According to the usual definition, a rough draft (brouillon) designates, very broadly, a work manuscript written with the intention of correcting it for use in the composition or definitive polishing of a text. To take only its literary sense, this definition carries the advantage of having a wide application and the defect of having, to make up for it, under these very loose specifications, a reduced comprehensibility. This explains why the word is frequently used and why a certain embarrassment often accompanies its use for philologists and literary historians, enamoured of terminological rigour. The difficulty is all the more palpable nowadays in that a new breed of researchers — specialists in literary genetics — have brought the rough draft centrestage by laying stress, over the last few years, on the remarkable benefit that the critical study of texts could derive by recourse to these genetic documents, in which the work of art becomes interpretable through the very movement which gave birth to it. Originating from widely diverging critical outlooks and working on corpora which are themselves in great contrast to one another, these literary geneticians do not constitute, in the official sense, a "school”, which would have its own terminology fixed in place by dogma. Their panoply of ideas and concepts can alter as they acclimatize themselves to works and to critical orientations chosen to interpret the avant‑texte1: a genetic study of the unconscious in the rough drafts of a poetic work will not use exactly the same terminological instruments as a narratological study on the compositional genetics of the writing of a novel. Whatever approach is chosen and corpus examined, however, genetic critics have grown accustomed to making to mesure, as precisely as possible, the categories with which they try to classify and interpret genetic documentary evidence. And among these typological conceptions, the category of the “rough draft” has, over the last twelve years or so, been the object of various attempts at definition, principally in the shape of monographs on genetic research.
The result of all this is that the signification of the word “draft” nowadays has rather a broad margin of uncertainty. For the textual critic, concerned neither with manuscript nor with genetic development, the word continues to mean, according to its usual definition, a vague generic term designating the approximate and negligable domain of all that precedes the finished version of the text: a sort of opaque space in which the structures of signification and style are not yet in place and which remains resistant to interpretive designs upon it. For the literary genetician, whose time, on the contrary, is devoted to understanding the pre‑textual process, the draft is an essential link in the chain of transformations which have led from the work project to the definitive text: a crucial moment in the  avant‑texte stage.
If, however, geneticians are in principle agreed as to the importance of the rough draft, the content of this consensus remains ultimately undecided, and the concept itself papers over extremely varying realities, according to the corpora studied and, above all, according to the type of genetic study envisaged. Microgenetic analysis, which sets up and interprets the total compositional development of a short textual fragment, opposes itself to macrogenetic research, which looks at one or several complete collections of genetic documentation, and which studies large‑scale phenomena, and they will not necessarily have the same outlook on the “rough draft”. Microgenetics might, due to the restricted dimensions of its object, adopt the hypothesis of an expanded version of the concept of the rough draft (and extend it, for example, to cover development from the very first initial workplans to textual adjustments made just before publication). Macrogenetics, more sensitive to the diversity of genetic ingredients, attentive to pre‑textual structuration problems, and studying objects of vast dimensions (thousands of pages long, for example), will have a tendency to privilege a more tightly marshalled definition of the rough draft, conceiving of it exclusively as the compositional space, completely distinct from manuscripts concerned with the pre‑planning, the structuring of the scenario, or documentary research. In short, each will develop their definition of the rough draft in accordance with their objectives. While this attitude appears perfectly justified when the definitional principle is clearly announced, it still seems that an overall consideration of the problem has not as yet been forthcoming. This will be the aim of the present typological study, within the specified limits of one genre.
Defined as instrumental to the composition and honing of a text, the rough draft constitutes a step almost always indispensable for the writer, which leads one to reflect that, in one form or another, drafts have probably always existed, even if few examples prior to the middle of the 18th century have survived to our times, with the exception of a handful of cases all the more striking because of their rarity2. By contrast, because of significant cultural and intellectual changes which modified thinking and behaviour from the latter half of 18th century Europe onwards, literary drafts have been preserved, by the writers themselves, with some care, right through the 19th and 20th centuries, and sizeable collections of manuscripts represent the great literary corpora of this period in archives today. In the case of certain works, practically every handwritten piece of work can be accounted for. Research in these archives shows that the use of the rough draft varies very considerably depending on genre and author, and sometimes, in the case of a single author, on the work. While non‑existent or very little used by some writers, the rough draft reveals itself to be, in the majority of cases, a decisive step in the creative work. Using different writing techniques, but comparable amounts of work, each page of definitive text called for five or six pages of drafts for novelists like Balzac or Flaubert; tricky passages possibly demanding up to fifteen or twenty pages of successive redraftings. In the Flaubert archives, the rough drafts of Madame Bovary come to thousands of pages (including scenarios, copies and documentary research, a total of around 3700 large manuscript pages); around 2500 sheets for L’Éducation sentimentale are to be found, written for the most part on both sides, and, together with the preparatory notes, close on 3500 sheets for Bouvard et Pécuchet. In contrast to the definitive manuscripts studied by philologists from the 19th century onwards, these “snarled‑up” (“embrouillés”) documents, covered in crossings‑out and additions, often difficult to order and decode, have only recently acquired the status of research object worthy of systematic analysis. By substituting the study of the writing processes (which implies a systematic examination of the whole collection of a work’s manuscripts) for the list of variants (compiled by philologists almost exclusively on the basis of the definitive manuscript or, at best, on the very last compositional states), genetic criticism has placed the rough draft at the centre of its investigations.
From the point of view of the genesis of a work, the rough draft can be considered as a sort of text laboratory, in which it becomes possible to piece back together an essential phase of the writer’s work, by tracing each one of the writing movements, observing, as if at the time it took place, choices, indecisions among the array of invented possibilities, bursts of speed and moments of discouragement or block in the composition, sudden inspirations or chance errors that sweep aside the difficulties and set the writing off again in a new direction. The rough drafts tell a kind of day by day story at once logical, symptomatic of affect and phenomenological, which is none other than the life of the writer at work: a secret tale, almost always absent from literary biographies, and which constitutes nevertheless the crux of what we would like to know about the author. Above all, however, it is in the rough drafts that the exact role that might have been played by this or that source in the composition can be identified and evaluated; the location and the bearing of autobiographical inspiration in the writing can be pinpointed; and the narrative, dramatic or symbolic structures, which will constitute the very foundations of the work, can be seen being built up, piece by piece. In short, the rough draft enables us to be present at the birth of the motivations, strategies and metamorphoses of writing, which more often than not, labours precisely at effacing its own tracks, and at rendering its mechanisms untraceable, secret or problematic in the completed form of the definitive text. The rough draft offers criticism an essential sphere within which to justify interpretations that the text often leaves in a hypothetical state, and sometimes even in a state of simple conjecture.

