In a section of his Essays in Critical Dissent entitled "The Philistinism of 'Research,'" the late F. W. Bateson laid down a challenge to bibliographers which, so far as I know, has never been taken up directly. The question he poses is roughly this: if the Mona Lisa is in the Louvre, where are Hamlet and Lycidas? what is the essential physical basis of a literary work of art? ¹ Bateson's answer, a somewhat surprising one given that Hamlet is one of his examples, is that the physical basis is "human articulations"; "the literary original exists physically in a substratum of articulated sound" (pp. 7-8). A book, he claims, has the same sort of imperfect relationship to the original work as a photograph has to the man photographed; it is a "translation" or "reproduction" (p. 7). It follows from this, Bateson argues, that the bibliographer is guilty of mistaking the secondary for the primary: he busies himself preserving the author's "accidentals," when the author's responsibility stops with the sounds; the bibliographer confuses the function of the author with that of his copyist (p. 8). To much of this the bibliographer will have a ready answer, but the importance of these criticisms lies in their level of generality; they call for a justification of certain bibliographical attitudes in terms of aesthetic theory and they raise, in vivid if eccentric fashion, several of the crucial issues in aesthetics today. Without presuming to speak for bibliography, I want to challenge Bateson's conclusions on these issues and to suggest that the physical appearance of books sometimes has even greater importance than textual bibliographers are willing to allow it. I believe that leading writers on aesthetics -- writers quite independent and even ignorant

     The most comprehensive and penetrating discussion so far of the mode of existence of works of art is to be found in Richard Wollheim's Art and Its Objects.² Wollheim's starting point is to consider the very simple hypothesis that works of art are physical objects; but as soon as he begins to do so the general category "works of art" begins to split into two: on the one hand, we find painting, sculpture, and architecture, and, on the other, music, drama, ballet, and, it would seem, literature. Of course, a strong case may be, and has been, urged against identifying even painting, sculpture, and architecture with physical objects, but at least in the case of those arts there are arguments to be made. If I am asked where the physical object which is the Mona Lisa is, I can point to the canvas in the Louvre, and similar possibilities present themselves with Donatello's St. George or Vanbrugh's Blenheim. But if I am asked for Ulysses, or Der Rosenkavalier, or Swan Lake, I am in difficulties. There are physical objects or events I call Ulysses or Der Rosenkavalier -- the book or score on my desk, the reading to the Joyce Society, or the performance by the Keele Amateur Opera Club -- but I cannot simply identify one of these with Ulysses or Rosenkavalier. As Wollheim, from whom I draw this argument, insists, we cannot identify the book on the desk with Ulysses because if we did it would follow that if that book were lost the work itself would be lost. That is obviously unsatisfactory, and, since we often refer to the book on the desk as my "copy" of Ulysses, the next step might reasonably be to look for the physical object Ulysses of which my book is the copy. The obvious candidate is the author's manuscript, or the fair copy which Greg and

     The question of the mode of existence of literary works, therefore, seems to be closely tied up with questions of identity and with differences between the arts. Since much relevant discussion hinges on these differences, it may be useful to list them before moving on to consider accounts of them and their usefulness to the bibliographer. At this stage it is probably wise to leave literature aside and concentrate on the distinction between painting and sculpture on the one hand and music, drama, and ballet on the other. The status of prints, architecture, and literature is often problematic.

	Painting and Sculpture	Music, Drama, and Ballet
(1)	identity unproblematic -- may be identified with some particular thing	identity problematic
(2)	atemporal	temporal
(3)	no distinction required between creator and performer	distinction required between creator and performer
(4)	creator as such produces those things of which the work consists	creator does not produce those things of which the work consists
(5)	work can be forged; identity depends on history of production	work cannot be forged; identity free of history of production
(6)	work is a particular	work is a type with tokens 3

There is nothing to be said at this stage about (1), for it was the starting point of our inquiry. The temporal/atemporal distinction is one which both Bateson and J. O. Urmson think important. Bateson seems to believe that it explains why the literary work is "invisible" (p. 10), but he must be mistaken because ballet is clearly visible and temporal; there

