The problematic iconicity of authors’ manuscripts is illustrated in a very striking way by one of Picasso’s works, which is at the same time pictorial and literary. It is an etching, presenting itself as a rough draft of two of Picasso’s own poems, dated « 16 mai XXXVI Paris » and « 18 mai XXXVI », complete with additions and deletions (very thick cancellations), blots, pen trials and intricate doodlings in the margins (hatchings and human figures, typical of Picasso’s style at that period, but not entirely different from the strange creatures that tend to crop up under the fidgeting pen of any absorbed writer).

The page has all the formal characteristics of a draft, and yet it can hardly be considered as a draft, since it is printed in several copies. It is a mechanical reproduction, a printing of the poems, but it seems very difficult  to consider it as an edition of the two texts. Since a crossing out is interpretable, from an editorial point of view, as an indication that a portion of the text is to be left out, what would be the significance of a multiple reproduction of a crossing out? Should it be supposed that Picasso drafted his poems on the surface of an engraving plate, and that the subsequent pullings should be considered literally as imprints of the creative process? The idea is suggestive but, in this case, we have internal and external proofs to the contrary. First, it must be noted that, in addition to the two dates  that signal the beginning of the two poems (May 16 and 18), the engraving itself is dated « 20 mai XXXVI ». This last date, inscribed in larger characters on the bottom of the page, is the genuine date of the making of the etching, that is to say that it acts as a kind of linguistic shifter, linking the visual and verbal statement of the print to the actual moment of execution. The other two dates mimic the act of dating a manuscript at the beginning of the writing session, but do not carry the same illocutionary value. We could say that they are « mentioned » and not « used ». In the same way, the stylised deletions of this page are “mentioned”, as are all the iconic features characteristic of a rough draft. This is confirmed by the fact that a real first draft of those two poems is extant. It bears the same dates (May 16 and 18) and the text is identical. But the engraving is not a literal reproduction of this draft, which includes other poems and different deletions and additions, or a second draft of the poem, for there is no textual modification and we can check that all the engraved additions are pseudo‑additions, pretending to introduce elements that were already present on the original draft. So it is neither an edition of the poems, nor a facsimile of the manuscript, nor a genuine draft, but a plastic reinterpretation of the iconic value of a rough draft.

This document is a very peculiar case of an already very particular category of writers’ manuscripts:  manuscripts copied post facto by the writer himself, to be used as gifts or to be sold to collectors as real working manuscripts. Documents of the first kind (gift manuscripts) can be genetically relevant when they are the only surviving witnesses of a state of the text, and they can be iconically very rich. It is certainly revealing, for instance, that Joyce made a copy on parchment, calligraphied in Indian ink, of his collection of poems Chamber Music and presented it to his future wife Nora, with their initials entwined on the cover, or that he sent a version of the poem « All day I hear the noise of falling waters » to his friend J. F. Byrne on a postcard that displayed a romantic‑looking photography of himself, but this kind of iconicity can be analysed in terms that are not fundamentally different from those that are applicable to the medieval manuscripts or the Victorian editions presented earlier in this conference.

Much more interesting from our point of view are the authentic‑yet‑forged manuscripts, for they bring an additional level of complexity, being governed, as we have seen in the case of the Picasso engraving, by a secondary iconicity: they not only copy the text of a draft, they imitate its external appearance. Or rather they imitate not its actual appearance but a « conventional » appearance of the working manuscript.

Although such manuscripts, admittedly atypical, do not belong to the category of working manuscripts, their importance for our subject should not be underestimated. They are important semiotically: according to Umberto Eco, the possibility of being used for lying is the touchstone for what belongs to the realm of the semiotic. And they have become historically significant: in the twentieth century, holograph authorial fakes have become rather frequent, simply because the exercise has become lucrative enough. The surrealists, for instance, penned a good number of them as a source of income, and Joyce himself sold the Rosenbach manuscript of Ulysses which is a complex case of a partly authentic, partly authorially forged manuscript. The question then becomes: why do 20th century collectors often prefer these foul papers rather than ornamented fair copies? Since the textual content is more or less the same and the metonymic (fetichistic) connection to the writer roughly equivalent, the significant difference between a fair copy and a working draft must be located in the appearance of the pages. In this context at least, the import of the iconic page becomes tangible (and convertible in monetary terms).