One moment in the writing process

Although the rough draft represents an essential document for the understanding of the work in the light of the overall process, of which the work is the effect, it nevertheless constitutes only a moment of this process: a central moment, to be sure, which forms the very heart of its genetics, but an intermediary and provisional moment, caught between two other universes, that of the initial conception of the project which in general precedes it, and that of the preparation for publication of the text, which follows it. Between initially hypothetical workplans, notes and scenarios, which permit the study of the project’s birth, and the finishing touches to an already completed text, only grooming its appearance by tweaking at a few minor details in the definitive manuscript and at proof stage, the rough drafts constitute the very universe of the composition: the moment when the project passes from the state of hypothetical schema to the state of textualized verbal material and becomes involved in a succession of metamorphoses which, at a certain stage in the process, will produce the decisive tip from possibility to necessity, from probability to proof. This could be termed the moment of coagulation of the work. The rough drafts enable the observation of this phenomenon: how something which, up till then, had been nothing but a hypothetical product of the work of writing, changes into a regulatory and causal authority. At a certain point of development, an advanced state of drafting settles like a dividing line across the history of the avant‑texte, showing that an insistent image of the text has acquired enough density and stability to give rise to and police all further transformation of its form by the activation of its own controls. It is the modalities and limits of this series of mutations that I hope to specify here, by demonstrating through a typological description how this compositional step, while possessing an overall identity and coherence within the genetic process, is far from constituting a simple transition, but is rather itself transformed by the way in which quite distinct writing activities are put to work, belonging despite their interdependence, to a logical, complex, mediated and diversified scenario.
The ideal way to arrive at my goal would be, it seems logical to believe, to set up this typology on the basis of a vast assemblage of genetic corpora which would embrace several centuries of literary production, in diverse linguistic cultures, and with as wide a spectrum of genres as possible at its disposal. The realization of such an undertaking could hardly happen overnight. Over and above its unrealistic side, however, it seems perhaps doubtful that such an ambition could even be pertinent from a scientific point of view: the genetic approach lends itself only poorly to theoretical speculation about transhistorical and transgeneric totalities; it seems better suited to relative generalities which allow for historical delimitation, for specific generic constraints, and for oddities which can become normative in the history of genres. I will come back to this. In contradistinction to this vast undertaking, the proposition of a functional typology which follows constitutes a partial approach, based on the study of a few corpora, in the context of a genetic study focused exclusively on prose works. Flaubert’s manuscripts, particularly rich and complex from a typological point of view, have been selected as a foundation for my thinking, without, that is, neglecting particularities appropriate to other narrative corpora, and with the ambition of creating a typological tool applicable to numerous other collections of genetic material3. The analysis of Flaubert’s manuscripts brings to light a considerable number of document types; this inventory may prove to be inadequate by excess for other genetic studies, for which the compositional documents are often more limited in number and type. In most respects, however, the system works (almost) as if compositional collected materials could be interpreted in this typology’s terms, on condition that certain stages or genetic categories are deemed void or non‑existent. The particularities specific to certain corpora will obviously entail a refocusing each time the typology is applied, which might, for example, cause redistributions of intensity to show up, in the relative importance granted by the writer to this or that phase of the process, or changing preferences as to the type of documents the writer uses for the composition proper, without, one hopes, these particular configurations disrupting the general typology to the point of calling the whole conceptual arrangement into question.
I will define the collection of genetic documentation (dossier de genèse) as the whole body of known, classified and transcribed work manuscripts and documents, connected with a text whose form has reached, in the opinion of its author, a state of completion or near completion. When fairly complete, the genetic documentation of a published work generally exhibits four overarching genetic phases which I have entitled the pre‑compositional, compositional, pre‑publication and publication phases4. Each one of these four phases can be broken down into several moments and several functions to which the individual types of manuscripts are related. The body of documents generally referred to as “rough drafts” of a work corresponds for the most part to manuscripts from the second phase (labelled “compositional”), but, in certain cases, this body can also include some entries grouped under the first phase (“pre-compositional”), prior to it, and certain mutations normally grouped under the third phase (“pre-publication”), after it.

To take just the compositional phase, the study of manuscripts often enables us to make out several steps which cause developed scenarios, general sketches and partial roughouts to succeed each other (with the function, for example, of developing the contents of an inital workplan or scenario, sometimes without shaking off the list‑like or sub‑compositional style which characterised the first phase). These are followed by notes on documentation to be used in composition; then a succession of compositional states in the true sense (that is, true rough drafts, which proceed progressively to full sentences, and possibly to the admixture of various documentary sources, in a general process of structuring and articulating the material‑turned‑text). Finally, we can distinguish a series of advanced rough drafts followed by fair copies more or less reworked (undertaking rewrites and specific corrections to an already stabilized base text), in the stage which immediately precedes the fair copy of the pre‑definitive manuscript, and then the final manuscript. These last belong to the pre-publication phase but can still sustain substantial modifications.

It can be seen that an understanding of the “rough draft” can only be constructed differentially: it only makes sense, because of its mediatory and heterogeneous nature, when brought into a relationship with a whole set of functions and documents which can hail from quite distinct moments of the process of the avant‑texte’s evolution.
In order to consider this more closely, I think this is the moment to present, in synthetic and synoptic guise, a general table of the stages, phases and operational functions which enable the ordering of different types of manuscript according to their location and status in the process of production of a work. The complete conception of the “rough draft” can then be seen represented in the context of an evolution in which it constitutes only a transition (more or less expandable according to the extension required for the concept) in the chain of transformational phenomena which lead from the initial idea to the published text.

Typology of Genetic Documentation

Functional Typology of Genetic Documentation

This typology, then, relocates the question of rough drafts into the context of a global evolution.