     Professor Wollheim points out in a brilliant short essay, "On an Alleged Inconsistency in Collingwood's Aesthetic," ⁴ that literature shares with music the potentiality for being created in the artist's head. Wollheim considers the following art-activities: writing a (short) poem; composing a tune; painting a picture; making a sculpture (he ignores cast sculpture, as I have earlier). The first two can be done in the artist's head, without materials, but the second two cannot. As he remarks, "a natural way of thinking about making a sculpture in one's head, or, for that matter, painting a picture in one's head, is as imagining making a sculpture or imagining painting a picture: whereas to talk of imagining

(i) that it is a sufficient condition of the making of a work of art-type that one should internally produce (e.g. say to oneself, play to oneself) a token of that type;

(ii) that it is a sufficient condition of a work of art's being a work of art-type that it should be expressible in a notation; ⁵

(iii) that the discrepancy between an art activity and the internalized version of it arises acutely, or more acutely, with work of art-particulars (p. 256).

     The type/token distinction stems from Charles Sanders Peirce in his consideration of graphs and signs:

A common mode of estimating the amount of matter in a MS. or printed book is to count the number of words. There will ordinarily be about twenty the's on a page, and of course they count as twenty words. In another sense of the word "word," however, there is but one word "the" in the English language; and it is impossible that this word should lie visibly on a page or be heard in any voice, for the reason that it is not a Single thing or Single event. It does not exist; it only determines things that do exist. Such a definitely significant Form, I propose to term a Type. A Single event which happens once and whose identity is limited to that one happening or a Single object or thing which is in some single place or any one instant of time, such event or thing being significant only as occurring just when and where it does, such as this or that word on a single line of a single page of a single copy of a book, I will venture to call a Token. (Collected Papers, ed., Charles Hartshorne and Paul Weiss (1933), iv, 423.)

     Wollheim's reference to notation in his essay on Collingwood shows some interplay between his own ideas and those of Nelson Goodman, and it is from Nelson Goodman that he doubtless derives his interest in notation. Goodman's principle of distinction between the arts is different from Wollheim's, but much more sharply separates those which are amenable to notation from the rest. In his Languages of Art he starts from the observation that Rembrandt's Lucretia can be forged, while Haydn's London Symphony cannot, and nor can Gray's Elegy: "Let us speak of a work of art as autographic if and only if the distinction between original and forgery of it is significant; or better, if and only if even the most exact duplication of it does not thereby count as genuine. If a work of art is autographic, we may also call that art autographic. Thus painting is autographic, music nonautographic, or allographic." ⁶

     Goodman considers the possibility that the difference between them is that autographic arts are one-stage, whereas allographic arts are two-stage, but he rejects this on the grounds that literature is allographic and one-stage. The ground of difference he is considering is, of course,

To verify the spelling [Goodman is using the word in a very general sense here] or to spell correctly is all that is required to identify an instance of the work or to produce a new instance. In effect, the fact that a literary work [or a musical one] is in a definite notation, consisting of certain signs or characters that are to be combined by concatenation, provides the means for distinguishing the properties constitutive of the work from all contingent properties -- that is, for fixing the required features and the limits of permissible variation in each (p. 116).

     There is no room within the scope of this essay to examine Goodman's discussion of the requirements of a notational system, even were it within my capacity to do so, but his conclusions merit our attention. A musical score, he argues, is a true score, that is, it is a character in a notational system: the score defines the work but there are no competing definitions as there are in ordinary language; the score, unlike an ordinary definition, is uniquely determined by each member of the class it defines (p. 178). It is because of the notational system that we are able to move from score to performance, or from performance to score, or from score to score without (if we do not make any mistakes) impairing