Commercial assessment can be explained here as the slightly delayed echo of a cultural evaluation: since the romantic period the disorder of the draft page is valued because it is felt to be symptomatic of the turmoil of « genius » at work, to bear the direct imprint of creation. For similar reasons (science is no more independent than business from the larger trends of our culture) writers’ working manuscripts have become an object of study: even if genius is no longer considered as a relevant scientific category, drafts are thought to provide the best clues towards an understanding of the process of literary invention.

We have been speaking of clues, symptoms, imprints... Without being unduly metaphoric, we might go on to say that the draft page is like a soft ground and the genetic critic like the hunter who reads the configuration of the tracks to reconstruct the past behaviour of his prey.   This lexical field indicates that the activity of the student of manuscripts clearly belongs to what Carlo Ginzburg has called the « indexical paradigm ». Working manuscripts are indices much more than they are icons in the Peircean sense: in this context it would be more accurate to speak of an « Indexical page ».

This is not to say that so‑called iconic phenomena play no part in genetic processes, but this part is usually limited. There are the cases when an iconic effect is sought in the final printed text: from calligramatic poems to the small coffin appearing in Cash’s monologue in Faulkner’s As I Lay Dying,  the newspaper headlines inserted in the text of the « Aeolus » episode of Ulysses, the theatrical typography of the « Circe » episode, or the marginal notes and footnotes in Chapter 2, Book II of Finnegans Wake... All these do not appear suddenly on the printed page. The visual idea is generally traceable at some stage in the manuscripts (although the actual result is the product of a transaction between this idea and the willingness and technical capabilities of the printers) and it is always interesting to note when it materialises. It appears very early in the case of « Circe » — it is introduced on the third page of the very first draft — and rather late in the case of « Aeolus » — at a stage (galley proofs) which precisely represents the interface between typographic elaboration and authorial intervention. It is also instructive to see how the visual innovation interferes with the text. In « Circe », the introduction of the dramatic typography coincides with a paradoxical change in the narrative technique: instead of a more objective perspective, it opens up the way for the intrusion of a radically subjective point of view. In  « Aeolus », the superimposition of the headlines triggers a new round of expansion of the text and the production of several new paragraphs. In the majority of cases, however, particularly in manuscripts of fiction, the final page lay out is not an issue at the time of  composition or does not play an important part in the genetic process.

Another type of iconicity in working manuscripts is represented by the images of all kinds, that accompany, duplicate, illustrate or relay the text in many different ways. Its minimal form is the very widespread habit of doodling, which can be very discreet and sporadic, as in the case of Flaubert, or extremely frequent, in writers as different as Beckett and Pushkin. Pushkin’s notebooks are visually very striking, with numerous sketches and caricatures in the margin. The relation of drawing to text can often be easily discovered. At one point, it is clear that Pushkin had difficulties in the process of describing one of the Finnish milkmaids of Saint Petersburg, stopped writing, made a sketch in the margin, and then went back to his description. A similar but not identical case can be found in Stendhal, who mentions Sancho Panza in a ridiculous situation, stops writing in the middle of the page, scrawls a tiny caricature and then resumes below, passing on to Don Quixote.

Stendhal provides us with many other examples of images, caricatures but also topographic sketches, diagrams, and even engravings by old masters, bound with the manuscript sheets... Even in his case, however, we should not make too much of this form of iconicity, for most of these images are concentrated in a single manuscript (La Vie de Henry Brulard). The drafts of his fiction, in particular, include less drawings – which does not mean that they are devoid of visual interest, on the contrary. Neither is there any drawing in Joyce’s manuscripts. Joyce was not pictorially oriented, perhaps because his eyesight was so bad. This is precisely the reason why we will chose him as our central example: he is more representative of the fundamental aspects of our subject than writers as visually gifted as a Pushkin (or of course a Picasso).