Process and Chrono-typology

This functional typology of genetic documentation is presented as a table with two ways in. The horizontal axis realigns the different categories of genetic documents, by sub‑group, according to their operational function and on the basis of their alliance with a definable moment of the genetic process. From left to right, the table can be read as the framework of a completed genetic study: it moves from process to documents, from the widest typological conceptions to the most finely‑tuned of descriptive categories. From right to left, the logic is instead that of a study which takes semi‑empirical data (the genetic ordering of genetic documentary material) as a starting‑point from which to piece back together the over‑arching articulations of the genetic movement, to identify the processes at work and construct a global interpretation of the collected evidence.
The vertical axis presents the genetic process as a continuum and enables the realignment of the different document types by the chronological and logical order which links each step of the process along the temporal axis. It leads from a provisional avant‑texte, where the hypothetical traces of the very first formulation of a compositional pre‑project may be perceived, still in the unnameable state of vague, unstable idea, right through to the possibility of handwritten corrections made by the dying author to the final edition of the work printed in his or her lifetime. Conceived of in this way, the table is intended to represent the complete spectrum of the transformations which constitute the very object of the genetic approach: the writing process, described according to its stages, its phases, its operational functions and according to the types of document which are linked to each of these sequences. This typology is a chrono‑typology: each of the denoted moments takes on meaning from the relationship of contiguity by which it establishes itself as the intermediary link in a chain of modifications developing along the axis of a temporality and logic: that is, of a teleology.
This teleology, however, with its heuristic and structural undertones, does not imply any finalist presupposition at all. The table presents the continuous unfurling of a general model of pre‑textual production which allows us to take in a full cycle, synoptically, going from a crystallization of a primitive pre‑project to the publication of the printed work. Yet this continuity is only a model, which, when used to interpret a real genetic evolution, must take into account all sorts of ruptures, doubts and reversals: until the “pass for press” stage is reached, that is, throughout the pre‑textual coming into being of the work, nothing can ever be considered as absolutely guaranteed. The inital processes can set off one or more “false starts” which in some cases end up in a blockage serious enough to put off the composition by several months, sometimes by several years; even in the compositional period itself, when the project would seem to be safely in place, radical doubts can lead the writer to give up his or her original idea completely, so as to follow up leads that nothing (or almost nothing) had prepared us for; and even at the far end of the journey, at the final step of the corrected proofs, the avant‑texte can still undergo considerable upheavals which will radically modify the image and dimensions of the text eventually published.
This typology then, should be seen, not as the picture of the actual genetic unfolding,  along the axis of time, but as the abstract diagram of the logical links which let us name and classify genetic documentation relative to its function. If, apparently at the last step of proof correction, the writer is not content just to tinker with details, and if he or she then takes up the practically finished text again, against all expectation, in order to plunge into a new campaign of large-scale rewrites, then it could reasonably be said that these corrected proofs function for the writer as a new tool for textualization, that is as a rough draft or a corrected fair copy. It can be said, for this particular case, that, contrary to type, the pre‑publication phase, theoretically given over to the manuscript’s finishing touches, has been given a compositional function. Although by no means frequent, these phenomena of functional regression are not exceptional either, which indicates that the overarching stages described in terms of phases and operational functions do not constitute successive sequences separated by fixed frontiers which may only be breached in an irreversible progression towards the finished state of the work, but instead constitute large-scale operational units whose boundaries can permit a certain mobility and which can be crisscrossed by multiple instances of reversals. A clearcut typological definition of these large‑scale units and their linking is nevertheless indispensable: it is by means of this that each genetic profile can be described and evaluated, notably in its own differentiation, simultaneously due to its partial adequation to this typological outline and to the specific divergence which distiguishes it from a teleological model in general.
The logical linking of genetic operations cannot be uniformly subject to any final treatment. The very principle of finalization which anticipates the finished form of the work may make its appearance in conjunction with highly variable modalities of the pre‑textual process, and at very diverse instants of it: the principle may be utterly absent at the start of the process and in some cases remain hypothetical or problematic until an extremely advanced stage of the avant‑texte; in other cases, by contrast, the writer’s work may exhibit, from very early on, an evolution under the sway of the need to complete a publishable text. The idea of finalization is rarely absent from the compositional stage and even more rarely from the pre‑publication stage, but in general the typology proposed here does not presuppose any precise determination of the moment when the avant‑texte takes on a final logic or irreversibly stable fulcrum. Finally, these chrono‑typological terms are, in principle, applicable to the analysis of a literary project, whether completed or unfinished, that has never been published.
On the other hand, this typology is clearly only suitable for the genetic interpretation of real documents, in the framework of a collection of genetic materials without too many lacunae, using which it would be possible to piece back together a verifiable chronology of the preserved items by internal and external critique. This chrono‑typology brings several concepts into play (stages, phases, operational function and type of document) whose meaning and use I want rapidly to specify before going on to talk about their application to the issue of the rough draft.

Stages: the Avant‑texte and the Text

The left hand column of the table – Stages – is devoted to the general distinction between Avant‑texte and Text. A horizontal bar cuts right across the table at the level corresponding to “Pass for press” in the column headed Document Type, the decisive moment when what had been in an ever‑changeable and mobile manuscript state up until that point becomes fixed in the suspended form of a published text. Everything above this dividing line belongs to the pre‑textual domain and is a matter for manuscript genetics (or pre‑textual genetics), while everything below belongs to the textual domain and is the province of a genetics of the text (or textual genetics). In trying to elucidate the overall process of the production of the work, the genetic approach sets itself the essential objective of piecing back together the avant‑texte on the basis of manuscripts: that is, a succession of partial and interdependent processes whose linkage builds up the picture of an overall process, interpretable as avant‑texte. Such research can be extended to include a genetic study of the writing processes noticeable in the printed text’s metamorphoses, when the textual stage, as is often the case, also exhibits numerous and substantial transformations of the work over the course of its various editions. Textual genetics differs from manuscript genetics, however, on a conceptual point appropriate to its object’s characteristics: textual variants, something not entirely assimilable to the transformational phenomena observable in work manuscripts.

The Avant‑texte Stage and the Genetics of Manuscripts

The avant‑texte designates, in the work of the writer, the chain of writing operations which have preceded the appearance of the text proper. The pre‑textual stage therefore designates that of the work’s process of production, insofar as it can be pieced back together by the analysis of the author’s work manuscripts, and then interpreted following a defined critical method. Even before piecing this evolution back together, however, such documents will have had to have been first of all inventoried, ordered, dated and deciphered, since they are neither legible nor interpretable in their raw state. The avant‑texte does not therefore mean the material manuscripts (a typological inventory of them makes up the fourth column, Document Type), but rather the critical discourse by which the genetician, having laid out the objective results of their analysis (transcriptions, relative dating, ordering, etc), reads them as successive moments of a process. The typological table’s first column thus offers conclusive and generic classificatory concepts; which could be used to interpret, in their logical concatenation, the synthetic results of research that has, prior to this, analyzed the documents (listed in column 4), according to their operational functions (column 3) and decided which one of the three pre‑textual phases (of column 2) they belong to. Without judging the contents of these classificatory concepts in advance, contents which are always specific and possibly missing in certain cases, they enable us to name, at the level of the abstract model proposed by this typology, a series of seven major stages constituting the avant‑texte as a chain of “partial processes”. The avant‑texte can be interpreted in its overall specificity by assessing to what extent each one of these processes makes its presence felt, and by looking at their nature, content, relative intensity and the way they link up together. It is these processes which are ultimately interpreted by the study of a text’s genetics.

