University of Virginia Library

The Description of Non-Letterpress Material in Books
by
G. Thomas Tanselle

One of the principal problems with which descriptive bibliography has to contend arises from the fact that books are made up of smaller units, each with its own separate history. The book as a whole may be regarded as a unit by the bookseller or the book-buyer; but, as with other manufactured products, its component parts are the units at earlier stages of production. Variations can therefore easily occur among copies of what is supposed to be the same "book," as a result of differences in the manufacturing history of the component parts or differences in their placement in the finished product. Bibliographers have given a great deal of attention to this problem in connection with the letterpress sheets in books and have classified the variations that may occur into those that produce "issues" and those that produce "states."[1] Mixtures of sheets from different impressions or of gatherings with canceled and uncanceled leaves are difficulties that bibliographers are accustomed to thinking about in a formal way.

Much less theoretical discussion has been devoted to the similar problems that are likely to occur when engravings, lithographs, and other materials produced separately from the letterpress sheets are intended to be bound with those sheets to form completed books. It is perhaps natural that variations among copies of the plates used to illustrate an edition of a work, or variations in the placing of those plates in different copies of the edition, have not generally been taken up in conjunction with variations in letterpress sheets. Inserted plates do seem to be further removed from the sheets than, say, a cancel leaf (or a whole cancel gathering) that was likewise printed separately from the rest of the sheets and inserted. One reason, of course, is that they normally result from a fundamentally different process of reproduction. When illustrations and

text are produced by the same process—as when wood blocks are used with type (both relief processes)—there is no necessity for printing the illustrations separately and inserting them. But when, for example, engravings are to be combined with a text printed from type, the illustrations and text must be produced separately, since one results from an intaglio process and the other from relief. They would not only be printed on different kinds of presses but perhaps even in different printing shops and on different grades of paper. Furthermore, inserted plates, if their content is visual rather than verbal, represent not merely a different medium of reproduction from the text but a different medium of expression as well. They also often have a life independent of the text they are combined with in a particular edition. However, despite the natural tendency to think about inserted plates as a separate question from the complexities associated with the letterpress, it is important in a bibliographical context to see the plates as but another example of a larger problem: the potential for variation produced in books by the joining together of discrete elements, whether they are gatherings, cancels, plates, or publishers' casings. Plates may indeed have an independent existence and be treated by art historians as separate entities. But when they are included in books as part of the publisher's conception of what the books are to consist of,[2] bibliographical descriptions of those books must take them, and any variations associated with them, into account.

In some books, of course, it is hardly appropriate to call the non-letterpress portions "insertions," for they may constitute the bulk of the books. A volume showing an artist's work in the form of engravings, or an atlas of engraved maps, or a tune-book of engraved music, for instance, might have a letterpress preface. Such books are likely to be discussed and described by art historians, cartographic historians, and musicologists, who bring to the task the traditions and vocabularies of their own fields, often derived from thinking about the engravings as separate pieces. That their language is sometimes different from that of bibliographers accustomed to dealing with printed books may be an inconvenience but is perhaps not a serious matter, since usually one can quickly learn to make the necessary adjustments. What is more important is that terminology reflects an underlying approach to the material; bibliographers and cartographers, for instance, may find themselves talking

at cross purposes, even when they think they know how the other group uses certain words, because they may not share a single conceptual framework. There has in recent years been some discussion among both cartographers and musicologists about the classification and description of books in their fields, and there has been a somewhat longer tradition of concern with bibliographical description on the part of those concerned with illustrated botanical and zoological books. Illustrated books in general have always received a certain amount of attention, but I think it is fair to say, judging from booktrade activity and library exhibitions, that interest in them has never been higher than it is at present. The moment seems propitious, then, for encouraging interdisciplinary discussion of the problems of bibliographical description that are common to all books with plates. A consolidated approach will produce sounder advances and be of greater benefit to all who deal with books than a situation in which each field considers the problem independently. In this spirit, I should like to discuss what seem to me the two major questions: the first, more theoretical and conceptual, is what relation individual plates, with their own states and variations, have to the variant issues of the books in which they appear; the second, more methodological, is what considerations are involved in devising a system for recording the non-letterpress elements of a book in a bibliographical description. Much of what is said will have had its origin in thinking about books that join engravings and letterpress; but the principles involved ought to be applicable to any books that bring together materials produced by different graphic processes.

I

In 1974 two works appeared that attempt to deal with the problems of bibliographical classification in cartography and musicology, and they can provide an instructive introduction not only to recent thinking in those fields but also to the nature of the difficulties such an undertaking entails. One is Coolie Verner's paper on "Carto-Bibliographical Description: The Analysis of Variants in Maps Printed from Copperplates," in the inaugural number of the American Cartographer (pp. 77-87); the other is the substantial volume that D. W. Krummel compiled for the International Association of Music Libraries, the Guide for Dating Early Published Music: A Manual of Bibliographical Practices. Both deserve credit for being pioneer efforts and for focusing attention within their fields on important but neglected questions; both also have short-comings, as their authors would no doubt agree, and a consideration of certain of those flaws can, I hope, serve a constructive purpose. I find

that I must be rather harsh on Verner's piece; but it seems to have achieved considerable acceptance in its field—I have heard it referred to as "the McKerrow of carto-bibliography"—and a careful examination of it is therefore in order.[3]

Verner, after explaining the kinds of events that occur in the history of individual plates of maps, concludes, "Terms used in descriptive bibliography [i.e., of letterpress books] are not applicable to cartobibliography because the precise conditions described by a term in the one are not found in the other" (p. 84). This statement, on the face of it, raises a number of problems. If indeed the "precise conditions" are so different, then of course different terms are desirable. But that observation hardly gets us very far. It is true that some terms used in the description of books have been employed with different meanings in the description of maps, and there is of course an awkwardness in such a situation. But to conclude that the same terminology will not fit both kinds of material is not a very productive solution: one of the goals of bibliographical description, as of other kinds of historical research, is to find connections, or patterns, or organizing principles, in the mass of surviving evidence. Even though letterpress is a relief process (or, in its more extended sense, a planographic one as well) and engraving an intaglio process, both involve presses of some sort and the transfer of inked images. It would seem profitable, therefore, to proceed from a premise of underlying similarity rather than to emphasize differences of detail.[4]

For example, in his brief discussion of issue, Verner cites definitions by McKerrow and Bowers and then concludes, "In view of the precise meanings attached to the term issue in descriptive bibliography and its inappropriateness for carto-bibliography, this term should not be used in describing maps" (p. 86). Why it is inappropriate for cartobibliography is not analyzed; instead there is the flat statement that the "very specialized application" of the term to letterpress material has "no corollary in carto-bibliography." Although there has been some difference of opinion about exactly how issue should be employed in the description of books, the various definitions share a common theme: an issue is a publishing unit within an edition or a printing, and variant features that provide evidence of publishing or marketing decisions (such as a cancel title leaf with a different publisher's imprint or an inserted series designation) are determinants of issue.[5] In the case of an individual map, the substitution of one publisher's or bookseller's name for another by means of a paste-over cancel would create a different issue; and even if situations producing issues are not common in connection with the publication of separate maps, the possibility of such situations should surely be recognized in any theoretical framework for classification. Furthermore, when several engraved maps are fastened together as a unit, they become in effect a book, and the concept of issue is as relevant to such books as to letterpress books. Issue, of course, refers to whole entities as they are made available for sale, not to particular constituent parts of them, and engraved maps inserted into letterpress books may become elements in one or another issue of those books; indeed, their presence could conceivably be the factor determining issue, if copies of the letterpress sheets were also published without the maps. Although variants in individual maps are not necessarily involved here, the point is that the concept of issue cannot be regarded as irrelevant to the study of the way maps are published.

The logic of Verner's discussion of edition is equally questionable, but his comments do serve to illustrate still more of the problems involved. Verner says that the definition of edition by McKerrow and

Bowers, when applied to maps, would mean "all the impressions printed from a single plate" (p. 86).[6] Although "impressions" is used here in a different sense from its usual one in bibliography (here it means individual copies), the statement seems reasonable, since the engraved plate is the counterpart of the setting of type in letterpress printing. But Verner proceeds to say, "In this use of the term it is synonymous with the carto-bibliographical term plate and therefore redundant." After another sentence, he adds, "although the term edition is not functional in carto-bibliographical description and should be abandoned, it will continue to be used with reference to a given group of impressions [copies] of a map and must be defined precisely. An edition, then, as Skelton specifies, consists of all of the impressions of a map printed from any given state of a plate." A term that is "not functional in carto-bibliographical description" is thus recognized to have a function in describing "a given group of impressions of a map." One may well ask why the latter activity is not part of carto-bibliographical description; perhaps "description" here really means "cataloguing," although much of the discussion deals with relationships among copies and not with single copies in isolation. That difficulty, however, is only one of many raised by this passage.