The danger of the notion of iconicity, when it is applied to manuscripts, is that it implies similarity and supports the misleading intuition that manuscripts must be similar to the final work that is issued from them not only in their verbal content but in their general make‑up, in other words, that the work must look like its manuscripts.

It is easy to find concrete examples to contradict this idea. The first page from the draft of Georges Bataille’s Le Coupable is literally oppressed by heavy and opaque cancellations, incarcerated behind a close hatching systematically crossing out entire paragraphs. In between those compact blocks a few sentences laboriously manage to surface:

« J’écrirai ces notes incapable d’autre chose. Maintenant il me faut me laisser aller à des mouvements de liberté, de caprice. Le moment est venu de parler sans détour. »

Who would guess from the surviving text the compulsive appearance of the manuscript page on which this declaration of spontaneity was produced? And who would guess from Faulkner’s dishevelled books, full of sound and fury, that they are issued from incredibly neat looking manuscripts?

Beyond such empirical refutations, one might question seriously the very notion of « similitude » as an instrument of analysis. Umberto Eco has convincingly shown that the sign‑icon‑index trilogy is the least rigorous of Peirce’s trichotomies, primarily because of the impossibility to define consistently the relation of similarity with its object which is supposed to characterise the iconic sign. Therefore, it seems better not to take the title of this conference in its strict Peircean sense and to consider that it refers to the page as a complex semiotic object, particularly in so far as non‑verbal aspects are concerned and the type‑token ratio is what Eco calls a « ratio difficilis », that is to say all cases in which there is no predetermined relation between an expression‑type and an expression‑token, imprints as well as pictorial representations.

In that perspective, let us look closely at a draft for the « Sirens » episode of Ulysses (Buffalo V.A. 5, pp. 14‑15 ‑ fig 1). It is a visually striking example, but it is not exceptional in this respect among Joyce’s manuscripts. The two facing notebook pages can even be considered as typical in their lay out: the main part of the text is inscribed on the right hand page in a dense column, while a broad margin, widening towards the bottom of the page, is reserved for corrections, as well as the whole of the left hand page. This general disposition and particularly the accentuated slant of the margin could almost be used as a clue to identify the writer.

The document, however, is exceptional in one respect: we have no previous version and no later version of a good portion of the material present on this double page. The main column of text is obviously a copy of a previous draft, but this first version did not survive. Nothing uncommon here, for most of the early drafts of Ulysses are not extant. But the lack in the other direction is much rarer, for Joyce very rarely discarded anything he had written, while here he simply dropped, or reused in other contexts most of the additions on the left hand page. The consequence is that we are deprived of the correlational evidence which is normally decisive for reconstructing the genesis, and we have to rely very much on visual clues. Since visual elements are meaningless, or at least ambiguous when taken in isolation, it is absolutely necessary to reconstruct another form of correlation. This will normally be found in the internal logic of the surrounding text and in the knowledge of the author’s usus scribendi . The problem is that this passage features a radical transgression of the narrative rules that obtain up to that point in the novel: the fragment « a rosery of Fetter Lane of Gerard herbalist he walks. Afar hands on whiteness laid », inscribed in the centre of the left hand page, is felt like a shocking intrusion in a passage dominated by Leopold Bloom’s monologue and an impersonal narrative voice, for it comes straight from the inner thoughts of Stephen Dedalus, a character who is entirely absent from this episode. Such a disruption of the continuity of reading conventions (and writing strategies) makes inferences very difficult. However, this particularity should perhaps be considered as emblematic of the status of draft material, for genetic studies inevitably encounter difficulties of this kind in so far as they are concerned with invention, as opposed to textual studies strictly speaking that are concerned with modalities of repetition. Correlations between states of the text are indispensable but they are elusive, since there is a potential discontinuity between each state of the text in the course of its creative elaboration. For our present preoccupation, this means that the verbal and non‑verbal signs that display themselves across the manuscript page can never be fully contextualized and therefore can never be interpreted univocally. This is a limitation of  genetic criticism — but it is also a condition of draft writing, for the author also finds himself in the position of an interpreter of his own graphic signs, and interprets them from the point of view of the development of his invention which is not always an accurate historical interpretation, for the author often becomes so obsessed by the retrospective obviousness of the new context that he no longer has access to the original meaning of those signs.

