouping by publisher

or series (or other characteristics encompassed by subedition).[19] This fact has recently been explored by James L. W. West III, who suggests that the concept of plating should supplant that of subedition.[20] His article on this subject includes a valuable summary of the means thus far proposed for detecting the use of various kinds of plates; but his presentation of the concept of plating requires some further consideration, which I hope can serve as an example of how one may deal with conflicts of classification that in turn produce alternative physical arrangements. West argues that the grouping of impressions into subeditions, following Bowers's plan, with appended notes as necessary to identify replatings, "works satisfactorily for books with a simple history of plating but is not as efficient in more complicated situations" (p. 257). In addition, he says, "the root word of sub-edition is edition, a term which implies that type has been reset." Questions of terminology are obviously of a different order from questions of concept, and West is well aware of the fact, for he explicitly points out that any term can theoretically be assigned any definition. It is surprising, therefore, that he should introduce this second objection to subedition, which is an argument against using the term to refer to a group of impressions of an edition, for such an argument is irrelevant to the more substantive question of whether a publishing or a printing concept should provide the intermediate classification between edition and impression, at least for purposes of arrangement. Without explicitly raising that question, he in fact opts for printing history, for he proceeds to say that "bibliographical terms, if they are to gain general acceptance, should if possible describe what actually happens at the printing shop." This statement is not self-evident, because the bibliographer deals with published objects, and an equally good case could be made that publishing history should dominate over printing history when the arrangements deriving from the two are in conflict.

There is something askew in the argument that subedition, incorporating

the word "edition," carries "mental associations for a bibliographer which are inappropriate to what the printer has actually done" and that a "new term would be helpful"—for subedition was not meant to refer to what the printer did and is perfectly adequate to designate the various publishing practices it was intended to cover.[21] The real question is not whether the term is satisfactory but whether the concept it refers to is the most appropriate one to serve as the basis in a bibliography for the primary subdivision of an edition into distinctive groups of impressions. Although West asserts the primacy of the plating classification[22] and of the arrangement it suggests, he does recognize that in particular instances one may not manage to determine whether separate sets of plates were used. He says, "If plating did not occur (or he cannot prove that it occurred), then the bibliographer will simply use edition, impression, issue, and state as he always has" (p. 258). Does he mean that even in such situations differences in publishing arrangements are not to be acknowledged by a grouping into subeditions?[23] Whatever was meant, I believe that this question brings us to a significant point. Since subedition largely involves the kind of difference that within an impression would produce an issue, subeditions are usually recognized as such, for the differences were intended to be noticed as distinguishing discrete publishing efforts.[24] The fact that one generally knows when one is dealing with a subedition and frequently may not know when there are duplicate plates cannot in itself be an argument for or against a specific classification, for the desirability of a classification is not affected by one's failure to ascertain the information needed to apply it. Nevertheless, the bibliographer deals with surviving physical objects— normally objects that have passed through a publishing process or have in some way been distributed—and the groupings into which they fall, within an edition, as a result of the publishing process would seem to be the natural basis for arrangement.[25] Details of production history, including the identification of platings, are an important part of the total story and should certainly be recorded; but it would be hard in most circumstances to make a convincing case that those details rather than publishing details should form the primary basis for arranging groups of impressions in a bibliography. One can of course concentrate on the printing history of a given work and write a study of that history. But bibliographers ordinarily do not stop there; instead, they concern themselves with physical entities that made their way into the world and are now before us—entities that may be made up of parts with differing printing histories but that present themselves to us as published objects. This fundamental fact must affect all thinking about bibliographical classification and arrangement.[26]

One might then ask why publishing details should not be dominant throughout, instead of adopting as basic the concepts of edition and impression, which refer in large part to printing history. Perhaps the most direct way to begin answering this question is to say that bibliographers actually use these printing terms as adjuncts of publishing history and are not focusing any less consistently on publication by adopting them. Indeed, the incorporation of subedition into the hierarchy illustrates the point. The bibliographer is interested not simply in identifying the impressions that derived from a single setting of type but in showing the varying publishing auspices under which they were released to the public. When, for example, a second publisher releases new impressions from the same setting as those released by the first publisher, the bibliographer cannot refer to the group of impressions thus created as an "edition," since that term has a well-established meaning that is broader. The adoption of a term like subedition reflects the bibliographer's view of an edition as a series of publishing units. Now it must be admitted that if this line of argument were carried to its logical conclusion, the result would be likely to satisfy no one: if, for instance, the seventh separate printing of a work to be released by a given publisher were actually the first impression of a second edition, it would be pointless not to reflect that fact in arrangement simply because it was judged to be a part of printing history, not of concern to the publisher, for whom all the impressions formed a single succession. In fact, the decision to order a new typesetting is a significant publishing decision,

just as the decision to order a new printing from previously set type or from plates is a publishing decision. Although one can recognize a kind of logic in the notion of recording printing facts in one sequence and publishing facts in a separate sequence, a bibliography produced on this plan would be extremely difficult to use: books are so inextricably the product both of printing and of publishing that classification and arrangement should be expected to take both into account. Separate outlines of the two approaches might provide helpful guidance in some complicated situations, but the basic historical account must attempt to deal with both together. Favoring the publishing concept of subedition over the printing concept of plating as the primary classification intermediate between edition and impression for purposes of arrangement is consistent with the thinking underlying the process of bibliographical description as a whole. And, one might add, it does not mean that the question of plating should be investigated or reported upon any less thoroughly.