Another is the idea that edition is synonymous with "plate" and "therefore redundant." It is true that "plate" is often used to refer to individual finished products rather than to the copperplate itself: we speak of a letterpress book illustrated with "plates." In this sense all the "plates"—that is, all the copies of an engraving—produced from a single copperplate would be analogous to a letterpress edition (all the copies from a given setting of type).[7] But this use of "plate" is not what Verner has in mind. "In carto-bibliography," he says, "the term plate applies specifically to the wood block, copperplate, or lithographic stone used to print a given impression of a map" (p. 85). In that case it is difficult to see how "plate" renders the term edition redundant, for the former refers to a printing surface and the latter refers collectively to all the copies printed from a particular printing surface. Edition in letterpress terminology does not, after all, refer to type formes but to all the sheets of paper that have been printed from a given set of such formes.[8] Precision

requires that the distinction between printing surfaces and printed copies be maintained in terminology. If "plate" were used for both purposes, the context would sometimes make clear what was meant, but in many cases confusion would be likely to result.

Perhaps Verner intends to make a similar point when he admits that edition should be used "with reference to a given group of impressions of a map." But the definition he approves for this purpose creates further difficulties: "all of the impressions [copies] of a map printed from any given state of a plate."[9] The result of this definition would be that a single copperplate could result in more than one edition. Since edition has been more widely used for books than for engravings and since the meaning of the term in that connection is well established, it would seem to be foolish, when applying the term to engraved material, to support a usage that is not parallel to the usage for books. There is surely a convenience in having a term to refer to everything produced from a single plate, stone, setting of type, and so on, and edition would seem to be the clear-cut choice for this term. Equating an edition of an engraving with a single state of a plate (whatever "state" means) can only invite misunderstanding.

What Verner means by "state" is of course relevant to understanding his comments on edition. He believes that there is "essentially more agreement between descriptive bibliography and carto-bibliography with respect to the meaning attached to the term state than to any other terms used conjointly" (p. 85). In carto-bibliography, he says, "state identifies and designates a particular period in the life of the plate; during another state the plate will differ from the preceding in some particulars." Any difference, in other words, produces a new state. Those accustomed to talking about letterpress material would be more inclined to think of state as referring to the printed pieces produced. This point is not of great moment, since normally a different state of type-impressions on a piece of paper results from a different state of the type.[10] (There is a more substantial point involved, however: description, whether of books or of maps, has as its primary focus the finished products, which in any case are generally the only evidence one has. When a copperplate or a stereotype plate has survived, it can be described, but such description is subordinate to the main business of bibliographical or carto-bibliographical

description—to describe and classify the printed or engraved items themselves.) What would be more disturbing to the letterpress bibliographer about this broad conception of state is that it ignores the distinction between the kinds of differences that reflect different publishing intentions, intended to be noticed by the public (differences that create issues, in other words), and those that result from accident or the attempt to correct errors. One may at first think this distinction to be more relevant for books, since state refers to a particular element in a book, whereas issue refers to the book as a whole. (A textual variant can produce two states of given page; but a cancel title leaf with a new publisher's name creates not a second issue of that leaf but a second issue of the whole book.) When one is dealing with a single map (or a letterpress broadside, for that matter), where the whole is made up of one piece of paper, not of physically separate units brought together, one may feel that the issue-state distinction is unnecessary. But even if the distinction is less important for practical purposes in such cases, it remains conceptually significant. And when maps are joined together into books, the possibility of having different states of particular maps and different issues of the whole is just as real as with letterpress books:[11] carto-bibliography cannot be limited to individual maps but must take into account the presence of maps in books, both in atlases and as inserts in letterpress books.[12]

In letterpress bibliography, issue and state are subordinate to printing, when an edition consists of more than one printing. For Verner this point is unnecessary, because he dismisses the concept of printing as irrelevant to the study of engraved maps. A printing has traditionally been defined as all the copies printed from a setting of type at one time— that is, in a single press run. Verner asserts, "This is not a pertinent characteristic in the study of early printed maps, since it is impossible to identify with certainty what copies were pulled at the same time so long as the plate was not altered by plate changes" (p. 85).[13] (One might

add that even if changes occurred, they would not necessarily coincide with the intervals between printings.) It may well be that distinguishing copies of a letterpress book printed at one time is often easier than determining which copies of an engraving were produced in a single run. Sometimes, for instance, a title-page date, a notice on the verso of the title page, or some other readily recognizable feature identifies a printing. But printings of books are by no means always easily identifiable, and frequently a considerable amount of bibliographical analysis of the physical evidence, along with the investigation of external sources (such as publishers' archives), is necessary before one is in a position to speak with any authority about what printings a given edition may consist of. To say that it is "impossible" to identify printings of engravings is surely an overstatement; to say that it is generally difficult and sometimes impossible would be more reasonable. But to claim that something is not "pertinent" because it is difficult (or even impossible) would seem to represent wishful, rather than logical, thinking. Bibliographical description, whether of books or of maps, seeks to establish historical facts. The fact, for instance, that a particular copperplate was used to produce a group of copies in March of one year and another group two years later is worth putting on record, when it can be ascertained. That facts are not easy to unearth does not mean that the effort should be abandoned. Verner provides no reason for believing that the concept of printing— the designation of copies produced in a single run—is irrelevant to the historical study of engravings.

There is a difficulty of terminology that must be faced in connection with printing. Descriptive bibliographers of books have traditionally regarded impression as a synonym for printing, and publishers as well as bibliographers use impression to refer to a group of copies, not to individual copies. However, scholars of cartography and of art history, who regularly deal with engravings, have customarily employed impression to signify each copy made from a copperplate. One must admit that this usage has as much logic as the other—indeed, it probably has more. Each copy of a book is an impression from the type, just as each copy of an engraving is an impression from the copperplate. The only problem is one of convention: impression is not conventionally used to mean "copy" by bibliographers of letterpress. This difference of usage is awkward, but nothing more: it does not reflect any underlying differences

of process in the two areas so drastic as to make a common vocabulary impossible. Both usages of impression, however, are so well established that it would be futile to try to eliminate one of them. Instead, one can avoid ambiguity either by using "copy" and "printing" rather than impression or by incorporating an explanatory phrase into any discussion where impression appears, if the context would not by itself make the meaning clear. Having these two meanings for impression is not ideal, but neither is it a matter for serious concern. It is a relatively superficial problem, on an entirely different level from the effort to think through logically the conceptual relationship between printing from engraved copperplates and printing from type.

That Verner's attempt to do the latter leaves something to be desired is further illustrated by his definitions of four terms that he says "are applicable only to printed maps" (p. 86). It is worth looking at these definitions, for their deficiencies are instructive, particularly in regard to the question whether description is to be focused on the copperplate or on the copies made from it. The first of these terms is "change": "Any alteration effected to a plate which prints on successive impressions pulled after the change is made constitute a plate change and create a new state of the plate."[14] This statement perfectly illustrates the problem through its inclusion, as a restrictive clause, of the words "which prints on successive impressions pulled after the change is made." By shifting here from the plate to the impression, the sentence succeeds only in making the nonsensical assertion that plate changes are changes that show up in printed copies made from the plate. Obviously one could change a plate and never print from it, but it would be no less a change for that. The fact of a change in the plate is not tied to whether the plate is actually printed from, though of course normally the only evidence one has for a plate change is a surviving impression made from that plate. One can legitimately say that two states of the impressions made from a plate, if they are of a certain kind,[15] are evidence of a change in the plate; but one cannot turn the point around and say that changes in the plate are changes for which evidence exists in printed impressions. The illogic of Verner's statement derives from a wavering as to whether he is talking about plates or impressions from them.[16] If his focus had been consistently

on description of the printed artifacts, with subordinate references when appropriate to the history of the plates as inferred from the artifacts, there would be fewer problems with his work.[17]

His definition of "variant copy" suffers in the same way. "The term variant copy," he says, "is used to identify an impression which can be distinguished from another one pulled from the same plate and state, where the differentiating characteristics occur on more than one such impression and do not result from alterations to the plate itself as in the case of watermarks in the paper." Because his emphasis in defining state and "change" is on the plate, he feels the need for a term to refer to other changes that affect the finished product. But if he had kept that product at the center, he could have recognized more easily that differences in the plate and differences in the paper both produce differences in the resulting artifact and that it is logical (if perhaps unconventional) to regard both as resulting in different states of the finished prints. Indeed, either could result in different issues as well as states: the use of two different grades of paper, for example, might signify two issues just as much as might certain kinds of changes made in the plate. Because the use of state to mean "state of the plate" is so well established among those who write about engravings, one will no doubt have to speak of "state of the print" (or some such term) when one is referring to the printed piece of paper. But logic dictates that state and issue, as terms for the classification of artifacts, must encompass all kinds of differences present in the artifacts at the time of their release to the public. This approach does not of course result in any blurring of the distinction between differences in the plate and differences in the printing, for the features determining issues and states would obviously be recorded in any description; what it does is provide a logical framework for showing relationships among printed copies, taking into account the paper and the inked impressions that together make up those copies. The term "variant copy," as Verner defines it, thus seems unnecessary; as a technical term it is unwise because it is conceived too broadly to make the distinction between issue and state. Not the least curious aspect of his

definition is its stipulation that "the differentiating characteristics occur on more than one such impression." If only one surviving copy is on paper with a particular watermark, one cannot be sure that others did not once exist; and even if only one ever existed, the use of that paper is still a historical fact.