In order to understand the characteristics of this document better, it is interesting to compare it with another artefact that shares some of its visual characteristics. Arbitrarily we have chosen a double page (fig. 2) from a newspaper, The Dublin Evening Telegraph for June 16, 1904, which was used as a source by Joyce for Ulysses and even for this particular passage. There are some obvious common points. Both make use of the full surface of the double page, and both occupy it in a not entirely sequential way: they are not meant to be read from the top of the left hand page to the bottom of the right hand page. Both openly acknowledge the de facto tabularity of writing (as opposed to the linearity of  spoken discourse). Moreover, they are composed of discontinuous but not unrelated fragments of writing (news of the same day in one case, material for the same passage in the other).

We even find one purely indexical element in the newspaper, a vestigial trace of the sedimentation of time. The stop press column, printed with a different system, is visibly paler, so that the practiced reader does not even have to read the heading: he is able to catch an instant glimpse of the fainter column indicative of the most recent news (or to be more precise, indicative of the most recent printing and indirectly of the most recent news). In the same manner, the student of the « Sirens » draft will use the change of writing instrument (pen then pencil then crayon) and the different shades of ink to try to establish the chronology  of the different marginal additions.

However, the differences between the two written documents are much more important. The most obvious is probably the difference in the circuits of interlocution: the newspaper is clearly meant for its readers, while the draft has no reader. No other reader, that is, than the writer himself: the signs on the page, iconic or otherwise, have no addressee other than their own writer. They are meant for his eyes only, but this does not mean that there is a perfect identity between the writing and reading agencies. The draft page is the locus of a dialogue between the writer and his later self or selves. This dialogue can be more or less intense, and the lag or deferral, inherent in the notion of writing, can be more or less important. The writer can read his own signs almost at the same time as he writes them, or months, or even years later: in the course of the composition of Finnegans Wake, Joyce left aside some drafts for more than fourteen years before he started work on them again. In some cases, the writer knows that what he writes is going to undergo an almost immediate revision and rewriting, and that his reading will be supplemented by fresh memory, so that the indications need not be very precise. This is clearly the case in our « Sirens » draft: the connecting lines are very loosely positioned, the relation between some fragments is not made explicit at all. It is one of the paradoxes of the semiotics of the draft page. The lack of indications is itself indicative: it tells us something about the circumstances of composition. We can also find more direct proofs of the reading and rereading tempo: when a writer reads his own page, he usually does so pen in hand, and different layers of revision can be set apart. On the « Sirens » draft, we can immediately distinguish in the margin a series of additions written in the same ink as the main column, which must have been done in the course of writing this column, or almost immediately afterwards, and others written in pencil, which belong to a later round of revisions.

As a consequence, the student of manuscripts is reading something that is not intended for him. Not only is he is an intruder in the circuit of interlocution (like someone who intercepts a private correspondence) but he reads the page in order to find there something that was not its primary purpose and interprets as an index something that was written as a series of straightforward symbols. In this respect however, he is perhaps not entirely divorced from authors who keep their manuscripts, instead of throwing them away when they are no longer useful, partly because of this secondary significance, because they bear the imprint of their toil. Beyond this, we could even consider that we share this condition of parasitic reader with the writer himself, who, in the process of revision, must become a decipherer of his own writing when it is inscribed on the page, and sees in them more (and sometimes less) than he has put, in an act of creative interpretation.