     The status of painting with regard to notation is relatively uncomplicated, just as, in a quite different way, that of music is; with literature, the position is more complex. Here is Goodman's conclusion: "The text of a poem or novel or biography is a character in a notational scheme. As a phonetic character, with utterances as compliants, it belongs to an approximately notational system. As a character with objects as compliants, it belongs to a discursive language" (p. 207). Goodman's meaning will become clearer if we look at what he says of the drama, which serves in some ways as an intermediate case between music and literature. "In the drama, as in music," he argues, "the work is a compliance-class of performances" (p. 210). Compliance is the inverse of "denotation," by which Goodman means reference in a way which excludes exemplification as a mode of reference. "Complies with" is interchangeable with "is denoted by" and "has as a compliant" is interchangeable with "denotes." A single inscription will denote many things (all the performances of Der Rosenkavalier, for example) and the class of these things constitutes the compliance-class of the inscription (p. 144). Hamlet, then, for Goodman, is the compliance-class of performances of Hamlet. The text of Hamlet is mainly a score; the dialogue is "in a virtually notational system, with utterances as its compliants." You get from text to performance and back again without losing the identity of the work, just as in music. The stage directions, on the other hand, are only supplementary instructions. It is readily apparent, therefore, that a literary text could either be taken as a score

(1) The text is a score; it denotes utterances; it is a character in a notational system. The work is a compliance-class of performances. The physical "basis" of the art is speech. The text is not an instance of the work.

(2) The text is a script; it denotes "objects;" it is a character in a discursive language. The work cannot be the compliance-class, because the compliance-class is a "set"⁷ of objects, and which objects belong to the class might be indeterminable. The literary work, therefore, must be the text or script itself and its utterances: "All and only inscriptions and utterances of the text are instances of the work" (p. 209).

     Even if all texts of literature had the same status as the dialogue in drama -- that of a character in a notational system -- they would not have the relation to the work of a copy or translation. One text might, of course, be a copy of another, but, provided it was copied correctly, it would be just as valid a score as the original; it would not, like the copy of a painting, be necessarily inferior to the original. Nor would the relation of the score to the work be that of a copy or translation: the text would define the work; the notation distinguishes constituent properties from contingent ones. If Bateson were right and the text of Hamlet were an attempt to copy the sounds in Shakespeare's head, or even those of the first performance, it would clearly be a failure: a mass of information about quality of sound, pitch, and duration would be lost; we would not have Hamlet. The play would, in Wollheim's terms, be a particular, and the work would be lost. It may still be correct to argue that Hamlet is a matter of human articulations; but, following Goodman, we must say that they are articulations compliant with the text. The text is a special definition of the work. The subsequent questions, what is required by compliance and whether there are redundant elements in the notational scheme, are ones of vital interest to bibliographers

     All this, however, is to suppose that texts of literature should be regarded as scores with utterances as their compliants -- the view most favourable to Bateson's case. As we have seen, Goodman decided otherwise, claiming that both utterances and inscriptions are instances of the work. If he is right, a text is as far removed as it possibly could be from being a copy of the work; it is, in so far as anything is, the work itself. This is a matter of some consequence and, at a fundamental level, the dispute over whether a text is a score or an instance of the work may underlie some of the controversy which has taken place in bibliography itself. For example, there has been much debate over whether an editor should ordinarily choose an autograph manuscript as copy-text or whether he would be better off choosing the first edition instead. There is a tendency for those who recommend choosing the manuscript to emphasise the role of the text as score -- the important thing is to avoid error in transcription -- while those who recommend the first edition are thinking of the qualities of the inscription available to the public. Certainly, those authors, Pope is an example, who take most interest in the qualities of the printed book are those who most lead the editor away from the manuscript. The same problems arise in debates about editing diaries, journals, and other unpublished matter, when it is sometimes argued that the precise details of the inscription are important, as they are not for published work.⁹ It may be that these issues will only be settled by closer attention to how particular authors did regard the inscriptions of their work.