The two remaining terms, "ectype" and "piracy," raise a somewhat different problem, but one of equal importance from a theoretical point of view. "The term ectype," Verner announces, "is used to identify a particular plate of a map which is very similar to that which it replicates but differs from the original only with respect to bibliographical details." He then adds, "This term has no counterpart in descriptive bibliography." The reason it has no counterpart is not that there is no parallel situation but that, as defined here, the concept is confused. If a plate is "very similar" to another one, it is nevertheless a different plate—just as a plate of a map representing a totally different geographical area would be a different plate. What links the two "very similar" plates together is their content—or what in connection with letterpress books might be called the "work" embodied in the physical object. It is conventional to distinguish between a "work" (an arrangement of words and punctuation) and a "book" (a concrete object that embodies one representation of a work). When a verbal work is set in type a second time, the resulting edition may be just as different from the first in typography, paper, and binding as an edition of some other work; it is related to the first only because the arrangement of type-impressions in it is meant to be a representation of the same work. In the case of maps, one would have to define what the counterpart of "work" would be. It would obviously not be simply a question of geographical region, scale, projection, and so on, but of individual attempts at such delineation. Defining what is meant by the "same" work or deciding when a version becomes "another" work is a problem relevant to all fields. Is is a problem, however, that relates to intellectual content rather than to physical form.

Bibliographers of books have not needed a word like "ectype" because their basic terminology relates to the physical aspects of books, and different settings of type are different editions regardless of the content of what is set; when they need to discuss several editions because those editions contain the same work (and they do, of course, have to decide what texts can be regarded as versions of the same work), it seems sufficient to refer to the first edition, second edition, and so on, of that work. The trouble with a term like "ectype" is that it joins considerations of content and of form: it refers not only to a separate edition (a distinct plate) but one that at the same time is related in content to another one.

Such a term, if a real need for it as a shorthand expression could be demonstrated and if it were precisely defined, could perhaps be condoned—though I wonder if it would not always inhibit clarity. In any case "ectype" is defined too vaguely to be of use: what is meant by saying that one map is "very similar" to another? Earlier in his paper Verner is more detailed on this matter but does not solve the problem. He explains that an ectype "will have an identical geographical and cartographical form with its prototype but differ from it in bibliographical details"—as opposed to a "derivative" map, which "may differ from a prototype in every way except the geographical information conveyed" (p. 79). This statement, in turn, is explained by adding that "an ectype will present the same geographical area at the same scale as the original it replicates, while a derivative may differ from its prototype with respect to the total area depicted, the scale, or the relationship among the bits of geographical data contained on the map." It would seem, in other words, that an ectype is an edition (and perhaps a version as well) of a work, whereas a derivative is an edition of a wholly different work that happens to draw on a particular earlier work as a source. Bibliographers will obviously need to make such a distinction; but among editions of the same work it is still not clear what advantage there is in having a term to refer to one that "may reproduce another with considerable exactitude but differ from its prototype in such detail that there is no possible chance to confuse the two." Whether a bibliographer judges a chance for confusion to exist seems an eccentric criterion of classification.

This point can be illustrated by considering Verner's fourth exclusively carto-bibliographical term, "piracy": "The term piracy applied to a particular plate of a map which is so like the original which it replicates that it is apt to be mistaken for that original but which was issued by a publisher other than the one who published the original map" (p. 86). If "piracy' 'is supposed to indicate an unauthorized edition, as the word would naturally suggest, the definition does not make the point clear. For is it not possible that an authorized publisher could use a new plate so close to the original that it is "apt to be mistaken" for it? Can an ectype—a plate "very similar" to another but clearly distinguishable from it—never be pirated? The terms are not mutually exclusive, for they mix up considerations of similarity of content with circumstances of publication. It would seem to be more sensible for the basic terms of bibliographical classification to refer strictly to physical evidence; these terms can then be modified, as the occasion warrants, with adjectives that express various other considerations, reflecting publishing circumstances or characterizing content, as in "English edition,"

"pirated edition," "revised edition," and the like. Verner assumes that traditional descriptive bibliography does not operate in this way: the term "piracy," he claims, "has no equivalent in descriptive bibliography where a pirated edition would be identified merely as another edition." A pirated edition would certainly be another edition, but that fact does not prevent one from calling it a "piracy" as well, or a "pirated edition," and those terms are in fact used and understood by bibliographers of letterpress books. The essential point is that "piracy" describes the nature of the publication history and the publisher's motivation, and it thus refers to a different set of considerations from those that result in the designation of "edition."

Verner in effect recognizes this point when he says that a piracy "should be designated as one of the plates of the original with the added designation of Piracy, e.g., PLATE III, Piracy" (p. 79)—which is parallel to such a phrase as "third edition, pirated." The third act of engraving of a particular map (a "work"), like the third act of typesetting of a verbal work, produces the third plate or third edition of the work, whether or not it is "pirated." But these clear lines are blurred by Verner's definition of a piracy as a "copy of a map that is so like its prototype that the two might be confused." One of course hopes that a bibliography, by setting the facts straight, will provide the means for distinguishing among items that might be confused; but the bibliographer's assessment of potential confusion cannot be the basis of bibliographical classification or the determinant of a piracy. This curious emphasis recurs in other statements: "The inclusion of such piracies in the description of an original map also eliminates the possibility that the copy might be considered to be a different map, thus acquiring validity as an historical document which it does not actually possess." The acquiring or losing of "validity" does not affect classification, nor therefore the arrangement of entries implied by the classification; if a particular piracy is found to be the third engraving of a work, then it naturally falls into the position thus dictated, whether its stature is enhanced or lowered by that placement. In any case, a piracy, simply because it may have been intended to be taken for something other than what it is, does not thereby lose "validity as an historical document." It exists; it played a role in the publication history of the work; and it is therefore a historical document. The fact that it is not the first publication of the work does not deprive it of validity as a document—though of course its significance will vary depending on the purposes for which one turns to it as documentary evidence.

Verner's paper obviously serves a useful function in focusing attention on some central issues in the description of engraved material, but

it is clearly unsatisfactory in many of the statements it makes about those issues. In regard to the central question of the relationship between carto-bibliographical description and the traditional bibliographical description of books, a more thoughtful approach had been expressed some thirty years before by Lloyd A. Brown. One paragraph of his Notes on the Care & Cataloguing of Old Maps (1941) is worth quoting in full: The precise definitions that Brown proceeds to offer, though they are generally sensible, are not those regarded as standard today[18] —a fact that is not surprising, because he was writing before Fredson Bowers's Principles of Bibliographical Description (1949).[19] But the approach Brown was using is one that we should not lose sight of, for he recognized that a difference of terminology between fields is in itself a superficial matter and that one should not be hindered by the traditions of individual fields from seeing similarities of processes and concepts underlying the different terms. Verner's contention that the terminology of letterpress bibliography is "not applicable" to carto-bibliography because the conditions being described are too dissimilar is not supported by the arguments he advances. Instead, an examination of his paper makes it all the more apparent that the concepts referred to by edition, printing, issue, and state in letterpress bibliographical classification are relevant to material printed from intaglio plates as well. The usage of certain words, such as "impression," is different in the two fields, but that kind of difference is not really a serious obstacle to thought. Letterpress bibliographers and carto-bibliographers cannot avoid coming to terms with the concepts that are common to both fields, for they often find themselves dealing with the same physical entities—books that include, or are entirely composed of, engraved maps. The crucial difficulty is not what differences there may be between relief and intaglio printing but the shift in thinking required to move from dealing with single-sheet entities to treating composite entities (made up of more than one engraving, or more than one letterpress sheet, or engravings mixed with letterpress sheets). If bibliographers are to meet the challenges posed by their materials, they must not stress the differences in these situations but search for the larger framework that includes them both.