The visual characteristics of the page play their part in this re‑interpretation, they are active factors in the creative process. It is generally admitted that painters do not simply reproduce on the canvas what they see with their eyes or their mind’s eye: every brushstroke alters the balance of the composition and calls for a response on the canvas. The page itself is not a mere receptacle for a pre‑existing text issued from a previous mental or material draft, it is a dynamic creative space in which different graphic elements interact. For instance the fact that an addition in ink shows through the porous page can trigger an entirely unexpected development, providing that this material accident encounters a favourable textual context.

There is another difference between our two documents, however, that is probably even more important: the draft is not a text, or a discourse, it is a protocol for making a text. It can be compared to a musical score, which is not melodious, not even sonorous, but engenders music; or to the colour names jotted down by the painter on a rapid pencil sketch, which are not pictorial elements in themselves (although they may acquire a secondary pictoriality, for instance in the paintings of Jasper Johns) but instructions towards a future picture.

If the draft as a whole can be taken as a set of instructions, each of its components does not have exactly the same pragmatic status. Some verbal chunks simply convey the implicit prescription « to be repeated in the next version » while others carry much more complex commands. For instance in the main column of our « Sirens » draft, half‑down the page, we can find the phrase « Or if not? ». The question mark, like the preceding words, is simply to be carried into the next version. But just to the left, in the margin, we find another question mark (« Molly likes left (?) side of her face best. ») with a very different pragmatic status. It is not supposed to pass into the next version but expresses a (self‑addressed) order that could be approximately paraphrased as: « Check whether it should be the left or the right side ‑ for reasons of symbolism or consistency... » We could say that here the question mark is « used », while it is simply « mentioned » in the first instance. But such a distinction is not sufficient if we look at some textual fragments on the left page. On top of the page, we can read the words, « Hee hee hee hee : Wait while you wait. Hee hee a preacher is he hee hee. Hee hee. I wonder did he. Hee hee. Practise what he. Hee hee hee hee Preach while you wait », while, near the bottom of the page, we read a very similar passage, « Wonder did he practise what he preached — wisdom while you wait. Hee hee hee hee. a preacher is he. I wonder did he. Practise what he ». It seems that both these fragments are not to be carried into the next version. They are variants of the same passage, but the alternative seems to be left open. The implicit instruction here seems to be « Choose later one of these two versions and disregard the other ». Another category of signs has a different but simpler status. For instance the word « (back) », just beneath the large B in the margin of the right page,  simply indicates that the B insertion is to be found on the back of the page. It is a pointer, operating at the same level as the large marginal « A », « B » and « F », or the figure « 28) » (which conveys the instruction to “copy this page after the page marked 27 and before the page marked 29”).

How do we, readers, discern the pragmatic value of those heterogeneous elements coexisting on the same page? How did Joyce himself retrieve from them his own unformulated commands? How are we to discriminate between, say, the two question marks? It is of course the context which allows the distinction, but the spatial as well as the semantic context is taken into account. On the draft page, signs are toposensitive.

Beyond the usual toposensitivity of writing (a word written to the right of another normally comes after this word), there is an extension (and even a distortion) of the relation of proximity, to the point that we often hesitate to determine the correct word sequence. But it is not only a laxity of the code, it is also a recoding of normally non‑pertinent features, such as the distance between two words, or the relative size of the letters. A large capital B does not have the same function as a small capital B — but how large has a letter to be in order to be identified as a pointer? It depends, again, on the context. The size is relative, and so is the position of words to one another: the « (back) » is attracted by the « B » rather than the « F » or the neighbouring words « to hide them » or « nibbling » because of a number of converging criteria, visual (closeness, direction of the ductus, writing instrument) and semantic.