The two approaches to arrangement can be efficiently illustrated by the simple example that West focuses on in his discussion of plating, William Styron's Set This House on Fire (1960). The two impressions published by Random House from that firm's original relief plates naturally come first. In West's proposed arrangement, the next entry is the impression distributed by the Book Find Club in 1960, because it was printed from new offset plates deriving from the first Random House impression, and it therefore represents the "second American plating" of the original edition. Following that comes the 1971 Random House impression, because it was not printed from the original Random House plates but from new offset plates made from the second impression sheets—and it therefore becomes the "third American plating":

• First American plating (relief): Random House First impression, March 1960 Second impression, September 1960
• Second American plating (offset from first American plating, first impression): Book Find Club First impression, July 1960
• Third American plating (offset from first American plating, second impression): Random House First impression, March 1971

Arranging the entries according to the chronology of plating means that the impression distributed by the Book Find Club intrudes itself into the sequence of impressions published by Random House.[27] Since only a single impression of the Book Find Club is involved, there is no great inconvenience here. But in a more complicated situation, there might be a succession of impressions released by a second publisher during the same time that the first publisher was also producing a series of impressions, and if one or more additional platings were ordered by either publisher, there could be a considerable amount of interspersing of the entries relating to the two firms.[28] An arrangement by subeditions would avoid this awkwardness, though obviously at the price of making the sequence of platings less readily discernible:

Parent edition: Random House[29]

• First impression (original relief plates), March 1960
• Second impression (original relief plates), September 1960
• Third impression (offset from second impression), March 1971

Subedition: Book Find Club

• First impression (offset from first Random House impression), July 1960

This system allows for the orderly expansion of the list of impressions under each subedition, whereas the other emphasizes the orderly recording of platings. There is no question in either case of concealing any information, nor even of whether one system is a better classification than the other, for each emphasizes significant facts and both can be defended. The only question is which provides a better arrangement for the printed pages of a bibliography, where one entry must follow another (unlike a stemma, where simultaneous sequences of events can be shown side by side). What gives an edge to the subedition approach is, as I have suggested, its emphasis on books as published products. Since the impressions coming from a single publisher or published in a single series are likely to be closer to one another in physical appearance (title-page design, binding, and so on) than they are to the impressions from another publisher or series, a practical result of arranging entries by subedition is that similar impressions, many of which can be treated in abbreviated form ("the same as X except for . . ."), are grouped together and cross reference among them facilitated. As concepts of classification, subeditions and platings are both unquestionably important; as a plan of arrangement in a bibliography, giving priority to subeditions is likely to be more appropriate.

Having arrived at this point, we are still left with what many would regard as the chief conflicts that produce difficulties of arrangement. I believe, however, that the line of thinking followed thus far prepares us for considering these other problems. Perhaps the central question —for it involves the others—is how to treat publication of an edition in more than one country.[30] Geography is not an element in the basic classification scheme, but publication history is, and it would obviously be possible to gather entries together according to country of publication, thus setting geography ahead of chronology as the primary organizing principle. If editions were never shared between countries, the problem would be relatively simple, though the best arrangement would still not be an obvious or easy matter to decide. What would be involved is whether a single chronological sequence of editions would be more, or less, useful than separate chronologies of editions under each country. Strong cases could be made for both approaches in particular instances, and I do not think a single rule on this point could fit all situations. The problem is often more complicated, however, for editions frequently are split between countries: among the possibilities are sheets issued with cancel title leaves and new impressions produced from offset plates. The further question in these cases is whether the description of a single edition, or a single impression, should be split up because parts of it were published in different countries. This question is significantly different from the one concerning the ordering of whole editions, for editions by definition are discrete entities, originating in separate acts of typographic composition. One can more comfortably contemplate the rearrangement of whole entities than a scheme of arrangement that necessitates the splitting up of entities and the entering of the resulting parts under different headings. The latter may be defensible, but the question is a more ponderable one.

The difficulty is to some extent illustrated by West's example. In discussing Set This House on Fire, West asserts, "In a descriptive bibliography, it would be much better to separate the American publication history distinctly from the British" (p. 264). After listing the three platings of the "First American Edition," he proceeds to a parallel heading, "First English 'Edition,'" and records two further platings (Hamish

Hamilton, 1961, and Jonathan Cape, 1970). Since of course the English platings do not constitute a different edition, he is forced to place the word "edition" in quotation marks in his heading and thereby to acknowledge that the American and English headings, though assigned parallel status, are not parallel in concept—unless one changes the concept from one based on typographic composition. In this particular instance, because of its simplicity, the grouping together of the two English platings (or subeditions, for in this case they coincide) causes no difficulty; indeed, they would appear at the same point (but without the heading) in a chronological arrangement, assuming that the 1971 plating were included with the other Random House impressions. But the awkwardness of the "edition" heading calls attention to the inherent problem, when impressions are split off from the edition of which they are a part. And there is is the further problem—not illustrated by this example from Styron—of parts of impressions (i.e., issues) that are released in different countries (and presumably by different publishers). Segregating entries by country may therefore require splitting up impressions as well as editions.

Although I do not hesitate to recognize—as I have already pointed out—that bibliographical classification and arrangement involve a mixture of printing and publishing concerns, I still find something anomalous in the proposal at once to classify according to plating and to arrange according to country. Whether one chooses to give precedence to platings or to subeditions as subdivisions of each edition, one will then find in many cases that an arrangement by country works against such classification, since impressions from the same plating may be published in different countries, as may issues of an impression from a subedition. One might at first think that arranging by country would entail no more problems than arranging by subedition, since each publisher (or series) determining a subedition must be located in some country. The significant difference is that nationality adds another level to the classification and causes editions—not merely subeditions or platings—to be broken up. West rightly observes, "Increasingly in modern books, the American and British manufacturing histories overlap or derive from one another" (p. 265); but this fact would appear to be an argument against, rather than for, the separation of the records for the two countries. In all but the simplest cases it would probably be helpful to provide an outline of the publishing history in each country, with references to the relevant entry numbers. The question at issue is not whether the publishing history of a book in each country should be conveniently accessible but whether nationality provides the most desirable arrangement for the descriptive entries themselves. Unless we are to abandon

classification by edition (and I see no justification for that), it would seem that keeping together all the entries dealing with parts of a single edition, rather than separating them by country, would in most cases make those entries easier to use and would present a more coherent account of the international intertwining of the printing and publishing histories of individual editions. In proposing this generalization, I have no wish to deny the usefulness of a country-by-country outline (in effect, a form of index to the entries) or to rule out the possibility of arranging whole editions by country of origin. The latter would serve little purpose if the former were provided; but my point is the desirability of avoiding the dispersal of the descriptions relating to a single edition.