• provisional avant‑texte stage: before any verifiable appearance of the compositional project, this process transforms the body of sources at the author’s disposal in his or her own manuscripts and work notes, into a pending structure potentially oriented towards this project.

• exploratory avant‑texte stage: before any verifiable decision made to undertake the compositional project, this process constitutes, for the writer, an informal exploration of possible ideas, during the course of which, a compositional pre‑project takes shape.

• preparatory avant‑texte stage: the initial process by which the project proper takes shape, a decision‑making process, but also one of conception and in certain cases, of pre‑planning.

• outlining at avant‑texte stage: the processes by which the conception and the pre‑planning of the project are reworked in a compositional framework, in terms of overarching and partial structuring.

• documentation at avant‑texte stage: the processes by which the writer, in certain cases, finds it necessary to equip him or herself with further documentation, in greater or lesser quantities, to sift and transform it ready for integration into the composition.

• composition at avant‑texte stage: the processes by which the composition proper is performed, through all the rough drafts to the definitive manuscript.

• post‑compositional avant‑texte stage: the process of definitive rewriting and checking performed, after the composition proper, on a fair copy of the definitive manuscript, then on the proofs which will be used in publishing the text. The final handwritten changes are recorded on the set of typed-up proofs bearing the writer’s signature under the pronouncement “bon à tirer” {French publishing practice; the English equivalent would perhaps be the phrase “pass for press”. See footnote 4 below}.

The Text Stage and Textual Genetics

Where it exists, the pronouncement “bon à tirer” (“pass for press”) marks the moment when the author decides that he or she can put an end to the general and local metamorphoses of the work, which can thus be manufactured and offered to the public in this form5. From this moment, we leave the pre‑textual domain for the textual history of the work: a history in which the author is still in a position to act upon his or her text6, and which can be diverted via the publishing of a “pre‑first publication” version7, in serial form in the press, or which ends directly in the publication and distribution of the “first edition” in book form. This is the “text” of the work, but, of course, it is not necessarily the final state of the text. The work might see several editions during the author’s lifetime, on the occasion of which the author would have every right to modify the text on new sets of proofs, and it would move into the “variant text” stage. These modifications can be substantial, and even lead to new partial recompositions or to restructurings, for which the author will make use of rough drafts, fair copies, a definitive manuscript, etc, under the same conditions as during the original composition. However important, these modifications (or “variants”) which, from new edition to new edition, can produce perceptibly different versions of the work, do not, nevertheless, have exactly the same status as the transformations to be seen in the collected genetic documentation for the original work. The mutations of the avant‑texte took place in a private writing domain where everything was possible at any time, including total production stoppage, even if the work seemed to be heading towards the achievement of a publishable text. By contrast, modifications post-publication are made in a public sphere where the book’s reality cannot be ignored: they successively affect versions of the text, all equally definitive of the ‘same’ work8, and which can claim the status of a completely separate text each time, without it being in general possible to recognize the logic of a process comparable to the pre‑textual one between them9. Textual variants (or “edition variants”) can be numerous and substantial: they necessitate the fixing of the text’s identity over and above all the modifications that can appear in the course of different re-editions. Apart from the exceptional cases10, the text of the modern literary work is therefore conventionally taken from the “last edition of the author’s lifetime”, to which the final handwritten corrections may possibly have to be added, marked for a future re‑edition by the writer, who was ultimately prevented from checking them by death. This definitive image of the work marks the farthest reach of the field of investigation proper to genetic studies.
Finally, in order to be completely coherent, the hypothesis of a further stage should be inserted after the textual stage, that of the post‑text, which corresponds to the publishing future of the work (its various editions) after the demise of its author. Many interesting cases fall into this category, from the posthumous edition of a text either completed or left unfinished11, such as Bouvard et Pécuchet for example, to the various transformations (by cuts, the introduction of errors, structural modifications, the appendage of critical baggage, paratextual metamorphoses, etc) which characterize the post mortem editions of literary works. It should be clear, however, that this post‑textual stage no longer has anything to do with the genetic perspective. Its domain would be rather that of critical reception and the history of the book.

Genetic Phases, Exogenetics and Endogenetics, Outlining and Writability

Phases

The second column of the typological table is concerned with phases, which, while taking into consideration all the reservations already noted, enables the abstract breakdown of the avant‑texte stage into three overarching sequences (pre‑compositional, compositional and pre‑publication). Each one of these sequences corresponds to one or several partial processes from the point of view of the pre‑textual stage, and from that of the operational functions angle, to one or several specific procedures or operations:

• pre‑compositional phase: a provisional, exploratory and preparatory process; of which the operational functions are: orienting, exploring, decision‑making, conceiving and pre‑planning.

• compositional phase: an outlining, documentation and compositional process; of which the operational functions are: structuring, researching and textualizing.

• pre‑publication phase: a post‑compositional process; of which the operational functions are: the finishing touches and preparing for publication.