We can pursue this line of thinking further by turning to the other 1974 work mentioned earlier, D. W. Krummel's Guide for Dating Early Published Music. In the section entitled "Differences between Copies" (pp. 30-48), Krummel[20] takes up questions of bibliographical classification, and from the outset we know we are in the presence of a person who has given sustained and methodical thought to these matters. A carefully balanced note at the beginning reports that some music bibliographers doubt whether "the terminology of general bibliography, growing out of experience with literary texts, mostly printed from movable type, should be used for musical editions, mostly printed from engraved plates"; but it concludes unequivocally: "The present study," Krummel says, "endorses the use in music bibliography of the terms of the general bibliographer."[21] The argument leading to that conclusion

is enlightened: "advantages are to be gained from drawing on the experience and the systematic thinking of several generations of general bibliographers, concerned with essentially the same problems." Recognition that the problems are "essentially the same" is an important step in itself. The comprehensiveness that it implies for bibliographical concepts is reflected in the definition of edition that immediately follows: "the whole number of copies printed from substantially the same printing surfaces (be they settings of type, engraved plates, or lithographic stones), at any time or times."[22] This essentially standard definition is made to include not only music but any engraved material—and, beyond that, all printed material, whether the printing surface is relief, intaglio, or planographic. The approach places engraved music—the immediate subject—in the larger conceptual framework of which it is logically a part. It recognizes that the nature of the printing surface does not alter the need for a term to refer collectively to all the exemplars representing a given act of drawing, engraving, typesetting, and so on. If this term can be identical in all fields, the chances for confusion are undoubtedly lessened; but establishing identical terms is less important than recognizing the identity of the concept to which the terms refer. The objection to Verner's term "plate," therefore, has more substance than simply the fact that it is a different word: it leads one away from the recognition of identity of concept because the word seems to direct attention to the printing surface itself, not to the copies made from it, and is awkward as a collective noun.[23] The importance of Krummel's integrative approach cannot be overemphasized; it is clearly the direction in which bibliographical thinking must proceed.

After defining edition, Krummel goes on to take up issue, state, and impression, in that order. His discussions are thoughtful, as one would expect, and (with one significant exception, to which I shall return) they reflect the standard definitions as commonly understood by letterpress bibliographers at the time Krummel was writing. His comments on issue (pp. 31-32), for example, are "quoted and paraphrased from

Bowers" and form a fair, if highly compressed, statement of Bowers's views as represented in the 1949 Principles. They therefore emphasize "re-issue" (as opposed to simultaneous issue, the possibility of which is not suggested) and the role of the title page ("The title page is critical in determining the issue"). Since the time of Krummel's book, however, I have argued for an extension of Bowers's essentially sound definition, in order to accommodate a wider range of manifestations of publishers' intentions for creating discrete publishing units—particularly those that result from practices employed in the period of publishers' bindings.[24] It is not necessary to repeat the argument in detail here. I shall simply say that publishers can and do create issues (that is, consciously planned publishing units, intended to be recognized as such) by means that do not necessarily involve title pages—as when they insert half-titles with a new series name or use a different binding (casing) meant to identify that series. And the concept of simultaneous issue (recognized, but not extensively discussed, by Bowers) need not depend on typographical alterations, for a group of copies on large or special paper—even without any other changes—would constitute a distinct publishing unit, planned along with the regular copies and perhaps published simultaneously with them: the large-paper copies would be an issue, but not a reissue. Krummel's basic definition of issue as "a consciously planned unit" (quoting Bowers) is certainly satisfactory, as far as it goes; but it is then interpreted with undue narrowness, citing reissues and title-page alterations. One example of the emphasis on title pages: "when a number of small editions are brought together and re-issued with one collective title page (as with an opera score assembled from the separate editions of its individual selections), the item is a re-issue because of the collective title page." But it could also be an issue without a collective title page, if other means were used to segregate it as a publishing unit—a publisher's binding, for instance (as opposed to a custom binding, which would probably signify only an individual binding-up of these particular selections, rather than an issue of them by the publisher).[25] The emphasis on reissues is further evident in the comments on state, where it is asserted that "By definition, a new issue is re-issued" (p. 34).

Krummel's discussion of state raises a few additional questions. He equates state with "variant" and defines it as "any form of musical publication which exhibits variations in content caused by purposeful alteration

of the printing surface" (p. 32). I have similarly argued, since Krummel wrote, that this concept should be broader: it should cover, in my view, any differences that the publisher does not regard as indicating a discrete publishing effort (and therefore that the publisher does not wish to call to the attention of the public). Such differences need not be in "content" or "of the printing surface"; they can be any kind of difference, whether in the printed image or the paper or the binding. Nor do they need to be "purposeful"; they can be accidents, as well as corrections of errors or attempts at improvement. Since bibliographical description aims at describing the forms of publications as they were published, it must recognize variations among copies that are inadvertent as well as those that are intended; in either case, they are differences and must be recognized. Krummel properly notes that publishers were "under no obligation to mention the changes, and in fact usually found it inconvenient and even undesirable to do so." It follows from this essential characteristic of a state that states (as opposed to issues) refer to particular constituent parts of books (a gathering, an inserted plate, the casing) rather than to whole copies of books. In this respect, Krummel's phrase "any form of musical publication," in his definition of state, is perhaps unwise, because by referring to the whole "publication" he does not clarify as much as he might the distinction between issue and state. Since for him a "new state contains a different text" (p. 34),[26] it is natural that he recommends to editors "in search of an authentic text" that they "must study all copies which make up an authentic edition to locate variant states" (p. 32). And of course this advice is correct. But his suggestion that states are of more consequence than issues to editors (with the reverse being the case for cataloguers) is perhaps stated uncautiously: distinguishing among issues, he says, "may prove quite irrelevant to the editor, since the musical text itself may be completely unchanged."[27] An issue may or may not contain textual variants (just as a state may be determined by textual or nontextual elements), but the question must be pursued in any case; the absence of textual variants does not make the existence of issues or states irrelevant to an editor, for an editor's preparation must include as full a knowledge as possible of the production and publication history of the work being edited.

The fourth standard term that Krummel takes up is impression,

which he appropriately defines as "all of the copies as a unit run off the press at one particular time (i.e., as one 'press run')." What is striking in his treatment of impression is not the definition itself, but its placement in fourth position, following edition, issue, and state. Although he does not say explicitly in his section on impression that he regards this concept as a subordinate classification to state, the point becomes clear when he cites an example in which the "third state exists in two impressions" and when he asserts, "It is axiomatic that two copies from the same impression display no differences at all in text" (p. 36).[28] Later, in a paragraph on the "hierarchic" relationship of the four terms (p. 39), he does become explicit, providing a diagram for an example showing an edition first divided into two issues, with the first issue then divided into three states, and the first state in turn comprising two impressions.[29] This conception of impression is the major departure Krummel makes from Bowers, and it represents the principal way that readers of Krummel's discussion are likely to be led astray. One may imagine how it came about: in Bowers's treatment of the period before the eighteenth century, the sequence of terms is indeed edition, issue, and state, because in that period type was almost never kept standing, and as a result each new printing of a work was likely to be a new typesetting, a new edition; in those few cases where a new impression from standing type did occur, Bowers "most reluctantly" (see Principles, p. 394) suggested the term "issue (re-impression)." Despite this special usage, Bowers makes clear in his detailed discussion of the classification of nineteenth- and twentieth-century books that issue and state "indicate alterations which separate the sheets of some copies of a single impression from others of that impression according to certain variations" (p. 395). The practice of bibliographers, almost without exception, is consistent with this view, regarding issue and state as subordinate to impression. The standard definitions of the four basic terms do not logically admit any other hierarchy: an issue, for instance, could not consist of several impressions, for a group of copies forming a discrete publishing unit could not comprehend the output of several different press runs over a period of time. Although impression as a term is less likely to be needed for books before the nineteenth century, the concept of course applies for all periods: in the earlier periods an edition generally consists of a single impression, but the identity of edition and impression in those cases does not mean that the concepts have altered. Krummel's book is intended to deal principally with music between 1700 and 1860, and thus Bowers's discussions in the later part of the Principles, as well as those in the earlier, are relevant. A concept of impression that apparently derives from Bowers's idea of "issue (re-impression)" for early books is not appropriate to the classification of books from the last two centuries.[30]