     An important challenge to Goodman's views and a defence of the idea that the text is a score rather than an instance of the work has come from Professor Barbara Herrnstein Smith; but the defence is mounted in such a way that it admits many exceptions and provides us with criteria for deciding whether or not the text is a score. I believe that close pursuit of this debate between Professors Goodman and

A silent reading of a poem, however, is or may be a much more specific and precisely determined activity than looking at a picture or listening to music. The reader is required to produce, from his correct "spelling" of a spatial array of marks on a page, a temporally organized and otherwise defined structure of sounds -- or, if you like, pseudo-sounds. The physical or neurophysiological source of the structure generated by the silent reader is of little significance here: it may originate somewhere in his musculature or peripheral or central nervous system, or the source may vary from reader to reader. What is significant is that the structure itself will not vary (p. 6).

     In psychology there have been a number of investigations of what is usually called subvocalization, but may also be called subvocal speech, or inner speech, or implicit speech, or covert oral responses. It is clear that there is such a thing because children, and adults, can sometimes be observed whispering or moving their lips while reading. Additional evidence comes from detailed investigations of adults (students about twenty-two years old) which were carried out in a medical setting by Åke W. Edfeldt in Stockholm. He measured muscle movement through electromyography, inserting needle electrodes directly into the speech musculature. He decided there was activity in the speech musculature during "silent" reading in all ten subjects, but he was cautious of generalising too far on this basis: "it was shown that silent speech probably occurs during all reading. It is, however, not possible to state categorically that such is the case, as some of the good readers showed less activity during reading than during relaxation." ¹¹

     Various other experiments giving similar results are reviewed by Eleanor J. Gibson and Harry Levin in their Psychology of Reading.¹² A different light, however, is thrown on the whole question by J. Baron's experiments, particularly that reported in "Phonemic Stage Not Necessary for Reading" in Quarterly Journal of Experimental Psychology, 25 (1973), 241-246. Baron found that his subjects could recognise whether or not a short phrase made sense just as quickly, regardless of whether the sound made sense. That is, they rejected "my knew car" just as quickly as "my no car," when, if they had been going through a phonemic stage, they should have taken longer over "knew car." However, it may be possible to integrate this finding with Edfeldt's, because it is generally agreed that subvocalization increases with the difficulty of the passages being read. Glenn M. Kleiman in "Speech Recoding in Reading" (Journal of Verbal Learning and Verbal Behaviour, 14 (1975), 323-339) presents a model of reading which he thinks accounts for the various results: printed sentence --> visual encoding --> lexical access --> working memory storage processing --> sentence comprehension.

     He believes that recoding into speech occurs after lexical access (recognition of the word and its meanings) and facilitates temporary storage of words necessary for sentence comprehension. This explains the disparity between Baron's findings (concerned only with lexical

     Although Professor Smith insists in her discussion with Professor Goodman that there is a necessary element of subvocalization or performance in reading poetry, she seems very ready to admit that there may be variations on that pattern. One such variation would be concrete poetry and the range of pattern poems, altar poems, and calligrammes. A poem such as Eugen Gomringer's "Silence" ¹³ makes its point without any vocalization:

silence	silence	silence
silence	silence	silence
silence		silence
silence	silence	silence
silence	silence	silence

Similarly, Professor Smith suggests that whereas lyric poems represent (or imitate, or picture) personal utterances, prose fiction characteristically represents "inscribed discourse," and it follows that a literary work in that genre would be constituted by instances of its own text (p. 8). This suggestion, though it is difficult to reconcile with the position adopted earlier in her essay, is rich in possibilities: the argument now is that whether or not a text is a score depends on what that text imitates: if it imitates the spoken voice, it is a score; if it imitates inscribed discourse, it is not. A glance at the development of prose fiction gives the suggestion credibility: early novels presented themselves as other sorts of written discourse -- letters (Pamela), journals and autobiographies