Endogenetics and Exogenetics12

Endogenetics designates any writing process focusing on a reflexive or self‑referential activity of elaborating pre‑textual data, be it exploratory, conceptual, structuring or textualizing work, and regardless of the nature of, or how far advanced, such elaboration might be. I will therefore call endogenetics the process by which the writer conceives of, elaborates and transfigures pre‑textual material, without recourse to outside documents or pieces of information, through simple reformulation or internal transformation of previous pre‑textual data. Although it not exclusively endogenetic, the rough draft domain is the endogenetic domain par excellence, the dominant process in the rewriting operations of textualization. In Flaubert’s‑work, thus, as in numerous other corpora, a draft fragment (in which the writer has roughed out and modified his or her sentences by crossings-out, additions and substitutions) will be neatly copied out again onto a new sheet of paper, which in turn will become the object of several correction and addition campaigns, at the end of which the new version of the passage will have turned into something twice as long, for example, as the initial fragment, by simple effect of the writing working on itself. Endogenetics can also, however, be seen at the outer reaches of the avant‑texte: in the work produced by the initial idea (taking the shape of a primitive workplan or preliminary notes in a preparatory notebook) through the use of the writer’s basic reflexive and imaginative faculties; or, again, right at the other end of the genetic development, in the corrections made to the typed‑up manuscripts, which can form the arena of very different kinds of transformations: either that the author, overcome by eleventh‑hour misgivings, sets to, to overturn what had been to all intents a finished text, or that, on the contrary, he or she is content to make the tiniest adjustments of expression, in order, for example, to avoid repeating a word or sound which had not been noticed up to that point. Lastly, endogenetics designates a procedure which can greatly overflow the domain of writing proper: a drawing or doodle can prove to be endogenetic, if it does not depict an external object, but is rather the projection of a fictive entity produced by the writing (the layout of an imaginary town – Yonville, for example – where the action of the novel unfolds; a person’s features; the spatio‑temporal schema of an action or an invented object, and so on). By the same token, an abstract organigram, geometrical or numerical representation of a system intended to structure the writing (as in Perec’s case, for example) belongs to the domain of endogenetics, whether it is in the form of a notebook full of lexical entries (lists of words and drawings of objects to be introduced into the narrative) or of a structural grid which affects the sentence form of the narrative.
Exogenetics designates any writing process devoted to research, selection, and incorporation work, focused on documentation which stems from a source exterior to the writing. It may be handwritten, or not; any documentary notes or copies; any quoted or intertextual matter; any results of inquiries or observations; any evidence of iconographical matter (that gives rise to a written transposition), and in a general way, any written or text‑image documentation, belongs by nature to the exogenetic category. Comments on things seen, on overheard or reported speech, sketches and drawings made on the subject in hand, friends’ letters giving useful information or anecdotes, reading notes, investigatory notebooks, newspaper cuttings, typescripts of interviews or conversations, printed textual fragments and marginalia, bibliographical references, confessions, essays and reports… the list is endless. The exogenetic empire knows almost no bounds, unless they could be said to the limits which, in good methodological faith, ought to circumscribe it to what is written and depicted (when what is depicted gives rise to a written transposition)13. The province of exogenetics extends only to written or drawn documents, excluding the empirical objects or data to which they refer: a live landscape is not exogenetic, even if its precise description may be found in the work; it is the external referent of a written or drawn document which constitutes itself as image in the genetic domain, whether by an investigation carried out by the writer him or herself at the scene, or by a sketch, or by notes taken by a friend on behalf of the writer, or again by any intertextual element that the writer has borrowed from the textual domain. The stuffed parrot that Flaubert had delivered to him when he was writing Un Cœur simple and that he put on his worktable so that he should be flooded with the parrot’s spirit, was certainly much more to him than a document, but it cannot be said to belong to the exogenetic domain, any more than can the Dead Sea, which he claims to have “seen” when he began writing Hérodias. On the other hand, the notes for his trip to the East, which include descriptions of the Dead Sea, and the notes he made from them for Hérodias, as well as the drawings of deer that he made in a notebook (and did not use) to compose la Légende de saint Julien, are exogenetic. The descriptions and topographical sketches of the Normandy coast sent in the post to him by Maupassant for Bouvard et Pécuchet are equally exogenetic, in a completely different way, despite not being handwritten, as are the scraps of textal idiocy copied out for him by his friend Laporte for the Sottisier, intended for the “Second Volume” of Bouvard, or the Mémoires de Madame Ludovica, those indiscreet revelations of Suzanne Lagier’s life of easy virtue which Flaubert used in drafting Madame Bovary and L’Éducation sentimentale. In short, exogenetics does not designate the “sources” of the work (such and such a real person, place, literary work, etc), but the locatable trace of these source‑referents in terms of documents (written or transposed) present in the collection of genetic evidential material. In the absence of any locatable trace in the work’s manuscripts, a highly probable source (for example the influence of Balzac’s Illusions perdues on the writing of L’Éducation sentimentale) could become the object of the same kind of study but under the heading hypothetical exogenetics, in accordance with an analytical procedure that clearly differentiates it from exogenetics per se.
Apart from its definitional limits, which exhibit certain peculiarities, the exogenetic domain seems to manifest, by its nature, a certain paradoxical dimension when it comes to its genetic evolution. For indeed, short of a particular decision on the part of the writer, to make the exogenetic signpost the very substance of the writing14, the pre‑textual exogenetic elements tend inevitably to be progressively converted into endogenetic material. The inscription of their original exteriority is a volatile one, and beyond a certain developmental point, the exogenetic mark becomes so intimately integrated into its endogenetic context that it becomes barely recognizable. This conversion of the exogenetic to the endogenetic takes place primarily as an effect of textualization, an operational function whose execution essentially concerns the rough drafts. After undergoing its multiple transformations, the initial exogenetic element (for example, a topographical comment, a detail or a situation borrowed from a literary work) may have become perfectly untraceable: it has metamorphosed into an organic part of the text, which, for the reader, points only to the writer’s imaginary and to the internal logic of the fiction, just like any other element of the work. This phenomenon is enough in itself to show the necessity of a fundamentally genetic approach to the question of sources.
If the exogenetic element can efface its difference so completely, however, and allow itself to be absorbed so thoroughly by the endogenetic that it ends up by disappearing, is this not because the exogenetic procedure contains within itself the principle of its own effacement by writing? Rough drafts demonstrate perfectly how, even for the most realist of writers, the will to referential veracity remains a secondary necessity before the organic primacy of the work: selected first of all for its reality factor value, the exogenetic detail is forced into the original context of the rough draft; but as the endogenetic logic develops, the writer can be seen abandoning, one after the other, over the course of the composition, all the realist characteristics which had been the inital reason for choosing such a demonstrative detail. As it integrates better and better into its context, the detail sometimes ends up by becoming utterly unrealistic, even if it avoids quite simply running aground in a definitive shipwreck. This phenomenon leads one to question the very sense of “documentation” and exogenetics, whose real role might then be thought to be less to inform the act of writing than to offer it dialogic elements, which give a motivational and heuristic kick-start to the endogenetic process. More often than not, it is the relative heterogeneity of the exogenetic fragment, its foreign or antagonistic relationship to the writing, which is at the origin of the exogenetic process. Exogenetics can never be dissociated entirely from a ruse, through which writing manufactures the conditions of a productive confrontation in which it pits itself against the challenge of development in a hostile environment. Alterity develops conjunctive and transformational faculties in the writing: it provides favourable conditions for the mastering of an identity. Under cover of being documentation, exogenetics puts to work an intertext which often functions as a countertext with endogenetic value.
Logically speaking, there is no such thing as a purely exogenetic element: every exogenetic fragment bears the primitive seal of endogenetics, and the opposition of the two concepts is only relative. Whatever the exteriority of the initial referential source (physical place, real itinerary, overheard speech, a scientific article, a table or a previous literary work), its actual exteriority counts, from the beginning, far less than the action of motivated research and particular selection by which the writer builds a viewpoint and then makes withdrawals on the documentational stock. Sifting is an endogenetic framing mechanism which singles out of all the possibilities in the referential world open to observation, an appropriate element, that is, an element which writing has already decided it can make its own, which it can appropriate. By withdrawing the piece of information from its source, exogenetics rips the element from its referential context (and so transforms it), then textualizes it and puts it into some kind of order, in the form of a written fragment, so as to reinsert it into the endogenetic milieu, still flexible, but becoming more and more constraining, which will finally distill it down to match its own conditions. This process is so powerful and plays such a regulatory role in the compositional drafts that many writers, such as Flaubert himself, have made a work rule out of it: good documentary research cannot precede composition, it should, on the contrary, be hammered out in the heat of the endogenetic “forge”, so that observation, selection and recording can take place right from the start according to the precise endogenetic requirements, like a sort of prelimited writing in which the internal and creative extent of the note‑taking already takes precedence over the exteriority of the information to be gathered. But, of course, many authors possess an ur‑exogenetic set of materials, with all that that entails of the writer’s personal prerequisites, (notebooks of quotations, bibliographical references, leftover previous documentary files, etc) which make up a quantity, a sort of inital stock, very much prior to a compositional project but which can play a weighty role in the research stage and the budding conception of the project itself: it is the provisional exogenetic material, itself always already shot through with the hypothesis of a future endogenetic appropriation.