Krummel proceeds to the difficult question of the integrity of copies. Although he deals with it briefly and from the point of view of cataloguers (likely the main component of his audience) rather than bibliographers, he does recognize the logical necessity of moving on to this matter. First he seeks to define copy and rather puzzlingly offers two very different definitions, adding that the second "would appear on general principles to be the better one to follow" (p. 46). It certainly is the preferable one: it defines a copy as comprising "whatever was known or likely to have been offered for sale in the earliest copy . . . including wrappers, catalogues, and the like."[31] The earlier definition takes copy to exclude wrappers, catalogues, and "other material superfluous to the content and the publishing event." But one is dealing with published items, and if a catalogue or a publisher's wrapper or casing was part of the entity as published, then it is part of "the publishing event," whether or not it is "superfluous to the content." However, the question of determining what in fact was a part of a copy as published is a central one in connection with books not in publishers' casings, especially when those books contain both letterpress and engraved material or consist of items also made available separately. As an illustration of the view that often "the precise entity is impossible to determine," Krummel cites the libretto of Don Giovanni, which "may, or may not, be seen as

an integral part of the Breitkopf & Härtel full score (1801)." Perhaps it would be clearer to say that for books not published in bindings or casings, or not now surviving in publishers' bindings or casings, one cannot always determine which of various separable units now bound together were placed on sale originally as combined entities. It may well be in many instances (as perhaps in the Mozart example) that the units were available both singly and together; in such situations, it would not be historically accurate to think of a single "precise entity," for there would have been three or more entities. In any event, "items which are composite in their texts," as Krummel calls them, are always a problem, even when the material is all letterpress, and they are likely to pose a particularly prominent problem when engraved illustrations, maps, and music are involved. As common musical examples, Krummel mentions collected volumes of series (the title pages of which may not reflect, in date or other respects, the series as a whole), volumes containing supplements (where the presence of the supplement may be a sign of a new issue or may merely result from an owner's binding together two independent items), anthologies in which the individual pieces are printed from plates previously used in other combinations in other volumes (and perhaps carrying individual pagination), and books that draw on "the same basic 'repertory' of plates," each of them resulting from "individualized assembly for a particular purchaser, much like certain early engraved atlases" (p. 47). These all raise interesting bibliographical questions, which cannot sensibly be speculated about on the basis of a single copy—which is all that a cataloguer, by definition, is called upon to examine. Krummel's comments, then, do not explore such books in any depth ("the busy cataloguer should be satisfied with the title page, although the editor and bibliographer will want to look further"); but it is noteworthy that he calls attention to the problem, which has not been much discussed and which is central to the bibliographical treatment of books combining materials printed from different kinds of surfaces.

The contrast between Krummel and Verner could hardly be more striking. Krummel's discussion is clearly the more reasoned and thoughtful; and, even if it requires some modification here and there, it demonstrates a sound and fruitful basic approach. But the process of examining the arguments of Verner, in addition to those of Krummel, has served, I hope, to show what the principal issues are. Underlying everything else are the opposite positions these two writers take on the question whether the terms—and therefore the concepts—of classification in letterpress bibliography are applicable to non-letterpress material. The answer to this question, as I trust the present discussion has shown, cannot really

be in doubt: the concepts that have been given the labels edition, printing (impression), issue, and state in letterpress bibliography are relevant to all printed matter, regardless of the nature of the printing surface.[32] Some of the terms themselves have not always been used in the same ways in different fields, but one should not be misled by variant usages into thinking that the concepts are different. No doubt certain terms—notably impression—will continue to be employed in more than one sense, but this situation should not be regarded as a serious problem, for the context should make the meaning clear.[33] What is of far more significance is that the concepts represented by the basic terms of letterpress bibliography be thoroughly understood by bibliographers of all kinds of material. The importance of this understanding is not simply practical—though of course there is a great practical advantage involved, since materials printed from different kinds of surfaces are often brought together in publication and must therefore be treated in a single bibliography. Rather, the importance of making these connections lies in the fact that the activities of printing and publishing are conceptually similar whatever the printing surface, and the bibliographical classification of printed matter will be misleading as historical scholarship if it does not reflect that fact. To have reached this position—as Krummel has—is actually to have taken a considerable step forward, and everything else one may say about the description of non-letterpress material rests on this basis.

Another central point is that bibliographical description in all fields must be concerned primarily with the printed items themselves, not with the printing surfaces that produced them. The bibliographer can infer certain facts about the plates, types, or stones from the finished product; but that product is the primary evidence, capable of being examined directly by the bibliographer, and is the object of the description. That the bibliographer's focus must be on the published entity is particularly obvious when one considers publications that consist of more than one unit—several letterpress sheets, or letterpress sheets joined with engraved music, maps, or illustrations. In these cases, although one must try to do justice to the component parts, one must also account fully for the composite entity, which is a fact of publishing history. Some parts of a book may have had a separate existence, as independent

entities, outside the book; but their role within the book is nevertheless an aspect of their history, and descriptive bibliographers, whether of books or of prints, have to be prepared to deal with such items both as separate entities and as parts of books. If one thoroughly understands the concepts behind edition, printing, issue, and state, one should have no difficulty in explaining clearly and precisely the relationship between such parts of a book and the book as a whole, without in any way slighting or obscuring the history of those parts as separate items. Knowledge of the printing surfaces and processes is of course important, but what one finally is describing are not copperplates or formes of composed type but the results of an act of making public the sheets printed from them, often in combination.

The classification of a published entity, therefore, cannot be governed by the classifications that may have applied to certain of its parts that were made available to the public separately. For example, since issue and state refer to variations within printings, the insertion of a particular issue of an engraving in all copies of a given printing of letterpress sheets does not result in any differences among copies. The fact that the map, say, was also available separately—and, as a separate, was a second issue—could appropriately be pointed out in the description of the inserted plates in the volume; but the volume as a whole would simply be a first printing, one of the characteristics of which is the inclusion of a map that—as far as its own separate history is concerned— is a second issue. If some copies of the first printing of the book contained the second issue of the map and others contained the first issue, the difference would have to be taken into account in the description of the book, since there would then be a variation distinguishing some copies from others within the same printing; but the result would not necessarily be two issues of the book, simply because two issues of the separate map were involved. There would be two issues only if the use of the second issue of the map became the occasion for a separate publishing effort (indicated, among other possible ways, by a cancel title, perhaps referring to the particular issue of the map included). Otherwise there would be a single issue (that is, a single printing) of the book, in which the map appears in variant states, indentifiable as the first and second issues of the separate map. Whatever created the second issue of the separate map (perhaps a different imprint) does not create an issue of the whole book unless the publisher calls attention to it as characterizing a discrete group of copies. Similarly, if copies of a single issue of the separate map are included in all copies of both the first two printings of the volume, the result is simply two printings of the volume; the presence in them of the same issue of the map does not make all copies

of the volume a single issue, because the letterpress sheets, containing the title leaf and other preliminaries, represent two different stages in the printing (and publishing) history of the volume.[34]

One should not conclude from these hypothetical examples that letterpress sheets take precedence over non-letterpress material in bibliographical classification. So long as we are talking about non-letterpress "insertions" into books that are largely made up of letterpress sheets, there seems nothing odd in allowing the letterpress material to play a determining role; but many books exist in which the engraved plates bulk far larger than the letterpress sheets (books of engraved music with letterpress prefaces, for instance, or suites of illustrations with brief introductory texts in letterpress). In any case, the relative quantity of the letterpress and the non-letterpress material is not the only factor to consider; equally important is whatever element—most commonly the title page—serves to define the book as an entity. Thus if a cancel letterpress title leaf, with a different publisher's imprint, is inserted into the sole letterpress gathering at the beginning of a volume consisting otherwise entirely of engraved music, a new issue is created, even though the engraved material is dominant in the volume and is invariant. If instead to accomplish the same purpose the entire letterpress sheet (gathering) were reset but the engraved music were still from the original printing, the result would still be an issue—not a new edition, even though the preliminary gathering would be a new edition, since the body of the book would still be of the original printing. Both the bulk and the title page play a role: that the reset material is such a small fraction of the volume (whether the rest is engraved is actually irrelevant) means that a new edition has not been created, even though the reset part includes the title page; and the existence of the new publisher's imprint in the

reset part means that the reset gathering cannot be regarded simply as a variant state of the preliminaries but must be considered a sign of a new issue of the book as a whole. Many volumes, of course, contain engraved as well as letterpress title leaves, and the engraved ones, by virtue of being title leaves, form a separate category from other plates that may be present. The evidence they provide must be evaluated in conjunction with that from the letterpress title pages; if the latter are more often dominant as indicators of issue, the engraved titles can also on occasion provide the evidence, and no generalization can be made that would relieve the bibliographer of the responsibility of weighing the evidence from both.