     In a suggestive and lucid article, Richard Shusterman has argued that the anomalous nature of literature stems from its historical development.¹⁴ Before the invention of the printing press literature was an oral art and manuscripts served as scores; this was the only way an author could reach a public of any size. The printing press, however, gave him direct access to a large number of readers; and we find him addressing the "reader," rather than the listener. Those genres, lyric and epic poetry, for example, which grew up before the period of printing retain much of their oral character, while newer genres, such as the novel, are essentially printed. One might add that there has been cross-fertilization so that some novels thoroughly exploit the potentialities of literature for encoding speech effects, while some poetry makes use of the potentialities of the type-written or printed page. The advantage of Shusterman's and Smith's work is that it points to a period, the beginning of the eighteenth century, when authors first became conscious of the anomalous nature of literature and of the importance of printing. They began to imitate letters, journals, and autobiographies; they noticed the lack of direct contact with an audience and the introduction of powerful intermediaries, the publishers; and they noticed that literature had got caught up in print and paper. It is the world Pope's Dunciad sets out to describe and that the Dunciad Variorum ironically exemplifies.

     When the Twickenham edition of the Dunciad first appeared in 1943, it provoked a review essay of special brilliance from F. R. Leavis. His introductory comments, witty and provocative, are the ones of immediate interest to the bibliographer:

Yes, one concedes grudgingly, overcoming the inevitable revulsion, as one turns the pages of this new edition (The 'Twickenham'), in which the poem trickles thinly through a desert of apparatus, to disappear time and again from sight -- yes, there has to be a Dunciad annotated, garnished and be-prosed

---

SB 037: Page 96

in this way. A very large proportion of the apparatus, after all, comes down from the eighteenth century with the poem, and the whole, though to read it all through will be worth no one's while, is enlightening documentation of the age that produced Pope and of which Pope made poetry. ¹⁵

     In the Dunciad Variorum Pope is playing with the transition of literature from oral to written that Shusterman discusses. He is exploiting the fact, as Sterne with his black page and squiggles does later, that literature is seen as well as heard. And he is perhaps the first writer to exploit significantly the opportunities that affords. Leavis's remarks begin with the appearance of the page. He is such a good critic that he immediately finds a meaning there: twentieth-century scholarship in its overweening way obstructs the relation between poem and reader; our literary culture has its priorities wrong. His mistake lies in believing that the meaning he finds is unintentional; in fact Pope put it there as a reflection on his own culture; it is part of his exploitation of the medium. Pope intended his poem to be hemmed in by scholarship: the work was designed not only to refer to the dangerous plight of literature

     The spirit which haunts the pages of the Dunciad Variorum is that of Richard Bentley. He was not, paradoxically, associated with Variorum editions, and his high standing in classical scholarship in large measure derives from the way his insistence on the importance of the editor's judgement led scholars away from the mechanical compilation of others' interpretations and judgements. Variorum editing was instead closely associated with Dutch scholarship, and E. J. Kenney in The Classical Text describes this period in Holland (1700-1750) as "the age of the Variorum quartos." It was the age of the notorious Havercamp, of whose edition of Lucretius H. A. J. Munro claimed, "there is not one week's genuine work beyond what scissors and paste could do."¹⁷ In calling his work a Variorum Pope must have had such work in mind -- a much wider frame of reference than his critics generally allow him -- and the Dunciad Variorum stands in ironic relation to OED's first recorded occurrence of the term (in Chambers Cyclopaedia of, significantly, 1728): "The Variorums, for the generality, are the best Editions." ¹⁸

     Nevertheless, Bentley's work dominated textual criticism in England in this period and directly affected Pope and his circle through the Epistles of Phalaris controversy and, most important, through Bentley's disciple Theobald.¹⁹ Bentley's most famous work, because it produced most controversy, was his edition of Horace, and from our point of view the second edition, published in Amsterdam in 1713, is especially