This approach can be extended to the treatment of issues: if impressions (or groups of them forming subeditions) are to be described under the editions of which they are a part, issues should be described under the impressions of which they are a part.[31] Scattering the description of a single impression poses the same problems as breaking up the entries for a single edition. If an issue bearing the imprint of a second publisher were given an entry independent of the rest of the impression on the ground that a different publisher is involved, the issue would have been accorded the status of a subedition, and yet its relationship to the parent edition would be very different from that of a subedition. As long as we are not giving up the classification into impressions, it would normally be desirable—for similar reasons to those already offered for the treatment of whole editions—to keep the descriptions of whole impressions intact. When the publisher of an issue turns up elsewhere in the record of an edition as the publisher of an impression, cross references could be provided, if the situation is complicated enough to warrant them, in order to facilitate reference to the full role of a given publisher in the history of an edition.[32]

A rule of thumb, simply stated, therefore might be to treat editions and impressions as entities, not splitting off any of their constituent parts for description under some other heading.[33] This rule would allow

for subeditions (or alternatively for platings), since any rearrangement they involve is of whole impressions, and it would not prevent the placing of nationality ahead of chronology for the arrangement of whole editions.[34] But in preserving intact the descriptions of individual impressions and editions, it fosters coherence in the basic historical account, which can then serve as a point of reference for any supplementary outlines or other aids that one may wish to provide. The following list of entries shows—for a purely hypothetical edition—what the arrangement might look like in a specific instance:

Knopf parent edition

• First impression (relief plates): March 1921
• Second impression American issue: July 1921 British issue: Cape, August 1921 [For Cape and Florin Books subeditions, see below][35]
• Third impression: June 1924
• Fourth impression (offset plates from third Borzoi Books impression): April 1930
• [For Knopf Borzoi Books and Knopf (London) subeditions, see below]

Boni & Liveright Modern Library subedition

• First impression (offset plates from second Knopf impression): December 1921
• Second impression: June 1922

Cape subedition

• [For Cape issue of second Knopf impression, see above]
• First impression (offset plates from second Knopf impression): January 1922
• Second impression Cape issue: November 1922 Cape Florin Books issue: March 1923 [For Cape Florin Books subedition, see below]

Cape Florin Books subedition

• [For Cape subedition and Florin Books issue of second Cape impression, see above]
• First impression (offset plates from second Cape impression): July 1923

Knopf Borzoi Books subedition

• [For Knopf parents edition, see above]

---

20




• First impression (offset plates from third Knopf impression): August 1924
• Second impression: February 1926
• Third impression: December 1927
• [For Knopf (London) subedition, see below]

Knopf (London) subedition

• [For Knopf (New York) impressions and Knopf Borzoi Books subedition, see above]
• First impression (offset plates from third American Knopf impression): July 1927
• Second impression (offset plates from first British Knopf impression): September 1928

If this edition were organized by country, impressions would have to be broken up; and if it were organized by plates, the sequence of impressions by publisher would be disturbed. It could not be arranged by country and by plates at the same time, any more than the present arrangement by subedition is compatible with either of those approaches. Something must be sacrificed in the basic arrangement, but the loss can be made up by descriptive phrases and cross references at appropriate points, as suggested here, supplemented by separate schedules or stemmata and a thorough index. A central idea underlying the arrangement shown here is that the orderly listing of impressions by subedition allows for the most efficient use of a bibliography, given the fact that impressions from one publisher or series are likely to display more affinities in physical details with each other than with impressions from another publisher or series. This belief may not hold true in every case, but I think it carries enough weight to be worth proposing as a norm. Anyone who has thought through the rationale for this approach will be in a position to see when another arrangement would be preferable. A successful arrangement must be the product of careful thought about the material, not automatic dependence on a predetermined pattern.

II

Having decided the arrangement of the material relating to a single edition and to the editions of a single book[36] leaves untouched the larger question of the arrangement of the bibliography as a whole—the arrangement, in other words, of the entries for all the various publications that are a part of the author's career. That this question may seem less urgent suggests how entrenched a particular approach to overall arrangement has become and indeed is an interesting reflection of the

history of the development of author bibliography. Most bibliographies today, as in the past, begin with a section describing the author's separate publications (whether book, pamphlet, or broadside), usually in chronological order; this section is followed by one (or several) recording books contributed to in some way by the author and then by one listing contributions to periodicals. The precise plan varies, but the general movement from separate publications to periodicals prevails.[37] The emphasis on books, with the concomitant subordination of periodical contributions, is achieved both by placing the books first and by according them detailed description, in contrast to the simple listings provided for the periodical pieces. Although one must admit that books generally are of greater significance and influence than periodical contributions, one must also acknowledge that this arrangement is an outgrowth of the original role of bibliographies as guides for collectors, who were (and have continued to be) more interested in books (including those containing the first book publication of pieces by the collected author) than in periodicals. Even though the trend from Sadleir onward has been to recognize bibliographies as works of historical scholarship—which in fact serve collectors much better than the overly simple guides did—the plan of arrangement has never outgrown its origins and comes to us now trailing its Wise-Sadleir-Soho lineage.[38] It is thus so widely accepted that it is generally not even thought of as an issue, whereas the treatment of an edition on a level of detail that would involve subeditions and platings is new enough to seem a pressing problem. The arrangement of individual works, however, has a more profound effect on the shape of a bibliography as a whole than does the arrangement of the editions of those works and their impressions.