Nor is there any point here in outlining further examples or kinds of situations: the number of possible permutations is large, and enough has been said to suggest how they can be dealt with by a consistent application of the principles underlying the basic concepts of bibliographical classification. Any system for handling non-letterpress insertions in letterpress books must be applicable as well to the more general problem of composite volumes containing items (whether or not printed from a different kind of surface) that have had an independent publication history. Judgment is involved in applying terms like edition and issue to these volumes, and bibliographers will not always agree in their assessment of complicated situations. But if they understand the basic concepts of classification and recognize the double level of classification required when independent entities are incorporated into larger entities, they are sure to clarify the publication histories of the volumes they describe. The object, after all, is to explain a course of events as clearly as possible to the reader, and one is free to use phrases and sentences in addition to the basic terms. Attempting to assign terms like edition, issue, and state should not be an exercise in applying arbitrary definitions but should be a process that allows one more clearly to see, and express, relationships. The same kinds of relationships exist among the elements of composite volumes in all fields; and asking all bibliographers—regardless of the subject matter they specialize in—to work within this one framework of classification is not to impose restrictions on them but to assist them in making their labors as productive as possible.

II

After one has thought about these questions of classification, the next step is to consider what pieces of information regarding non-letterpress material ought to be recorded in a descriptive bibliography and where

and how they should be incorporated into an individual description. In some respects these matters take care of themselves, if one has a thorough grasp of bibliographical classification and is familiar with the standard approaches to the description of other elements of books, such as type and paper (since those methods can easily be adapted to the description of non-letterpress material). There is some value, nevertheless, in taking up the implications of certain decisions about the form and content of the description of non-letterpress material, especially since such material has not been given very extensive coverage in discussions of bibliographical description. Indeed, Greg said in 1932 that book illustration is "a subject whose appeal or relevance lies primarily in the realm of graphic art and not of bibliography" and that engravings, "separately printed and stuck into otherwise already complete books, show the minimum of relevance to their surroundings."[35] These remarks, it must be recognized, actually occur in a passage arguing that bibliographers should give attention to illustrations: even though such work may contribute "mainly to the history of art," it is also "work essential to booklore, work that can be justified on the most rigid bibliographical grounds." Yet the textual orientation of Greg's approach to description is evident: although the connection of some illustrations with books "may be of the most superficial and fortuitous character," others (especially in "books of a technical character") "may be said to form an integral part of the text," and in any case illustrations may provide evidence relevant to "the history of the transmission of the text." The central point called for by Greg's discussion—but a point that does not emerge clearly from it—is that the descriptive bibliographer's task is to describe books as published entities and that illustrations, when included in books, are naturally part of what the bibliographer must describe; the degree of connection between the illustrations and the text is irrelevant. Greg comes close to making this point when he says that illustration "claims the attention of bibliographers merely on the ground that these products of art happen to be found in books"; but his inclusion of "merely" in the sentence is revealing. Some people will be less—and others more—interested in the art than in the text, but both are parts of the physical object to be described. The books that Greg was principally concerned with did not require him to consider the problem of inserted plates in much detail, and little is said about them in his comprehensive (and major) statement on descriptive bibliography occupying most of the fourth volume (1959) of A Bibliography of the English Printed Drama to the Restoration.[36]

J. D. Cowley, in Bibliographical Description and Cataloguing (1939), provides a brief section on "Plates and Insets" (pp. 105-106), which begins rather shakily but recovers in the second of these sentences: "Typographically, I suppose, books may be held to be complete without any plates that they may be intended to have when they are published, because the plates are printed separately and are not included in the signed gatherings. But since in bibliography we are describing books as they were published, we must account for the plates in the technical note." His view is that they should not be included in the formulaic signature collation but should be entered in the form of a brief register (e.g., "plates, Front., I-VI") following the pagination statement, with a description "in a separate paragraph dealing with illustrations or in the analysis of the contents." Although he does not recommend giving the location of plates (normally, at least before the nineteenth century, "it is practically impossible to determine the exact position which any particular plate or inset was intended to occupy"), he suggests, somewhat inconsistently, that inserted maps, tables, and the like, not part of a regular sequence of plates, be specified by location ("plates [6]; inset tables following pp. 56, 72, 104").[37]

Ten years later Fredson Bowers, in Principles of Bibliographical Description (1949), also devoted very brief space to engraved title pages and plates (see especially pp. 287-289 and 446), but what he says is characteristically incisive and logical and provides a sound basis to build on. His recommendations are essentially the same as Cowley's, except that he stipulates the recording of the positions of plates—as in "plates [3] (opp. sigs. C4v, F1, G4v)"—using signature or page references to the facing pages.[38] But he provides a more thoroughly reasoned statement of the desirability of limiting the signature collation to material that went through the printing press and of recording non-letterpress insertions separately. There is no question that signature collations, which can

become complicated enough when there are letterpress insertions to account for, are made less easily usable by the inclusion of other kinds of insertions as well.[39] Bowers has argued this position so effectively (pp. 198-201) that it should now be regarded as conclusively established. The place for a register of plate numbers (or an indication of the number of plates, if unnumbered) is—he (like Cowley) believes—at the end of the register of pagination; a description of the plates can occupy a separate later paragraph, although Bowers would make an exception for frontispieces and engraved title leaves and would include them in the contents paragraph (p. 289). Bowers's treatment of engraved material recognizes its importance: he never implies that it is less deserving of the bibliographer's attention than the letterpress. The brevity of his discussion of this subject—the absence of the wealth of detail that he provides for other parts of the description—reflects his experience (and the tradition of bibliography out of which his work grew), which largely involves books more noteworthy for their verbal content than for their illustrations and in fact containing relatively few illustrations.

The most thorough contributions to the question of the bibliographical study of inserted plates have emerged, not surprisingly, from the examination of botanical books. Unlike atlases and music books, in which engraved leaves often form a solid block of material without intervening letterpress leaves, botanical books consist of engravings inserted into letterpress gatherings often enough to make the problem of their relationship a pressing one; and unlike literary books, in which illustrations are often regarded as serving only a decorative function, botanical books normally contain illustrations that are essential parts of what the works as a whole are intended to convey. Gordon Dunthorne is generally recognized as a pioneer in the bibliographical treatment of plates in his Flower and Fruit Prints of the 18th and Early 19th Centuries (1938). He was writing before Cowley and Bowers, of course, and his work now seems rather rudimentary, although many more recent bibliographies have not been as detailed in their handling of plates. What he did recognize was the value of noting plate marks, imprints, and watermarks, and he attempted to work out a concise method for recording them (using abbreviations such as TC for "top center" to indicate positions on the plate). A more significant landmark is Allan Stevenson's work in the second volume (1961) of the Catalogue of Botanical Books in the Collection of Rachel McMasters Miller Hunt. The remarkable introduction to that volume, which in explaining the methods employed becomes a 65-page exposition of the fundamentals of descriptive bibliography, offers

the most extended discussion of the formal description of plates yet available (pp. clxvi—clxxvi). It agrees with Bowers in excluding plates from the signature collation but extends the point by objecting to tacking a register of plate numbers on to the end of the pagination statement. Stevenson argues instead that there should be a separate collation of plates (parallel to the collation of printed sheets), accompanied by descriptions of the plates and detailed plate lists (which would be the counterpart to the contents paragraph for the letterpress material).[40] He devotes his longest discussion to the plate lists, which are essentially lists of the botanical names for the plants depicted: his procedure is to quote the polynomials or binomials from the plates and then to append in brackets any corrections or modern reductions, sometimes citing authorities and intermediate forms. This part of Stevenson's work clearly required extensive research,[41] and it reflects the importance of the plates to botanists as evidence in questions concerning priority of nomenclature. Although the specific problems would be different in other fields, the kind of attention given here to the content of the plates is exemplary.[42] More recently the field has been given yet another advance in an excellent paper read before the April 1975 meeting of the Society for the Bibliography of Natural History by Gavin D. R. Bridson.[43] The point of view underlying the paper is summed up in Bridson's statement that "the aim should be to record and describe plates with the same degree of precision already accorded to letterpress aspects of the book" (pp. 471-472); he quotes some of Bowers's remarks on the need to go beyond description to analysis and then calls for "the same analytical approach" to be applied to plates (p. 473). Bridson praises Stevenson's work but surpasses Stevenson in the demands he makes on bibliographers; his paper is in effect a supplement to Stevenson, taking up the aspects of plates that Stevenson enumerates for description and outlining further complications connected with them. He convincingly shows why bibliographers need to have a thorough understanding of graphic processes (more than just the ability to identify the various media) and how a detailed description, as with letterpress, serves to facilitate further discovery. Anyone undertaking to describe a book with plates should give thought to Bridson's remarks.