     To stop at this point, however, would be to simplify Pope's purposes and to consider the chosen format too superficially. Not all the notes to Pope's poem are ludicrous, and this suggests that he felt it needed explaining and amplifying: we need to know who Gildon is (I, 250), what the relation between Baeotia and Ireland might be (I, 23), the nature of the link between Saturn and lead (I, 26), and why the poet says that Curll's waters in their passage "burn" (II, 178). ²¹ A careful reading of the notes suggests Pope's approval of certain sorts of annotation, and a close look at the format confirms this. An examination of contemporary editions shows that the Dunciad Variorum resembles Bentley's Amsterdam Horace much less than it resembles the Geneva Boileau of 1716, and the similarities between the edition of Pope and the edition of Boileau are too great, I believe, for them to be coincidental. We know that Pope had a copy of the 1716 Boileau from Professor Mack's listing of his library; it was given to him by his friend James Craggs.²² We also find the parallel between Pope and Boileau emphasised in the "Letter to the Publisher" in the Variorum, a "Letter" supposedly written by Cleland but doubtless composed by Pope himself:

Having mention'd BOILEAU, the greatest Poet and most judicious Critic of his age and country, admirable for his talents, and yet perhaps more admirable for his judgment in the proper application of them; I cannot help remarking the resemblance betwixt Him and our Author in Qualities, Fame, and Fortune; in the distinctions shewn to them by their Superiors, in the general esteem of their Equals, and in their extended reputation amongst Foreigners . . . But the resemblance holds in nothing more, than in their being equally abus'd by the ignorant pretenders to Poetry of their times; of which not the least memory will remain but in their own writings, and in the notes made upon them (p. 13).

     The aim of the editor of the Geneva Boileau, Claude Brossette, was to keep the author's work as fresh and clear for posterity as it had been for his contemporaries. Brossette presents himself as no ordinary editor; he is the author's friend, conveying information entrusted to him by the author himself, and thus giving others a correct understanding of his works: "Ce n'est donc pas ici un tissu de conjectures, hazardées par un Commentateur qui devine: c'est le simple récit d'un Historien qui raconte fidellement, & souvent dans les mêmes termes, ce qu'il a apris de la bouche de l'Auteur original" (p. VI). He deliberately distinguishes this sort of edition from those of the classics and criticises such editions in a way which would appeal to Pope: "Au défaut de ces connoissances, les Commentateurs qui sont venus aprés, ont été obligés de se renfermer dans la critique des mots, critique seche, rebutante, peu utile; et quand ils ont tenté d'éclaircir les endroits obscurs, à peine ont-ils pû s'élever au dessus des doutes & des conjectures" (p. VIII).

     The authority of his edition and its notes is expressed in the mode of presentation -- a mode which he remarks on and which is remarked on in turn by the Amsterdam edition of 1718 which departs from it -- he gives his notes the titles Changemens, Remarques, and Imitations, each group serving a special purpose. If we compare the Dunciad Variorum and the Geneva Boileau side by side (Plates 2 and 3), we see the strong set of similarities between them:

(1) The notes are plentiful and are placed at the foot of the page, as those in the Amsterdam Horace are. I believe Pope was the first English poet to be annotated in this way. The notes to Pope's Odyssey are at the foot of the page, but the notes to the Shakespeare are not.

(2) The notes are divided into sections by headings.

(a) Two of the headings, Remarques/Remarks and Imitations, are the same. Pope has no Changes.

(b) The headings are in italic capitals.

(c) Both use two columns with a dividing rule for Remarks; the Variorum also uses them for Imitations.

(3) The styling is similar.

(a) Both print Vers or VERSE in full.

(b) Both use italic and a square bracket to separate the lemma from the annotation. Further quotations in the notes are in italic.

     Pope aimed, therefore, to have his works presented in an edition to equal Boileau's. The Geneva edition has a splendid frontispiece with a portrait of Boileau and an inscription which poses this question:

Boileau sut remplacer Horace,
Seul il sut remplacer et Perse et Juvenal;
Mais de cet auteur sans égal
Qui remplira jamais la place?