Actually there is much to be said for the traditional arrangement of works in a bibliography, even if it is not what historically underlies the arrangement. Certainly collectors interested in writers' careers should not have neglected periodicals, for contributions to periodicals may have been more important in establishing the authors' reputations than were separate book or pamphlet publications. Similarly, anthology appearances,

also neglected, may play a greater role than book-length publications in extending a writer's influence.[39] Even so, form of publication is a basic fact that can serve as a useful principle of organization. Placing books first, contributions to books second, and contributions to periodicals third need not imply decreasing interest for collectors; it can simply refer to basic differences in type of publication, differences that form a natural basis for division into sections.[40] But however natural such an arrangement may seem, one should be aware in using it of other possible approaches that are thereby being rejected. In particular, chronology is being put in second place, taking over only within each section and not determining the overall order of entries in the bibliography. One could argue that the course of a writer's career might be better shown by a single chronological record of all the works, regardless of their length or their form of publication, arranged according to the dates of the first publication of each.[41] Readers of bibliographies, accustomed to the convention of describing books in considerable detail and periodical pieces in simple listings, might object that an overall chronological arrangement would result in an awkward mixture of single-line entries and multi-page descriptions. This objection is of course based on purely formal considerations and does not affect the logic of the arrangement. Besides, there is no reason why contributions to periodicals cannot be described as fully as separate publications are: they are equally a part of the author's career, and some of them may be more important than some of the books and pamphlets. In making these points, I am not arguing that a single chronological arrangement is necessarily preferable to the conventional plan; indeed, I think that the usual division into sections according to the type of publication may well be more appropriate in many situations. What I do wish to suggest is that bibliographers who choose the conventional arrangement should give some thought to what they are doing and not adopt it simply as a matter of routine. As with the presentation of any other research, the form should grow out of the special requirements of the material. The conventional scheme, though it was shaped by an old-fashioned and narrow approach to book collecting, can still be justified, but it is not law. Bibliographers should think the question through in each instance, taking into account the problems raised by the career they are dealing with.

If one decides that there is good reason after all to separate the description of those books, pamphlets, and broadsides wholly or substantially written by the author from the treatment of composite publications (containing some work by the author along with work by other writers), one still has to consider what arrangement is best within these two categories. Whereas all the separate publications by an author—whether books, pamphlets, or broadsides—are usually kept together in a single chronological section,[42] the composite publications are normally divided into at least two sections, one for periodicals and one for books. The latter section (called "Contributions to Books" or a similar phrase) is sometimes further subdivided and in any case (whether subdivided or not) is likely to contain several kinds of items, as far as publication history is concerned: contributions (pieces written for, or at any rate first published in, a particular book—e.g., as an introduction or as a contribution to an anthology); collected pieces (writings for a periodical, now first collected in book form);[43] and appearances (pieces appearing in book publication for at least the second time). These distinctions are legitimate but have often been made by bibliographers primarily in order to focus attention on "firsts," following the lead of many book collectors in the older tradition. The first two of these classifications, referring to books containing the first publication in book form of pieces by the author, are sometimes described in considerable detail; but the items in the third group, not involving first book publication, are apt to be given abbreviated entries, or not even listed at all (as often happens with school and college anthologies). Frequently, however, this last category (especially the anthologies) plays a crucial role in the establishment of an author's reputation. And as far as textual significance is concerned, one cannot rule out any printing (even in an anthology) that appeared during the author's lifetime, at least not until one has investigated its publication history.[44] (Printings that prove to

have no significance for establishing the author's intended text are still of textual interest historically: it is important to know, for instance, precisely what text of a particular work was encountered by thousands of readers in a popular anthology.)

These points do not in themselves constitute an argument for or against placing contributions, collected pieces, and appearances together in a single section, but they should be kept in mind in thinking about the question. Firstness, after all, is not the only criterion of significance; and if non-first appearances in books are relegated to a separate section, the reason cannot be that they are less important than the other categories. However the entries are finally arranged, the bibliographer must make clear in some way which of these categories each entry falls into.[45] The trouble with some of the "Contributions to Books" sections in bibliographies is not that they mix unlike items but that they fail to comment adequately on the publication history of each item. Whether such a section should be subdivided (or whether indeed it fails to include enough and should incorporate contributions to periodicals as well) is a matter to be settled only by considering the nature of the author's publication record. In some cases separate sections will appear appropriate and helpful, and in other cases they will not. If division is decided upon, it will not necessarily be limited to the three categories I have mentioned, for further categories may be required by the material. Donald Gallup, for example, in his bibliography of T. S. Eliot (1969), gathers books containing letters by Eliot into a separate section. It is also worth noting that he places books edited by Eliot in the same section with books that have original contributions by him, whereas a bibliographer dealing with a different author might decide to create a separate section for books edited by that author. One might, in the case of an author who contributed many poems and stories to composite volumes, divide the entries by genre, recording the poems in one section and the stories in another. One might do the same thing, of course, for contributions to periodicals:[46] it is important to remember that a division between first printings and later printings or between verse and fiction is applicable to periodical pieces as well as to pieces in composite volumes and may on occasion be worth making for periodical pieces.

The central point, here as elsewhere, is to think about the categories that are most useful in a given situation and not to accept mechanically any plan simply because it has already been used for a different body of material and has achieved a measure of acceptance.