The principal issues that emerge from these discussions can best be considered in two categories: those that have to do with the actual description of the non-letterpress material and those concerned with how to incorporate those descriptive details into the formal description of the entire book. The former are the more significant, because they are substantive, but the latter are not without substantive implications. Certain decisions about the substance of a description, of course, may obviate certain questions of form: if, for instance, the treatment of plates is reduced to the simple statement that there are so many engravings after paintings by a particular artist and so many after paintings by another, the question whether the information about plates should be presented all at once or distributed among the paragraphs of the description does not arise. And such brief statements may be regarded as appropriate on certain occasions. As with the other elements in a description, various levels of detail may be adopted, depending on the nature of the material and the primary purpose that the bibliography or catalogue is intended to serve.[44] The parts should normally be kept in proportion to one another, and if the letterpress is given abbreviated treatment, the plates should probably receive similar treatment. If the plates are the principal interest of the volumes and if the audience for the bibliography consists largely of art historians, there is naturally good reason to go into some detail about the plates; but in that case one should usually provide more information about the letterpress as well, recognizing that evidence from the one, as part of a combined object, may be relevant to interpreting evidence from the other. There is no doubt about the value to historical scholarship of bibliographies that describe rather than merely identify, and wherever possible descriptions should be conceived on a scale that will allow the treatment of plates to include the kinds of details to be enumerated below. In any case the cardinal rule, exemplified in Bridson's paper, is that descriptive bibliography cannot logically slight any element that goes to make up the finished book, and plates must be dealt with as fully as letterpress.

Stevenson designates four groups of "the main identifying and descriptive details" to be recorded in his plate descriptions: "1) designation of type of plate and range of subjects; 2) titles, artist signatures, imprints; 3) sizes of typical plates, in inches and tenths; 4) position of

plates" (p. clxxi). We may postpone consideration of the positions of plates, for that is a separate matter: it has to do with the physical relationship of the plates to the letterpress, whereas the other items relate strictly to the plates themselves (and would apply as much to the plates as separate items as they do to the plates as parts of books). Bowers's account of what is called for in describing plates in nineteenth- and twentieth-century books is similar: in addition to transcribing the lettering, "The analysis should state the method by which the plate was made and printed, and the color of the ink, together with the medium of the original. When the letterpress [or the other transcribed lettering] does not make the subject of the illustration clear, a brief account may be given"—"but," he adds, "it is impossible to describe a plate in sufficient detail to provide a basis for the detection of unknown 'states'" (p. 446). It is not always possible to do this for letterpress either: in both cases one cannot know whether the details chosen for inclusion will be those that vary in unexamined copies. But one will have looked at multiple copies of the books to be described, and one can report on variations discovered in them, both in the letterpress and in the plates. Bibliographers should of course have photocopies of the plates with them when they examine copies of the books at different locations (and ideally should compare copies on the Hinman Collator); in their published bibliographies they should always cite references to published reproductions and indeed should include as many reproductions as possible. Their aim should be—for plates as well as for letterpress—to provide those details that their specialized knowledge (developed from their examination of numerous copies) tells them will be most useful both as a historical record and as a means of uncovering further variations.[45] Many of the details listed by Stevenson and Bowers (e.g., transcribing titles and imprints, providing measurements) are similar to ones that would be given for letterpress material; and the general procedure should be to furnish for non-letterpress material the same categories of information provided for letterpress material, or corresponding categories when the match is not exact. To the details mentioned above, therefore, should be added some comment on the typography or lettering and on the paper. Any satisfying approach to the description of non-letterpress material must be amenable to handling not only interspersed plates but whole sections of books—such as multi-plate sequences devoted to musical compositions, each perhaps with its own title page. Such groups of plates, with consecutive "pages" or leaves, resemble in some respects sequences of letterpress text pages, and such symmetry as can exist in the treatment of the two is desirable.

Taking all these considerations into account, I would suggest the following as the categories of information to be provided in a detailed description of non-letterpress material in books. They need not be presented in this order or be separated into these nine units; and although the wording here applies most directly to individual illustrative plates, this outline can, I think, be readily adjusted to accommodate other kinds of non-letterpress material:

Descriptions on this level will not always be regarded as feasible: even Allan Stevenson normally listed plates selectively.[52] But when other aspects of a book are to receive detailed treatment, there is no justification

for failing to accord the non-letterpress elements a correspondingly detailed treatment. The space required, when a book has numerous plates, is considerable, but a detailed contents paragraph for a complicated book, with quasi-facsimile transcriptions included in it, can also be formidable; in neither case, however, can the value of the detailed record, or the logic of including it in a description, really be questioned. And the space involved is not as great as this enumeration perhaps suggests, for in most books a certain number of details will be the same for all, or nearly all, the plates and need not be repeated each time. For that reason Dunthorne and Stevenson, for example, first report various characteristics that apply to all or most of the plates in a volume and then list the titles of the individual plates. But the alternative approach of conceiving of the record as a list of plates with descriptive details contained in each entry does not mean that recurring details have to be set forth time after time. The particular arrangement will vary with the nature of the material; the listing above—let me repeat—is meant only to suggest the details to be covered, not a necessary order. Those dealing with other kinds of non-letterpress material than individual plates may find that several of these items coalesce. Multi-plate pieces of music, for instance, may have their own engraved title pages, transcription of which may automatically take care of items 1, 3, and 4 above. The form the description takes is not as crucial as the recognition that these categories of detail contribute directly to printing and publishing history and therefore to the history of the transmission of information in whatever field the books are concerned with.

In addition to the description of the plates (or other materials) themselves, there is the question of recording their positions in the books in which they occur. Some difference of opinion has existed on this matter. Whereas Bowers takes it as a matter of course that plates and insets have "their positions noted in the description" (p. 288), Stevenson (despite his listing of "position of plates" as one of four items to be included in descriptions) remarks in an often cited (and misused)[53] statement:

This is one of Stevenson's less happy passages, and several points require comment. He seems to conceive of the concept of ideal copy as denoting a standard of correctness that the publisher wished to achieve rather than as a historical concept referring to what the publisher in fact did achieve. Ideal copy is meant to distinguish the form or forms of a book as released to the public from the defective or altered forms in which one may encounter individual copies as a result of what has happened to those copies since publication.[54] The bibliographer, as a historian of printing and publishing, is interested in establishing what forms the publisher released and does not wish to record as part of a description variations that result from accidental or intentional alterations inflicted upon copies during the course of their separate post-publication histories. Thus in the face of a "dizzying" array of variations among copies in the placement of plates, or in their states, one cannot simply throw up one's hands but must instead attempt to ascertain, through an analysis of all available evidence, the nature and causes of the variations. If there is reason to believe that some or all of the variations, however numerous, represent forms of the book as published, then they are encompassed by ideal copy and must be described; any variations produced by the decisions of individual binders or owners would not be recorded. Distinguishing between these situations is frequently a difficult task for books not released in publishers' bindings; in many cases, of course, the plates were simply made available as a group, but sometimes with printed directions to the binder or with indications of position on the plates themselves. In any case, the point that affects bibliographical treatment is not how many permutations and combinations there are but who was responsible for them. Furthermore, Stevenson's remark that he was lucky in being concerned with only one copy of most books is doubtless true but hardly pertinent, since it has no bearing on the question whether he should record the positions of plates. Since he was writing a bibliographical catalogue of a collection, and not a bibliography, his task by definition was to describe the copy or copies of each book in the Hunt collection, rather than to establish the range of forms that constitutes ideal copy for each book. Thus whether or not to report the positions of plates depended entirely on what level of detail he wished to set for his descriptions and had nothing to do with the relation of the Hunt copies to ideal copy in each case. It is perfectly logical to record, in a catalogue of particular copies of books, the position of the plates in those copies, regardless of how the plates got in those positions. Of course, the catalogue will be even more useful if it sets forth whatever evidence is available in those copies bearing on the question of the responsibility for the present arrangement. In any event, whether the subject is a catalogue or a bibliography, Stevenson's view that the position of plates "matters less than many have supposed" rests on illogical arguments.