Outlining and Writability

The demarcation of the activity proper to each phase rests on a second conceptual opposition: writability and outlining. Writability [le scriptural] designates any writing process centred on verbalization and textualization, at whatever state of progress towards a polished lexical and syntactic form, but excluding all that is specifically directed at the organization or structuring of the narrative. Outlining [le scénarique], by contrast, designates any writing process focusing on planning, organization and structuring of the narrative, by large or small units, in the prospective form of workplans and general or partial outlines which will be put to use as plot-handling notes; it also designates the retrospective form of summaries or provisional assessments which can be used to appraise the compositional state of play. Writability falls in principle more under the jurisdiction of a microgenetic approach, outlining comes under a macrogenetic heading. It goes without saying, however, that these two concepts, useful for making out and specifying quite contrasting writing operations, are totally interdependent throughout the writer’s work, as he or she shuttles constantly between the two, perfoming reciprocal and successive reworking operations, uninterpretable except in terms of interaction: the structure modulates in response to textualization as much as textualization modifies under the constraints of restructuring.
The concepts of writability and outlining can be crossed with those of endo- and exogenetics, and each of the four outcomes can again be specified by the name of the phase to which they belong (or of the process in which they take part): thus, for a novel whose initial scenario has not been transferred from the structure of a previous work, the initial workplan could be defined as a “pre‑compositional endogenetic outline”. If it turns out, on the other hand, that the writer has borrowed the primitive workplan from a pre‑existing model, it would be better to talk in terms of a “pre‑compositional exogenetic outline”. The rough draft, in principle, comes under the heading of “compositional writable endogenetics”.

Operational Functions and Types of Documents

The second to last column of the typological table is devoted, still in the form of general model, to the operational functions which define, in terms of process or productive mechanism, the set of possibilities specific to each phase and, if need be, to the interior of each phase, to the “moments” or “steps” of this productive process15.
The final column, headed “Document Type” provides, in terms of descriptive categories, a detailed list of genetic documents which can be found for each of these moments or steps. These categories only concern documents that emerge strictly from the genetic domain (the collected documents, or dossier, of the genetics of the work)16, and only make up, in their attempt to be exhaustive, a series of strictly hypothetically possible documents. Most concrete collections will only comprise in reality a selection of these categories, the absent kinds corresponding either to documents which have existed but which have not been preserved, or to genetic operations which have not taken place. These descriptive categories are, on the other hand, dependent on a triple relativity which limits their field of usefulness to a zone circumscribed by history, genre and the genetic profile of the collected documents they are supposed to describe.

Historical Relativity

The descriptive categories apply to the types of documents to be seen in the period of around a century and a half which extends from the end of the first third of the 19th century to the 1980s17. Over this period, several of the concepts used here (like “scenario”, “sketch”, “roughing out” [“ébauche”] , etc.) were developed by writers to refer to their own work, with considerable variations in meaning according to auther and period. The term “scenario”, for example, obviously does not have the same meaning for Flaubert (who diverted the word from its theatrical sense to mean the detailed workplan of a narrative work18) as it would for a 20th century novelist for whom the term refers almost necessarily to cinematographic writing. But these historical variations of the genetic lexicon, whose impact should certainly be measured in the personal vocabulary of writers, remain quite easy to identify and do not seem to put obstacles in the way of establishing a general terminology, if every descriptive category used is given a clear definition, sharply delimited by its operational function and its specified adherence to one phase (endo- or exogenetic, writability or outlining, pre‑compositional, compositional and pre-publication). The historical evolution of of writing techniques between 1830 and 1980 (media, writing implements, inks, etc) also does not seem to pose an insurmountable difficulty to the listing out of genetic documents and operational functions19. The terminology used here to designate the different types of documents is therefore offered as valid right the way through the defined historical period, within the bounds of the inventory and definition.

Relativity of Genre

These descriptive categories concern the genetic documentation of a narrative work (novel, récit, novella, tale, legend, etc.) and could only be imperfectly applied, doubtless, to the genetic study of a dramatic or poetic text. Substantial modifications remain to be thought through for each genre: some categories wil become redundant, other concepts will have to be formulated, according to substitutions which might be made to chime in partially with the general schema of this typology.

Genetic Relativity

The descriptive categories serving to designate the different types of documents and operational functions, make up a general table which in principle, can be applied to the majority of collections of genetic materials, but on condition that each corpus be considered to constitute a particular interpretation of this chrono‑typology. The relative importance of the genetic steps is essentially variable according to writer. The table represents the moments of writing and the types of documents without marking any for special consideration: all the elements are given the same weight, although, of course, a writer’s work is individually characterized by the appearance into it of sizeable blocks of time which structure the genetic process in a unique way and create the image of a personalized procedure inventing its own habits and making to measure its own system of preferences from among all the possible writing media20. Each individual genetic process will follow writing’s virtual steps to its own rhythm and following a unique itinerary which in most cases will not take in all the documents grouped together here, and which could even, in places, upset the successive genetic operations as given here, by reversals, which it is impossible to symbolize in a two‑dimensional table. Finally (and this is the limit of all general typologies) some genetic material collections for novels, especially atypical ones, will prove uninterpretable in the terms of this chrono‑typology, unless it is borne in mind that this table can be used to measure typical differentials, in relation to an abstract model which does not claim in any way to reconstitute the norm but to organize a virtual structure.