The amount of detail recorded for individual items is, as I have said, a separate matter from the arrangement of those items. Nevertheless, the two questions continually intersect, as the foregoing has made plain. Sometimes the implied reason for the separate listing of a certain category of material is the difference in the treatment accorded it, which in turn reflects its perceived importance. These factors are not entirely separable, therefore; and if a formerly subordinated category is shown to deserve fuller treatment, there is then no formal reason for its segregation, though there still may be a substantive reason. The general issue of the relative elaboration of detail deemed appropriate for different entries is still referred to by the label Falconer Madan gave it in 1906, when he spoke of "degressive" bibliography.[47] Although this is not the place for a full consideration of the "degressive principle,"[48] I think it fair to say that the idea of giving more extended treatment to some materials than to others is generally accepted; in any report of research—or in any piece of writing, for that matter—one must decide which are the principal aims or emphases of the work and which are the subordinate ones and then adjust the relative proportions accordingly. One cannot be faulted simply for giving less attention to certain matters or certain editions, for one cannot do everything; what is significant is the intelligence with which the decisions to emphasize or subordinate are made. Bibliography is not the only field in which such decisions are often made unwisely; but the frequent failure to understand that descriptive bibliographies are scholarly accounts, not mechanical compilations, has meant that some conventional practices in the field have been routinely accepted without sufficient examination of what rationale justifies them. Bibliographers do sometimes give evidence of having thought about why they have decided to describe in less detail editions published after the authors' deaths; but it is rare indeed to find that they have thought about why they simply list, rather than describe, contributions to periodicals or why in many cases they

describe books contributed to by an author in less detail than books entirely written by that author. Are not composite volumes books, with all the potentialities for variation, in both text and physical makeup, that other books have? And are not periodicals the same as composite volumes, being printed matter made up of contributions by various writers? Why, then, has the custom grown up of giving them slighter description, or none at all?[49]

In large part the reason is that collectors have not in the past been very interested in these kinds of material. Books containing original contributions (or even pieces first collected in book form) have elicited greater interest than posthumous editions or periodical pieces, and these degrees of attention are still often reflected in bibliographies (even those by bibliographers who understand that in part they are writing publishing history). A defensible argument can in fact be made for these proportions, but only on different grounds. As a chronicler of an author's career, one cannot justify downgrading periodicals or classroom anthologies or late editions, or even the less severe subordination of books with original contributions or with first reprintings in book form: all these publications have played their roles, small or large (often very large indeed), in the author's career. One may of course decide, as in any research, to focus on certain aspects of one's subject or on certain events or publications; but some forms of publication are not inherently less in need of full description than others, for all are printed items, in which production history and textual content continually intersect. The justification for giving some classes of publication less detailed treatment is more convincingly made on practical grounds: one may have to cut down somewhere in order to make the task manageable enough to complete at all. A natural place to cut may seem to be composite volumes and periodicals, for one may feel that these publications, involving many authors, are not primarily the responsibility of the bibliographers of the individual authors included in them. A considerable duplication would indeed result if such books and periodicals were extensively described in the bibliographies of all the authors appearing in them; the preferable situation would clearly be for bibliographers to write separate descriptive bibliographies of individual periodicals and of specific

groups of anthologies.[50] But research cannot be forced into such orderly paths, and scholars cannot very responsibly adduce the possible future duplication of effort as a reason for not pursuing in the present certain lines of inquiry relevant to their subjects. They can say that they are focusing on books wholly written by an author, rather than books and periodicals contributed to by that author, simply because they find the former a more interesting subject and because they do not have the time or energy to deal with the latter. Such reasoning is hard to quarrel with, even when it results in bibliographies that seem to many people to have obvious deficiencies. But one may indeed protest if the basis for proceeding in this fashion is the idea that periodicals and books with contributions do not deserve or require so thorough a treatment—or if the bibliographer is following, without examination, the notion that bibliographies are always done this way.

The same thinking applies as well to late or posthumous editions and printings of the author's own books. Some of the discussion concerning the degressive principle has centered on the question whether full attention to such editions and printings would not involve shifting the focus of the bibliography from the author to the printing and publishing history of a later period. As long as one recognizes, however, that an author's text is inextricably bound up with the printing and publishing process and that later editions and printings—with the particular texts they contain—are crucial for understanding the course of an author's reputation and influence, one cannot defend abbreviating their treatment on these grounds.[51] One can simply declare that one is dealing only with printings appearing during the author's lifetime, or during any other specific period, for one is naturally free to set—indeed, must set—limits to the scope of one's work. The decision to use (for instance) the date of the author's death as the dividing line between full and shorter entries can be respected (though not welcomed) if it is made for practical reasons—but not if it results from the bibliographer's belief that

sound bibliographical theory requires, or at least supports, the distinction and that later printings are more properly treated by the student of printing and publishing than the student of the author. My point, as before, is that bibliographers should think about each situation in its own terms and not complacently accept some preconceived plan. Degressive descriptions are certainly defensible, but the categories offering the best candidates for less detailed treatment are not necessarily the same from bibliography to bibliography. Whether the decision to give certain items less full treatment precedes or follows the decision to group the entries for those items in a separate section, the two questions are related. What we must avoid—or else bibliographical scholarship will suffer—is allowing a particular arrangement, which often implies particular levels of detail, to become so established that we accept it without thinking about how it contributes to the ultimate goals of our work.

III

Another question related to arrangement is how best to number the entries. The numbering can of course only reflect, and not determine, the arrangement of the material, and in many ways it is a trivial matter. Unquestionably there is a convenience in having entries numbered for reference, but one might assume that any system of numbering (or lettering), as long as it enabled one to locate an individual item, would serve the purpose. And so it would. Difficulties arise only when one wishes to make each number convey more information than simply a position in a sequence of entries. Bibliographers have been increasingly concerned with inventing entry numbers that reveal the place of an item in the general scheme of organization of the bibliography in which it appears. In early author bibliographies entries were either unnumbered or (as in Sadleir's Trollope) assigned simple consecutive integers, from 1 onward. Even after letters were introduced to identify the sections of a bibliography, the entry numbers still might be a straightforward numerical series within each section; thus in John Slocum and Herbert Cahoon's bibliography of Joyce (1953), entries 3, 4, 5, and 6 in section A all are concerned with Chamber Music.[52] A further development, generally associated with the Soho series and widely imitated in other bibliographies, was to assign a single number to each work and then to identify various editions with an attached lower-case letter ("A3b" would refer to the second edition of the third separately published

work by the author). This system was sometimes used illogically, with the appended letters occasionally referring to selected impressions (often subeditions) or issues as well as editions; and the Pittsburgh Series in Bibliography, beginning in 1972, sought to refine the system. Although the Pittsburgh volumes have varied somewhat from one to another, the general approach was set by Matthew J. Bruccoli, general editor of the series, in his bibliography of F. Scott Fitzgerald (1972). He employed references such as "A8.1.a*," in order to specify impressions ("a") and issues ("*") as well as editions ("1").[53] There can be no doubt that his reference numbers follow a more rigorously logical system and convey more information than those previously employed—though at the cost of some rather cumbersome combinations.