In fact it matters very much simply because the plates are a part of the book, and one has not described the book without describing their relation to it. But beyond that, since plates are part of the content or "text" of a work, establishment of their positions is a substantive, not merely a formal, matter. Bridson points out, for example, the importance for the study of nomenclatural priority of reconstructing—in the case of books published in fascicles—exactly which plates appeared in each fascicle;[55] and he suggests some techniques of analysis, such as examining stab-holes, offset, staining, and the like, that can help in determining those original arrangements (p. 477). When directions for placing the plates—either as a list or as a statement on each plate—are published with a book, bibliographers have sometimes contented themselves with recording this intended order (which is, of course, important), without further investigating whether the plates originally appeared in a different order, as in fascicle publication. The separate parts of books published serially—whether novels or scientific works with plates—were intended to be bound together, and in their original form many of them are understandably scarce today. But bibliographers must attempt to locate them (not only for the arrangement of their contents but for the information printed on their wrappers) and, whether or not they can be found, must analyze whatever evidence is available in an effort to learn as much as possible about the history of the publication of the books being described. Bibliographers, in other words, cannot stop with a report of what publishers intended but must try to uncover the history of what they actually did. The fascicle order of plates, then, is important in its own right as a fact of publishing history but is also important for the light it may shed on textual and nomenclatural matters.[56] Naturally there may be times when there is insufficient evidence to enable the bibliographer to say anything about the locations of the plates in a volume; but the absence of such information ought to mean that the bibliographer

has attempted to secure it and failed, not that it is excluded as a matter of policy. The bibliographer should feel just as obligated to provide the locations of plates or other non-letterpress items as a signature collation of the letterpress sheets; indeed, the two are complementary, for it takes them both to form a complete record of the physical structure of the book.

The practical question of how to list these locations therefore must be faced. We may begin by accepting the point established by Bowers, Stevenson, and others that it is a mistake to try to incorporate this information within the collation formula for the letterpress sheets. Of course when one is describing a book in which the non-letterpress material falls in a single block and is not interspersed through the letterpress leaves—as in a fascicle with the plates at the end or an atlas with a letterpress title page and introduction—there is no problem anyway, since the letterpress and non-letterpress do not intrude on one another. This kind of situation could easily be handled in some such manner as the following:

When the plates are interspersed the record is more complicated, since the location of each plate must be specified. One could attach the record to the end of the collation and pagination line, as above, or make it a separate line:

• 8°: A 8 B-D8, 32 leaves, pp. [4] I 2-59 60; plates I-V (facing B2v, B6, C4v, C8v, D3).
• Letterpress. 8°: A 8 B-D8, 32 leaves, pp. [4] I 2-59 60.
• Plates. I-V (B2v, B6, C4v, C8v, D3).

The difference between the two is obviously not great, and both are clear and relatively concise (the word "facing" or "opposite" can even be dispensed with). However, logic and convenience may give a slight edge to the second approach, since it may be thought to convey more clearly the fact that the plates do not simply follow the letterpress and that the record of plates is in effect another collation formula, going back over the part of the book covered by the first one and focusing on a class of material excluded from it. In any event, the use of signature reference notation shows directly how the plates fit into the structure of the letterpress gatherings; in the case of later books, where references in general are more commonly made to page numbers (and plates may even be labeled with the numbers of the pages they are to face), one might still wish to keep the signature references as well—as in "I-V; B2v(16), B6(23) . . ."—simply because they provide an easier way of seeing at a glance the structure of the volume as a whole, sheets and plates combined. Signature or page references should be provided again later at the point where each plate is described; otherwise the task of finding a particular plate would involve the annoyance of having to count the location references, a significant inconvenience when the number of plates is large. Of course, the collation of plates could be offered in an expanded form, with each plate number directly attached to a location —as in "I (B2v), II (B6), III (C4v) . . ."—but such an arrangement would take up considerably more space, and more important, it would not as readily show any irregularities in the sequence of numbers on the plates. Indeed, when there are irregularities, it is a good idea to give a bracketed total in any case—as in "I-X IX-XXII [24]."[57]

The inclusion of the plate collation adjacent to the signature collation (clearly the right spot for it) leads to the question where the other details about the non-letterpress material ought to be placed in the description. Conventionally there is a paragraph labeled "Plates" or "Illustrations" that brings together these details. Yet an exception is (rightly) made for an engraved title, which is treated along with the letterpress title page. Thus two principles are at work: the handling of the engraved title and of the plate collation supports the idea that the focus of the description is on the book as an entity and that both letterpress and non-letterpress material will be brought into the description under categories—such as "title page" or "collation"—that refer not to a method of printing but to a characteristic of the completed book; the presence of a "Plates" paragraph, on the other hand, suggests an organization in which the method of production of the material takes precedence over its function in the book. The point is perhaps most clearly illustrated with reference to the description of paper: should the paragraph on plates include a description of the paper used for the plates, or should that description come in the paragraph on paper, which takes up the paper found elsewhere in the book?[58] If one answers that logically the paragraph on paper should include all papers in the book, whether used for letterpress or non-letterpress, one is really taking a stand on the question whether or not a book is fundamentally "complete" when its letterpress is complete. The implied answer is that if the book included plates

upon publication, it cannot in any sense be regarded as a complete book without them. If both letterpress and non-letterpress are required to make up the book, then the paper for both comes under the heading of paper for the volume. In this view, is there any justification for a paragraph on plates? Where would the characteristics of plates listed above, other than paper and perhaps typography and lettering, be covered? Those characteristics—involving transcription of wording and description of content—are actually the counterpart of what is covered for letterpress in the contents paragraph. Perhaps there should be two contents paragraphs, corresponding to the two parts of the collation, the first labeled "Letterpress Contents" and the second "Engraved Contents," or some equivalent phrase. The description of the non-letterpress material would then be parallel to that of the letterpress, with information distributed as appropriate in the sections on title, collation, contents, typography, and paper; the notes section could likewise include additional information about plates, such as the report of their bibliographical status as separately published items. The logic of this arrangement is perhaps most easily seen in connection with music books and atlases, where the engraved portions constitute the principal contents: it would be inappropriate to relegate them to a "Plates" section, reserving "Contents" only for letterpress. But the same point can be made about any combination of letterpress and non-letterpress. This approach requires only modest shifts in the arrangement presently regarded as standard; but I think these changes in wording and placement reflect a sounder view of the published book as a physical entity.

Books containing non-letterpress material present such a multitude of different situations that bibliographers will inevitably have to make their own adjustments in individual cases, and it is not my aim to try to suggest a simple formal arrangement to be used on all occasions, for that aim is unrealistic. Rather, what I am concerned with is the point of view or general approach that underlies bibliographers' treatment of non-letterpress material. Central to the view I have expressed here is the proposition that the same standards and procedures of classification and description are appropriate and desirable for both non-letterpress and letterpress material. This view further implies that both are equally deserving of the bibliographer's full attention: both are part of the contents, physically and also intellectually, of the books in which they appear together. Non-letterpress material may seem more directly relevant to the intellectual content of books in some fields than in others. One may, for instance, think first of natural history books when reading a comment like W. M. Ivins's on pictures "made to convey visual information":

"the story of prints is not, as many people seem to think, that of a minor art form but that of a most powerful method of communication."[59] But the point is applicable to prints in all fields.

Nevertheless, it is true that the exact nature and role of non-letterpress material varies from field to field, and interdisciplinary cooperation is especially important for the growth of bibliographical knowledge in this area. Bibliographers must take the broad view, and see their own particular problems of description in the context of non-letterpress description in general, if they are to benefit as fully as possible from relevant advances already made and if they are to produce bibliographies that meet as fully as possible the needs of users from different fields. Allan Stevenson was well aware of the mixed audience he was serving in the Hunt catalogue, and he has a great deal to say about the use of his catalogue by botanists and students of the graphic arts as well as by collectors and other bibliographers. "Those parts of the description are most successful," he says, "which prove of value to the botanist and these others at one and the same time" (p. cxciii). Or, as he puts it more succinctly, "Bibliography prefers to serve not one but all." In order to do that, all must cooperate, and Stevenson expresses the hope that bibliographers and scientists will "agree to use their special funds of experience and information to the common end of good descriptions." All who deal with the history of books, regardless of the field, do of course have common goals, and all their work is interrelated. The study of non-letterpress material in books provides a particularly forceful illustration of this basic point.

Notes