However there are some meaningful absolute positions on the writing space, somewhat like the advertising section, the specialised columns, and the stop press area on our newspaper example. Frequently (but not in Joyce), what we called the dialogue between the writer and his later selves takes the form of an explicit commentary, inscribed in the margin, like the annotations of a teacher. The margin is also the designated place for the additions to the text. In our draft, the left page is used as a supplementary margin, and this wide surface accommodates the expansion of the main column, but also a freer flow of invention, elements that have not found their perfect formulation or that are not yet integrated into the textual sequence. The metatextual question mark of « Molly likes left (?) side of her face best. » could not be found in the main column because it turns the sentence into a tentative note. On a wider scale, the whole transgressive fragment mentioned above (« a rosery of Fetter Lane of Gerard herbalist he walks. Afar hands on whiteness laid ») is boldly written across the page but is not anchored to the main text. Much of the work which is visible on the left page is an attempt at integrating it into the narrative logic of the main text.  This attempt will not be successful and most of the content of the page will be dropped and partially reused at another location in the episode.

So there is an undeniable specialisation of regions of the page, but this should not be considered as rigidly determined. For instance, almost half way down the main column, a small  interval between paragraphs is occupied by a pencil addition written in a loose, sprawling hand (« his voice unfolded »), that is very vaguely referred to the text. It is a marginal note transposed into the text, or we should rather say that a textually meaningful blank is turned into a small internal margin.

Even beyond such occasional transgressions, it is clear that a complex game of attraction and repulsion reconfigures the page dynamically. In this respect, drafts are different from charts or diagrams, in which each position has a predetermined meaning. To grasp this distinction, it is instructive to look at the drafts of a chart. Unfortunately we have no draft of the famous Ulysses schema, but many examples of diagrammatic manuscripts can be found. If we look for instance at the table (« Tableau de contes ») that the Marquis de Sade drew for his projected book of tales, we see a strict arrangement of lines and columns in which each position has a precise significance —but it is visible that in the course of actually trying to write some of the tales, several successive changes came to alter that rigid disposition, titles are cancelled, replaced by others that get inserted in between the lines, so that they fall outside of the rigidly assigned positions... An extrinsic temporal dimension interferes with the conventional use of the two dimensions of the page: the toposensitivity of the draft page is combined with a chronosensitivity and the spatial lay out can be partly  translated in terms of temporal succession.

According to Roy Harris’s « integrationnal » theory of writing, « It makes no difference whether the individual writer sets down the forms in this sequence or that (any more than it matters whether an artist begins by painting the grapes or the apple or the oysters in his still life). What matters, as in the case of the still life, is that the forms shall be finally arranged in such a way as to articulate the desired composition. » This may not be true in the case of Chinese ideograms, and it is certainly inaccurate for on‑line systems of writing recognition, but it is completely false in the case of genetic criticism. To the geneticist, on the contrary, the sequence of inscription is one of the most important messages to be deciphered on the manuscript page.

Without going into great details, we can distinguish on the « Sirens » draft in the margin of the right hand page, a first layer of additions to the main text, which comprises the following insertions, from top to bottom: « Music hath charms owls and birds », « nibbling », « Doesn’t half know [...] Cockcarracarracarra », « Molly likes left (?) side of her face best. », « Read out in court », « Write her letters [...] Lionel. Leopold. » It is difficult to ascertain a priority between those, although it seems that the long insertion, « Doesn’t half know [...] Cockcarracarracarra », came after « Music hath charms owls and birds », for it is situated much lower in the margin than its point of insertion (the « F » on the fourth line), which suggests that its natural space was already occupied. Then came a second round of additions, in pencil: the « A » insertion, « Basses under us [...] to hide them », starting on top of the page and working round the pre‑existing « Music hath charms owls and birds », and lower down « then hastened » and « chorusgirl’s romance ».