The discussion of entry numbering has not stopped there, however, for Craig S. Abbott, the author of another Pittsburgh bibliography (of Marianne Moore, 1977), has written an article on the subject, proposing a system that goes beyond Bruccoli's by taking subeditions into account (although he confusingly calls them "issues" and thus uses "issue" to mean two different things).[54] Each of his numbers contains a single period, which follows the edition designator; after the period, the figure indicating impression can be combined with lower-case letters to mark subeditions (if the letter precedes the impression number) and issues (if the letter follows the impression number), as well as with parenthetical numbers to mark states. An Abbott number might therefore look like "A9b.a6b(2)." The same number might even become "A9b.a6[b](2)," since italics specifies simultaneous publication[55] and brackets denote indeterminate order. Nor is this all, for if one accepts the suggestion that roman numerals be used for volume numbers and parenthetical letters for binding variants, one might have "A9b.a6[b](2)II(b)." Abbott's proposal has not put an end to the devising of systems: James L. W. West III, in his Styron bibliography (1977), sets forth and uses still another scheme[56] designed to convey almost as much, if in part rather different, information. An example of West's notation is "A4.I.a.1† b," in which

roman numerals refer to editions, lower-case letters to platings, arabic numbers to impressions, and superscript letters preceded by daggers to issues (superscript letters preceded by asterisks signify states). Both Abbott and West recognize the need for a level of classification between edition and impression, but one chooses subeditions and the other platings. As we have already noticed, these two approaches cannot be employed simultaneously for the actual arrangement of entries; they could, however, be combined in reference numbers, but such numbers—containing elements that do not relate to the position of the entry in a bibliography—would raise the fundamental question of what the function of reference numbering is. Indeed, the systems of Bruccoli, Abbott, and West already raise this question, for each seems to be moving beyond the use of the number as a locator for an item and toward the idea that the number is part of a notation system for recording in shorthand fashion many of the significant aspects of any copy of a book.

One may begin to think about this matter by observing that the complexity of a reference number does not increase its usefulness for locating an entry in a bibliography. Simple whole numbers, unencumbered with letters or decimals, will serve as well as complex combinations; in fact, they are preferable for this purpose, for they are generally easier to print in an index and (unless they are very long) to remember. The motivation for elaboration is obviously the desire to make the numbers meaningful and not merely arbitrary, and up to a point making them meaningful may also make them easier to remember—and at the same time makes an index particularly efficient, since each reference conveys some further information without the need for explanatory phrases. But as the impulse to elaborate grows, a point can be reached where the numbers are no longer very efficient simply as references, and their primary function becomes the provision of a coded statement. The latter approach is taken for granted by Abbott throughout his discussion. "The importance of reference numbers," he says at the outset, "has increased in recent years, largely because bibliographies that seek to be definitive must now reflect the findings of extensive machine collation." To anyone who thinks that reference numbers exist primarily to facilitate reference, the idea that their importance is affected by machine collation will be incomprehensible. The connection is not greatly clarified by his next sentence: "The numbers are needed to signal the ordering of more complex bodies of work and more complete descriptions of the various impressions and issues of an edition." Complex organization and description do not require complex reference numbers unless those numbers are meant to convey fairly detailed information about the items referred to. And whether the numbers ought to perform that function

is not a question with an obvious answer. Actually Abbott appears less interested in the labeling of entries than in the construction of a symbolic language for referring to books. By providing for the specification of binding variants, for instance, he is going beyond what would be called for in an entry number, since binding is only one of the matters that would be covered in an entry, and since other parts of the entry would deal with features that do not vary, or vary independently of the binding. His system seems headed in the direction of offering a means for denoting specific copies of books. He also explicitly claims another use for his reference numbers: "exact reference in apparatuses such as indexes of textual variants." An example of what he means is the use of index entries to sketch the textual history of a work by arranging the references in chronological order (thus often placing periodicals or anthologies before books by the author) and substituting other marks for commas to separate the figures (an equals sign to mean no revision and a plus to signify revision): "Ba55a=Ba55b=A1+A3.1+A3.2" (p. 73). This suggestion (though it is unlike the previous one in not tending toward the specification of individual copies) again assumes the value of making utilitarian references convey additional meaning: if reference numbers are to encode information about the items referred to, then an index citing such numbers is perforce informative, even without explanatory phrases, and one is naturally tempted to make it still more informative through the addition of a few more symbols.

Obviously one can use any reference number, however bizarre or unwieldy, to locate an item; one just accepts whatever number is provided and makes do with it. For this reason the whole question, as I have said, may be considered trivial. The only thing that makes it worth discussing is the possibility that a complex number may offer little, if any, advantage to compensate for its awkwardness. There is no question that many of the aims proposed as reasons for the elaboration of reference numbers are in themselves worthwhile; the issue is whether there is any point in attempting to achieve those aims by means of the reference numbering. The use of a classification scheme for reference numbers does frequently occur in certain areas, as in the arrangement of books on library shelves, where the motivation is to maintain a meaningful order and therefore to allow for indefinite expansion at any point. For the same reasons subject checklists may sometimes employ numbering systems akin to library classification schemes. In a descriptive bibliography of an author or publisher, on the other hand, or of an area or genre during a particular period, expandability has not generally been a major concern, even though new printings or editions of particular works may occur, and new works may be discovered. The elaborate

systems that have been used in author bibliographies in recent decades do provide for the addition of later editions and impressions to a sequence already begun but do not conveniently allow for the future discovery of editions, impressions, or works that would need to be interpolated between those already recorded. Embedded in these systems, therefore, is the same problem of interpolation that exists with a simple system of single sequential numbers. Letters or decimals can always be added, but if the references already contain letters and decimal points the result will be even more awkward. In any case such additions would decrease the utility of the references as conveyors of information, for numbers or letters would then not necessarily indicate the absolute position of an item in a series: "A9," for instance, would not mean the ninth book or pamphlet if "A6.5" had been inserted between "A6" and "A7." (The same would be true, of course, if numbers were skipped over in the original plan in anticipation of insertions.) It is clear, in other words, that the increased complexity of reference numbers has not produced an improvement in the way in which insertions are handled. Any possible benefits from their complexity must be looked for in another direction.