Typology of Rough Drafts. Two Conceptual Fields: Rough Drafts of the Work and Compositional Rough Drafts

In this functional typology of genetic documents, the term rough draft appears by name (with the qualification “compositional rough drafts”) under the column‑heading “Document Type” but only in the last section of the compositional phase, corresponding to the operational function “textualization” and the stage “avant‑texte compositional process”. Nevertheless, as the two markers placed in the “Stage” column indicate, it must also be accepted that in the wider sense, the concept of Rough Drafts occupies a much larger place in the pre‑textual process and can serve as a generic term to designate a vast body of genetic documents: the zone which corresponds to the compositional work, stretching from initial workplans or scenarios to the definitive manuscript, and which can even include in certain cases, all or some of the manuscripts documenting the composition. Such an extension could seem excessive, but in fact corresponds to the usual definition of the word which was recalled right at the beginning of this study (“a work manuscript written with the intention of correcting it for use in the composition or definitive polishing of a text”), a definition which however vague it might be when brought in without any other specification, seems in fact to correspond to a certain reality simultaneously in the writer’s work and in the genetic analysis of documents. If I talk for example of the “rough drafts for L’Éducation sentimentale” by Flaubert, I mean by that a whole body of manuscripts: simultaneously, the roughouts and sketches of different episodes in the novel, the overarching scenarios which served to build up each of the parts of the narrative, the partial workplans for chapters, the thousands of pages of rough drafts that were actually used, in parallel to the workplans and scenarios, towards the progressive textualization of the story, the more or less corrected fair copies leading to the definitive manuscript and even the plot‑handling directions, research notes and reading notes, the little drawings and other notebooks hastily scribbled over during the composition to nourish or structure the fiction. In short, “Rough Drafts” is a good expression for a body of documents, which are by their very nature obviously rather diverse, but which are also equally interdependent, to the point of not being really interpretable in isolation, since they have only been produced by the writer in order to contribute, each in their own way, to one and the same organic compositional attempt. I think therefore, that we must resolutely look to save this habitual use of “rough drafts” in its wide sense, even in the act of loading it with a precise and contrasted typological signification intended make it understandable in real terms. It is necessary and even indispensable, since there is indeed a functional interdependence between the different elements which make up this vast genetic zone of composition. Among all these elements, however, one particular type of document is to be found which is also called, precisely, “rough draft”. We must start, then, by resolving the logical contradiction posed by its inclusion. The problem is not insurmountable. For the general concept, the first letters will be capitalized and the generic plural used: “the Rough Drafts”, that is the Rough Drafts of the work, sub-section of the “work’s manuscripts”. As for the concept of the rough draft, the object itself properly speaking, serving to define a compositional document which is a specific result of the textualization function, and which constitutes a writing step made between the initial scenarios and the definitive manuscript, all ambiguity can be waived by designating it and qualifying it by its operational function: “compositional rough draft” or “textualization rough draft”. Thus conceived both in the wide and in the specific sense, the typological space of the rough draft is a double one.

1  See note on the use of the French word avant‑texte, at the beginning of this edition of Yale French Studies.

2 Some manuscripts of Petrarch’s work, for example, as well as Montaigne’s Essais, Pascal’s Pensées, etc.

3 While remaining for the most part more attentive to Flaubertian genetic documentation, I have tried to integrate most of the obsevable mechanisms in the genetics of Balzac, Hugo, Zola and Stendhal (in the 19th century) and of Proust, Aragon, Sartre, Giono and Beckett (in the 20th century).

4 ‘Les phases de la genèse et la question de l’édition téléologique {‘Genetic Phases and the Question of Teleological Publication’}, paper given at the bilateral Franco‑Russian conference ‘Methodology in the Critical Edition’, at Pushkin House, Leningrad (Saint Petersburg), October, 1989 (papers published in Russian). Text reprinted in a simplified form as ‘Genetic Criticism’ in Introduction aux Méthodes critiques pour l’analyse littéraire (Introduction to Critical Methods for Literary Analysis), ed. by D. Bergez (Paris: Bordas, 1990), pp. 5‑40.

5 The practice of the editor requiring the author to sign to the definitive state of the changes to be made to what they have agreed between them as the final set of proofs, has a history, albeit relatively recent. That being said, for material reasons of lack of space, few of these documents seem to have been preserved. Editors’ archives, however, perhaps hold some wonderful suprises in store. In France, IMEC (the Institution for the Conservation of Contemporary Publishing) has set itself the task of safeguarding, inventorying, and making this type of document accessible.

6 Either by last minute corrections as the editor marks up the manuscript for the printer, or by intervening as the text is actually being manufactured: for example by means of instructions about the layout or typography handed to the editor, or even, later still, at the printers, by making adjustments to the material type‑setting itself, as numerous 16th century authors were in the habit of doing (a procedure which caused interesting cases of variants within the print run of a single edition). This is, of course, not to mention the set of authorial paratexts which the author must supply as quickly as possible to the editor, if it has not already been done: the preface, the final version of the title, the subtitle, the table of contents or the synopsis, the dedication, etc. The paratext can also be subject to the genetic approach.

7 The issue of pre-first publication editions by serial in the press comes up for a great many 19th and 20th century texts. It is far from simple to resolve, since, as serial‑novels of the 19th century, Balzac’s works, and the celebrated case of the Revue de Paris edition of Madame Bovary, admirably demonstrate, considerable variations can very frequently be noted, between the pre-first publication edition and first publication in book form. In certain cases, frequent in contemporary literature, the edition of the work in pre‑first publication form in the press can simultaneously constitute the conclusion by publication of one textual stage, and the starting‑point (pre‑compositional) of the avant‑texte of a much more significant new version, which will end in the publication of the work in book form: a published novella could for example turn into the initial scenario of a vast novel. But would we still then be talking about the same work?

8 Unless this rewriting ends in the publication of a radically different new work, nevertheless given the same title. Apart from these exceptional cases, the dilemma of how to mark the threshold, beyond which a variant version of the text has moved far enough away from its initial model to be evaluated as a different work, becomes a very legitimate one. Reciprocally, how far can the invariant text bear modification and still stay the same, without suffering an overall shift in identity? The novels of the Comédie humaine, among others, form an exemplary case for which to pose the problem, especially in terms of structural variation. A clarifying typological study of this subject should perhaps be undertaken, which would be capable of defining these transformations of the text in genetic terms.

9 Although, of course, some very famous cases do exist, in which the published versions of the text are explicitly presented as stages of a truly continuous rewriting process: Montaigne’s Essais, for example. In a less spectacular way, and with variable modalities, a great part of the literary production of the 16th and 17th centuries (for which we do not, in general, have access to any work manuscript) seems to demonstrate that writers used successive editions of their works as stages of rewriting, a phenomenon which validates a call for a true genetics of the printed text.