     Pope's interest in the format of the Dunciad Variorum and its intimate

     My response to Bateson's challenge is, therefore, to claim that the existence of works of art as print and paper is not less but more important than bibliographers have generally taken it to be. Of course, Bateson's case stems from a confusion of work of art-particulars and work of art-types: just because plays and poems are not objects like paintings and statues, it does not follow that they are not physical, or even that they are not visual. The weakest case for the bibliographer is that literary texts are scores, and that does not mean that they are copies of the work; it means that they define the work -- as Goodman puts it, "the work is a compliance class of performances," performances compliant with the text (p. 210). It is essential, therefore, that all elements of the score are retained from copy to copy. All elements, that is, that are truly part of the score, that are necessarily there. Complicated questions parallel to those of synonymy are raised here which bibliographers may need to consider further. It seems, however, that it is also valid to regard works in the post-Gutenberg era as having inscriptions as their instances; some works are imitations of written works and may exploit the written, or printed, medium. I have tried to indicate how this might have been done in the case of the Dunciad Variorum. My discussion of that work is necessarily incomplete, but I hope I have shown that the presentation of the text is open to exploitation by the author, that it can carry specific associations, and that a richer understanding of the relation of author, book trade, and public may lead to better interpretation of literary works. There is here a whole field of bibliographical investigation, one concerned with the significance of format and habits of reading, and bibliographers are only just beginning to cultivate it. ²⁵

Essays in Critical Dissent (1972), pp. 7-10; first published as "Modern Bibliography and the Literary Artifact" in English Studies Today, 2nd series, ed. G. A. Bonnard (1961), pp. 66-77. Bateson draws on the chapter on "The Mode of Existence of a Literary Work of Art" in René Wellek and Austin Warren, Theory of Literature, 3rd ed. (1949; rpt. 1963).

Art and Its Objects: An Introduction to Aesthetics, 2nd ed. (1968; rpt. 1980). The second edition adds six new essays. Some of the relevant issues are raised (inconclusively, I believe) by Tom Davis and Susan Hamlyn, "What Do We Do When Two Texts Differ? She Stoops to Conquer and Textual Criticism" in Evidence in Literary Scholarship: Essays in Memory of James Marshall Osborn, ed. René Wellek and Alvaro Ribeiro (1979), pp. 263-279. A response (to which this essay gives extended support) came from G. Thomas Tanselle, "Recent Editorial Discussion and the Central Questions of Editing," Studies in Bibliography, 34 (1981), 57 n.67.

The idea of such a list comes from J. O. Urmson's "The Performing Arts" in Contemporary British Philosophy, 4th series, ed. H. D. Lewis (1976), pp. 239-252. Only the first four items in my list are Urmson's and he would dispute the validity of (6). The source of (5) and (6) will be apparent later. Urmson elaborates these ideas in "Literature" in Aesthetics, ed. G. Dickie and R. Sclafani (1977), pp. 334-341.

On Art and the Mind (1973), pp. 250-260; first published in Critical Essays on the Philosophy of R. G. Collingwood, ed. M. Krausz (1972), pp. 68-78.

It should be noted that (i) and (ii) are not necessary conditions; prints are work of art-types and do not fulfil these conditions. Nelson Goodman argues that (ii) is not even a sufficient condition because there could be a notation based on the history of production, "Comments on Wollheim's Paper," Ratio, 20 (1978), 49-51. But a lot hinges on the meaning of Wollheim's "expressible."

Languages of Art: An Approach to a Theory of Symbols (1969), p. 113.

"Set" is my own desperate unphilosophical choice. It is easy to see that "tree" might denote a class of objects in the world, but it is more difficult to see how the whole text denotes in that way.

This is to suggest that there may be a theoretical sense in which certain features of the text are "accidental," as opposed to Greg's practical sense. Tom Davis's acute observations on "accidentals" in "The CEAA and Modern Textual Editing," Library, 5th ser., 32 (1977), 63-74, have received too little notice, though they met with a shrewd critical response from G. Thomas Tanselle, "Recent Editorial Discussion and the Central Questions of Editing," SB, 34 (1981), 36-40. It will be apparent that my frequent proviso, "if we copy exactly," explains the bibliographer's emphasis on the holograph or first edition.