Those who have devised these complex systems obviously did not intend them to facilitate insertions but were concerned instead with creating references that would convey considerably more information than merely the location of an item in a bibliography. In fact, allowing for insertions, as we have just noticed, would in some respects work against the provision of information; the irony of this situation is that the former is a clear responsibility of reference numbering, whereas the latter is not directly related to the job of reference at all. This confusion of priorities is curious, since these systems were specifically designed for use in numbering entries; but from the results one can see that the designers were really interested in something else. In effect the goal of each (an explicit goal of Abbott's) was to develop a system of notation for referring to particular printings of particular editions of particular works; they then proposed to use this notation for numbering entries, without examining whether it was appropriate for that different purpose. Many of the situations in which bibliographical citations are used involve contexts that make such notation superfluous. A dealer's catalogue entry, for example, would make clear that the item described is a second printing or a third edition even if the appended citation included a reference number encoding that fact; the citation is purely a reference to a standard authority, and a simple arbitrary reference number would be no less informative in such a context. In an index to a bibliography a case can be made for reference numbers that contain

some informative elements, in order to eliminate the need for certain subheadings; but reference numbers of considerable complexity rarely offer additional benefits in an index, for they (however informative) cannot substitute for the entries themselves and are no more efficient in guiding one to the places where full information can be found. Abbott's idea that a notational system can be useful in an index to show textual revision is doubtful for similar reasons. Stringing the notations referring to particular editions together with plus and equals signs cannot possibly convey much about the textual history of a work and can neither substitute for the textual information that should be present in the actual entries nor direct the reader more effectively to those entries.[57]

Numbering for reference purposes, one must conclude, is not likely to be assisted, and may be made more awkward and inconvenient, by the assumption that some kind of work-edition-impression-issue notation can serve to form the reference numbers. But to say this is not to imply that such notation may not have value for other purposes. Formularies for concise notation are always potentially useful, to be drawn on at those times when saying the equivalent in words would be cumbersome and would impede the flow or reduce the clarity of a statement or discussion. However, a system for reducing phrases like "the second issue of the second printing of the third subedition of the first edition of the author's sixth book" to symbolic form cannot be expected to play as central a role in description as the formulary for signature collation does. In setting forth a collation formula one is making a substantive statement that would nearly always be less clear if expressed in words, and one can use symbols that remain constant in their signification from book to book. The places where one may at first be tempted to use a work-edition-impression notation, in contrast, will often prove to be places where the substantive content of the notation is not necessary (and any kind of reference number will do as well) or places where words seem preferable in the context; and while the general system can remain the same from book to book, the effective meaning of the symbols will vary. In bibliographical discourse there is frequent need to refer to a particular leaf or page—as a physical object—in terms of its place in the structure of a book, quite apart from any concern with what is printed

on it; but there is relatively little occasion to cite, for instance, the third edition of an author's second book as such, without intending that the reference be translated into a particular title, publisher, and date.

Nevertheless, I would not wish to deny that a standard notation for this purpose may be desirable. My only concern is that symbols not be needlessly multiplied. Some of the more elaborate possibilities derived from these proposed systems, like the one cited earlier tending toward the specification of an individual copy, lead one to anticipate an eventual row of symbols beginning with a Library of Congress classification number (and perhaps an International Standard Book Number), followed in turn by a number of the Abbott or West type (including symbols for states of sheets and binding variants), a signature collation, a "fingerprint" (as proposed by John Jolliffe),[58] a National Union Catalog symbol for the library holding the copy, and the call number or shelf mark of the copy. What purpose might be served by this long sequence I am not sure, though I am willing to believe there is one. But the piling up of symbols can, I think, be extended beyond the point of usefulness. A potential danger is the blurring of the distinction between reference to individual copies and reference to a scholarly description of an entire edition, impression, or issue[59] —which is to say, the distinction between an entry in a catalogue and an entry in a bibliography. Reference numbers attached to entries in a bibliography cannot be references to specific copies but must be references to accounts based on the examination of a number of copies. This central distinction can be maintained, and maintained efficiently, by a two-part plan operating within each entry: (1) a record of copies examined is placed at the end of the entry, each copy (sequentially numbered) identified by library symbol and call number or accession number, followed by an indication of defects or peculiarities that result from the post-publication history of the copy (and therefore not encompassed by the concept of ideal copy);[60] (2) any variation, among copies on which an entry is based, that results from the printing and publishing process (and is therefore encompassed within ideal copy)—such as states of a given leaf or sheet—is documented parenthetically

by the citation of the numbers assigned to the copies containing the variation in question. Whether or not one wishes to distinguish in form the references to entries from the references to copies (as in "copy 5" of "A9b"), the distinction is basic and must not be lost sight of.