10 Such exceptions, nevertheless, are numerous and sometimes concern texts of primary importance: the text of Madame Bovary, for example, cannot be taken from the last edition of Flaubert’s lifetime, which gives an atypical and relatively retrograde state of it. The next‑to‑last edition must be chosen in preference.

11 Not yet properly speaking a text, and only becoming one through editorial intervention, sometimes going against what would have been the author’s wishes. Contemporaneous examples are numerous. This issue raises a legal problem, notably in terms of the abuse of moral rights of the rights-holder. The opposite problem is also raised, however: practically completed manuscripts might very well not be distributed as the author would have wished, because of the rights-holder obstruction. While judicial texts remain obscure as to these matters, in France, as in many countries, significant legislation nevertheless exists which allows the Lawlords to opt for fair solutions.

12 We owe these terms to Raymonde Debray‑Genette (‘Génétique et Poétique: le cas Flaubert’, in Essais de critique génétique (Paris: Textes et Manuscrits, Flammarion, 1979) {‘Genetics and Poetics: the Flaubert case study’, in Essays on Genetic Criticism (Paris: Textes et Manuscrits, Flammarion, 1979)}, one of the very first theoretical texts published on these issues). In spite of their relevance, these terms, proposed in relation to the study of Flaubert’s manuscripts, have hitherto not been assigned their rightful place in genetic criticism, perhaps for lack of having been assessed in the context of a definitional study both less specific and more systematic.  I am therefore taking the liberty of borrowing these terms, giving them a specific signification with a more general value.

13 The exogenetic quality of iconographical documents poses some difficulty. If it is a matter of a hand‑drawn sketch linked to a written transposition of it, there is no problem. But what if we are faced with a reproduction chosen and used by the writer with no traceable sign of direct transposition? The lithograph of Brueghel’s The Temptation of Saint Anthony hung on the wall by Flaubert when he was writing la Tentation… does that count as an exogenetic entity? Probably not. By the same token, do the photos of the Middle East that Flaubert found in scholarly works, and on which he based certain descriptions of Machærous, make up part of the Hérodias exogenetics? To stay true to our method, we ought doubtless to stick to the notes made about these photos and recorded in his notebook; although this need not prevent us from studying these notes in conjunction with the source documents. But what about a childhood photo, associated with autobiographical writing? And how should we treat a photo (or video film) made by the writer him or herself to document the composition? In how far are such images different in nature from a hand‑drawn sketch or drawing? And furthermore, how should we define the genetic status of a cassette recording made by the writer to serve in the composition? If it is an oral improvisation of the writing, a sort of autodictation recorded on a dictaphone, the material obviously has an endogenetic quality; if it is a recording of music, of oral testimony, or of an interview subsequently used by the writer in composition, the material seems exogenetic, but even then, should we not reserve this qualification for the written notes by which the sound data has been transposed or transcribed and sifted?

14 Exogenetics more or less signalled as such (through all the modalities of the quotation, the pastiche, the parody and through hypertexual mechanisms in general) represents an important dimension of literary writing which offers itself for interpretation in the form of the palimpsest, as Gérard Genette’s theoretical research has demonstrated. It is nevertheless true that the greater part of the genetic approach to such phenomena still needs to be worked out.

15 The term “moment” should designate the least finalized operational functions (associated with an objective that is still hypothetical or indeterminate) while the term “step” [étape] should be used to signal functions that concern a clear teleological procedure (associated with a determined outcome).

16 It will have been noticed that the typological table does not include elements which in other respects could have an absolutely essential genetic research value, but which do not belong properly speaking to the genetic domain: the writer’s correspondence, for example, which, as regards external criticism, constitutes an irreplaceable database for the dating and interpretation of the collected genetic papers; even if, as in Flaubert’s case, this correspondence offers a veritable compositional diary. The same goes for “attestations” and other documents whose role might be decisive but which fall outside the actual collected genetic materials. On the other hand, a request for information formulated in a letter by the writer to a friend or specialist, and by the same token, its reply in letter form, even when not handwritten, or any other non-handwritten note giving the writer information which will be transferred into the work’s toolbag, all these do constitute completely separate genetic elements.

17 From the 1830s, the date from which there are enough collections of genetic material to establish a typology, to the 1980s where the spread of personal computers, word‑processing and desktop publishing seems to have modified the attitude of certain writers and transformed several aspects of the writing process.

18 The term scénario is not recorded in Littré and many other dictionaries of the 19th century. The encyclopaedic Larousse gives: “Scenario: theatrical vocabulary direction, in metaphorical sense: ways in which one prepares to trick, seduce, win”. Scénario was also used to speak of the action or the development of a play, notably in a written form. It is probably on the basis of this usage that Flaubert forged the personal sense that he gave the word for narrative writing: the scenario (initial detailed workplan) is in opposition to the scratching [pioche] (composition: as much the action – the act of composing – as its result: manuscripts snarled up in crossingsout). This opposition was already to be found in Balzac, in 1837, still denoting theatrical writing, in a passage from les Employés: “An author of plays […] is made up first of all of a man of ideas, whose job it is to find subjects and build the scaffolding or scenario of a piece of vaudeville; and then a scratcher [piocheur] whose job it is to compose the play”. The terms “pioche” and “piocheurs”seem themselves derived from the vocabulary of sculpture where they were used to refer to the initial work of sanding down; but a metaphorical reference should no doubt also be acknowledged to the vocabulary of the penal colonies. A fine lexicological study is waiting to be done on the history and transfer of metaphors in the genetic lexicon of writers and artists.

19 The steel nib progressively replaces the quill pen, and causes faint changes to the way manuscripts look, and later, towards the end of the 19th century, typing starts to make its presence felt and takes over, for certain writers, from the former handwritten fair copies, definitive manuscripts and copier’s manuscript, but does not, it seems, bring with it noticeable revolutions for operational functions in the process of writing, nor for the types of document produced by the authors.

20 Proust used his notebooks for very different purposes from Flaubert; Stendhal, who wrote such prodigious quantities of marginal directions, avoided making outlines, the idea of a workplan or a scenario seeming to him to put an obstacle in the way of his imaginative faculties; as for Balzac, he had, as we know, a personal method of composition which, by the correction of successive printed versions, led him to rework “rough drafts”, in a way quite atypical for his time, as well as to follow a developmental model relying rather more on outlines; Aragon, according to him, condensed his outlining effort to the very first words of the project; Beckett used typed‑up copies as corrected fair copies which could end up coming before new entirely handwritten versions; as for Giono, he gave a decisive importance to the initial choice of title, whose enigmatic formulation was supposed to serve as the regulatory concept for the structuring of the plot and the development of the composition, etc.