The left page was used as an overflow when the margin was full, in order to inscribe additions such as « Woodwinds like Goodwin’s name mooing like cows. », « doublebasses lying there [...] their harps not said », « shell out », « want to keep your eyes open ». But the whole cluster organised around Shakespeare and « a rosery of Fetter lane... » pre‑dates those additions. The way it is written — across the page, in a large hand — proves that the page was empty —while the long passage, « doublebasses lying there [...] their harps not said » is boxed in between pre‑existing inscriptions, squeezed in an exiguous space.

This simplified analysis is sufficient to show how the disposition on the surface of the page can be  translated in terms of a linear chronology. Genetic chronology, however, is anything but a simple matter: the temporality of invention is non linear, involving complex processes of anticipation and retrospection. Even material clues can be misleading: there is a page of the manuscript of Stendhal’s Lucien Leuwen on which only the margin, a very narrow margin, is occupied. If the centre of the page was covered with writing, we would be ready to affirm that the cramped marginalia was inserted after the main text... In this case, a functional reading must prevail over a chronological interpretation of the marginal situation. For Stendhal, obviously, writing on the periphery of the page did not have the same meaning as writing in the centre.

Moreover, the chronology of inscription on the surface of the folio is not the only time‑line to be considered. The spatial lay out is also indicative of a stage in the genesis of the work. Our « Sirens » manuscript is visibly a draft, which is not a first draft (the dense textual column indicates that it was fair‑copied from an earlier version). In the case of Flaubert, for instance, many stages are clearly differentiated: notes, scenarios, general scenarios, partial scenarios, local scenarios, stylistic scenarios,  summaries and recapitulations, drafts, fair copies... For some of these categories, he used different kinds of papers (different in size and in colour) and occupied their surface in a different way. The spatial disposition can also be indicative of the different stages of a writer’s carrier. In the case of Joyce, the evolution is blurred by his deteriorating eye‑sight. But Flaubert again offers a perfect example of a continuous mutation in his management of manuscript space, from his early works to his great novels and to his final tales. To make this brief survey of the chronosensitivity of the manuscript page less incomplete, we should add that it reveals not only relations of succession, but changes in the speed and the rhythm of writing. It is for instance immediately evident that the column on the right of the « Sirens » draft and the page on the left were not written at the same tempo.

A few concluding remarks. The physiognomy of the draft page is a most individual characteristic, like a kind of thumbprint of the writer’s mind: it is a space of complete freedom, that he occupies as he pleases since it is meant exclusively for  his own eyes. It bears the imprint of his will, of his habits, of his compulsions. And yet the signs used on this surface are not very different from public, socially accepted systems. Very few Pepys or Leonardo bother to invent private systems of writing or even of proof‑marking and those are usually a mere deformation of public ones. In the same way, the organisation of the page is usually derived from editorial practices or school models. This suggests that the inner or private dialogism that expresses itself on the surface of the draft is perhaps not so different from ordinary social intercourse, that its rules are distinct but not entirely divorced from public conversational rules, or at least that they are strongly contaminated by those. The very fact of using a socially accepted system of signs in the private sphere cannot be neutral, it is bound to impinge on the autonomy of that sphere. Therefore it should be possible to establish significant connections between drafts and the iconic pages of public manuscripts and printed books presented during this conference. This can be a useful reminder, because the fascination for the individuality, the « uniqueness » of the manuscript page is dangerous if it leads to a purely empathic reading of its direct expressivity, to what E. H. Gombrich calls a physiognomonic interpretation. As Gombrich suggests, prevalence of context (historical, textual, genetic context...) over expression must be the rule.

It would also be a mistake to oppose the page lay‑out and the textual data as representing, in Lyotard’s terms, the figural and the discursive. The figure is also present in the text, but the figure of the page helps us to understand how the text acquires its figural status.

Because of its very obviousness we run the risk of missing this major clue, like those names on geographical maps that are undetectable because (to quote « The Purloined Letter ») they « stretch , in large characters, from one end of the chart to the other » and yet convey the most important information, or like « The Purloined Letter » itself, invisible because it is so blatant, and yet so powerfully active...