We now return, by what may seem a circuitous route (but is, I believe, a necessary one), to the matter of arrangement. Entry numbering can be expected to reflect arrangement, but some kinds of numbering, when adopted as a standard and rigidly applied, can also determine arrangement. For example, I have heard people refer in a general sense to "A items" and "B items"—by which they mean books wholly written by an author and books contributed to in some way by the author. This usage, roughly reflecting a widespread (but by no means universal) bibliographical practice, in turn influences other bibliographers to set up their bibliographies in the same fashion. Although bibliographers should certainly survey the practices of their predecessors, they should not be content to have imposed on them a system of arrangement (or a system of numbering implying an arrangement) that does not emerge from the circumstances of the particular bibliography they are working on. There is not, in fact, any great uniformity in the definition of the B section, so that general allusions to "B items" are not very precise. If one uses the term simply to mean "books contributed to by the author," then one is referring to items sometimes included in the A section (e.g., books edited by the author) and others at times separated into C or D or E sections (e.g., if books containing original contributions, collected pieces, appearances, and letters are segregated). Presumably one might deplore this lack of uniformity if one could adduce an overwhelming reason for the necessity of a term like "B items." But it serves no useful purpose—any more than the presence of the "B" in dealers' citations tells one something not already stated in their descriptions. Certainly there can be no reason for such a usage that is worth the price of having to submit to a single scheme of arrangement prescribed in advance. In speaking about a particular bibliography, one might wish to refer to "B items" or "F items"; the terms would be clear in the context. But when such terms are used to refer to bibliographies in general, their function is shifted from the descriptive to the prescriptive, making reference numbering an organizing force rather than a reflection of organization.

It follows that bibliographers should feel no less bound to use "A" for a section dealing with books by an author than to use "B" for a section recording some or all of the other books with which the author was associated. One should first plan the arrangement of the bibliography and then see what numbering system would be most appropriate in

that particular instance. It could well be that one would decide to use "A" for the author's first published book, "B" for the second, and so on, with appended numbers for editions. The description of the various editions of books wholly by an author often occupies as much as two-thirds or three-fourths of the bulk of the bibliography;[61] prefixing an "A" to all the entry numbers in such a long section, as is commonly done, would seem to lengthen those numbers needlessly. Instead, one might assign letters sequentially to the succession of books, thereby saving one character in each entry number, and switch to the use of letters for whole categories of works at the point where the record of contributions to books begins. It is true that labeling the author's books "A" through "S" and then using "T" for all the books contributed to, for instance, means that letters are not used in parallel fashion throughout. But I see no disadvantage resulting from that fact. One can always use sectional divisions, like "Part I," "Part II," and so on, to mark the large units of a bibliography, without wasting spaces in the entry numbers for this purpose. Although the entry numbers throughout a bibliography must form a single sequence if they are efficiently to serve the function of reference, there is no reason why any given element in those numbers need have the same "meaning" throughout. It is only because bibliographers sometimes place the classification function ahead of the purely reference function in their numbering that they may think otherwise. But there is nothing illogical or impractical about employing different forms of numbering in different sections of a bibliography where different levels of detail are recorded;[62] one could still, if one wished, use in any particular section of a bibliography a system emphasizing classification, in which each element must consistently signify the same thing.

Devising a system of numbering that conveys some information is not necessarily a bad thing, for, as I have noted, such numbers can be useful in an index. As long as one resists the temptation to overload the numbers with informative elements—to the point where one loses sight of the primary purpose of the numbers and makes them less efficient for reference—a system that builds some meaning into the numbers can in many cases, through its mnemonic value, actually facilitate the job of reference. How much detail is too much cannot be arbitrarily stated. Let me suggest one approach that might seem appropriate in

some situations: information down to the level of impression could be expressed in three characters by using capital letters to stand for works, arabic numbers for editions, and arabic numbers following decimal points for impressions, as in "C1.2." Eliminating the traditional letter standing for the whole section of the author's own books effects a significant economy. As a result, one may feel freer to insert one more element, representing subeditions, which should be taken into account if impressions are to be specified. Subeditions could be indicated by lower case letters following the edition numbers, as in "C1b.2," clearly reflecting the fact that a subedition is a part of an edition and has its own sequence of impressions.[63] I would in most cases be reluctant to encode further information, and even these four characters would no doubt seem excessive for some bibliographies.[64] My suggestion of this scheme is in the spirit of what I have said earlier, and therefore I am not in any sense proposing it as a standard. I intend it only as an example of what one might come up with if one recognizes the folly of feeling bound to any single system previously employed and understands the distinction between reference and classification. It would obviously be inappropriate, without modification, for an author who published more than twenty or so books, since there might not be enough letters to cover those books and any additional categories that would have to be included; nor would it probably be appropriate if books contributed to by the author were treated in the same detail as books wholly by the author, since both categories should then receive the same kind of reference numbering, and again the number of letters might be insufficient. One solution in these instances, rather than the awkward doubling of letters, might be a reversal in the use of letters and numbers, letting numbers stand for books and letters for editions.[65] That solution illustrates how the reference numbering grows out of the particular situation; but I am less concerned here with the solution in a specific instance than with the process of thinking involved. One must first decide on the arrangement of the material; then one can consider whether the entries thus arranged are better referred to by a straightforward series of consecutive numbers or by a sequence of more complicated numbers that reflect in some degree the classification scheme already adopted; and if the latter approach (which is by no means an obvious choice) is selected, one can then think about the system that will best serve the function of reference and at the same time reveal some (how much is a further question) of the information conveyed by the arrangement.

The numbering of entries is clearly dependent on the arrangement of the material. And the arrangement of the material emerges from the process of thinking about the material. The question of arranging material in a bibliography is inextricably connected to the whole process of historical research and writing, to the problem of finding order in the raw material of history. The way in which results are presented is central to the meaning one has found in the historical record. I have ventured some opinions here about such matters as the ordering of impressions and issues, the degressive and segregated treatment of anthologies and periodicals, and the numbering of the resulting entries, but I regard my comments less as solutions than as examples of thinking. Writing a bibliography is not a mechanical task, and the urge to follow mindlessly a set of instructions will produce no better results in bibliography than anywhere else. The quality of a bibliography, like any other intellectual product, depends on the quality of thought that has gone into it. The arrangement of the material is a basic reflection of that thought.

Notes