I am aware that these disputes are usually couched in other terms, often concentrating on the author's intentions; but that may not be the best approach. The best guide through these mazes of controversy is G. Thomas Tanselle in "Greg's Theory of Copy-Text and the Editing of American Literature," SB, 28 (1975), 167-229, and "The Editing of Historical Documents," SB, 31 (1978), 1-56, reprinted in Selected Studies in Bibliography (1979), pp. 245-307 and 451-506.

On the Margins of Discourse: The Relation of Literature to Language (1978), p. 5. The essay first appeared as 'Literature as Performance, Fiction, and Art," Journal of Philosophy, 67 (1970), 553-563, and formed part of a symposium on Goodman's Languages of Art. An alternative view, which I shall not explore, is that in reading we are imagining the work performed.

Åke W. Edfeldt, Silent Speech and Silent Reading (1959), p. 151.

The Psychology of Reading (1975). Of particular importance are the sections on subvocalization (pp. 340-350) and models of the reading process in the mature reader (pp. 438-482).

Concrete Poetry: An International Anthology, ed. Stephen Bann (1967), p. 31. My observation, even if it holds true here, would not hold for all the poems in the anthology; some hardly depend on visual effect at all.

Richard Shusterman, "The Anomalous Nature of Literature," British Journal of Aesthetics, 18 (1978), 317-329.

The Common Pursuit (1952), p. 88. The review first appeared in Scrutiny, 12 (1943-44), 74-80. The fifth volume of the Twickenham Pope was edited by James Sutherland (1943).

The Correspondence of Alexander Pope, ed. George Sherburn (1956), ii, 503 (28 June 1728). An account of the arrangements for publication can be found in Sutherland's introduction to the Twickenham volume; it is supplemented by my studies, Pope's Printer, John Wright (Oxford Bibliographical Society Occasional Publication No. 11, 1976) and "Lawton Gilliver: Pope's Bookseller," SB, 32 (1979), 101-124.

E. J. Kenney, The Classical Text: Aspects of Editing in the Age of the Printed Book (1974), pp. 115 and 156. Munro's discussion of Havercamp is in his edition of De Rerum Natura (1886), i, 19.

As David Foxon has suggested to me, variorum probably started as a trade shorthand for cum notis variorum. The notebooks of Thomas Bennet show him writing "Cornelius Nepos variorum," "Boethius Variorum" and the like as early as 1686 (The Notebook of Thomas Bennet and Henry Clements, ed. Cyprian Blagden and Norma Hodgson (Oxford Bibliographical Society Publication, n.s. VI, 1956), p. 27).

The best study of the scholarship and controversies of the time is still R. F. Jones, Lewis Theobald (1919).

John Nichols, Illustrations of the Literary History of the Eighteenth Century (1817), ii, 621. Quoted in Lewis Theobald, p. 173. The 1713 Amsterdam Horace was the second edition: Q. Horatius Flaccus, ex Recensione & cum Notis atque Emendatibus Ricardi Bentleii, Amstelaedami, Apud Rod. & Gerh. Wetstenios.

References are to the Scolar Press facsimile of the Dunciad Variorum (1966).

"Pope's Books" in English Literature in the Age of Disguise, ed. Maximillian E. Novak (1977), p. 240. The Boileau is a two volume quarto: Oeuvres de Mr. Boileau Despréaux. Avec des Éclaircissemens Historiques, Donnez par Lui-même, A Geneve, chez Fabri & Barrillot.

The Works of Alexander Pope, ed. W. Elwin and W. J. Courthope (1871), ii, 261. Pope's Works II published in 1735 contained a list of "Variations" at the back of the volume. It seems likely that Warburton changed Pope's conception of the ideal edition.

In Marxism and Form (1971), pp. 390-400.

I was helped in the preparation of this paper by the generosity of Professor D. F. McKenzie, who allowed me to see the script of his 1976 Sanders Lectures, "The London Book Trade in the Later Seventeenth Century," and a pre-publication copy of his essay, "Typography and Meaning: The Case of William Congreve." My approach is more conservative and Anglo-Saxon than his and he might well disagree with much that I have said here, but his work provided an important stimulus for my own.