Take a simple woollen sock: it is liable to become worn and a hole will appear, at the heel or near the toes. This annoying imperfection can easily be mended with needle and thread. Sigmund Freud, in his 1915 article on ‘The Unconscious’, shows that for an obsessional neurotic the sock itself can be considered as a hole that is filled by the insertion of the foot (with all the connotations that you can imagine). But for a schizophrenic patient, pulling on the sock results in stretching the knit and opening the meshes and creating, or rather revealing, a multiplicity of holes that we cannot hope to fill, since they are built into the very fabric of the sock. We can darn the sock; we can fetishistically obturate the hole at the core of the sock itself, but we can never get rid of the myriad gaps created by the mesh of the knitted fabric.
In 1844, Alexandre Dumas (and Auguste Maquet, the ghostwriter who was drafting a large proportion of the novels that appeared under Dumas’s name) wrote to the editor of the paper that was serialising their joint work, to ask him to mark the proofs for them, authorising him to correct what Dumas calls ‘les niaiseries courantes’ (‘the usual blunders’) (Mombert, 2022, p. 38). In the light of this, it would be understandable that a textual scholar should consider that they are the last link in a chain of collaboration, from Maquet to Dumas, from Dumas to the editor of the paper, from the editor to the textual critic, and that as such he or she is entitled to correct what he or she considers to be imperfections. A paradoxical authorisation based on a conception of shared auctoriality.
One could think that this sort of attitude, this delegation of authority, this indifference to the perfection of the text, is characteristic of a serial writer like Dumas (who was writing novels and plays on an industrial basis, with the help of a team of hacks), but it is well known that William Butler Yeats (Yeats, the Modernist poet!) was happy to delegate his authority on such matters, and considered that the editors at Macmillan understood punctuation much better than he did.
It is clear that the proofs, which are supposed to be the field where the final imperfections of the text are eradicated, will have a very different aspect, and a very different value from a genetic point of view, in the cases we have just mentioned (Dumas or Yeats) and in the case of writers like Honoré de Balzac or James Joyce, who made a very different use of proofs, adding a very large proportion of the final text at this stage. These writers used the proofs to correct a much less superficial kind of imperfection: a radical incompleteness as opposed to a superficial defectiveness.
We will soon come back to Joyce’s proofs. But let us first have a look at another example. In 1986, a ‘New, Corrected Edition’ of William Faulkner’s Absalom, Absalom! was published. The various emendations introduced by the editor, Noël Polk, are listed in an apparatus at the end of the volume. Some of the emendations are motivated by a reference to the manuscript, or the typescript, considered to have greater authority than the printed text of the first edition. But some emendations are not explicitly motivated. They correct another kind of imperfection, a departure from an implicit rule that Faulkner failed to follow. For instance, an open parenthesis must be closed, and on this basis, a closing parenthesis is added (although some might think that there is something very Faulknerian in a parenthesis that is never closed).
Some of the corrections deal with the chronology that Faulkner appended to his book, ostensibly to make things easier for the reader (and this is certainly not superfluous). The emendations introduced by the editor are based on inconsistencies: the date of Rosa Coldfield and Quentin Compson’s visit to Sutpen’s Hundred is indicated as 1910, while in the book, it happens in 1909. This inconsistency is not the result of a transmission error: it is Faulkner’s error.
We can understand how it occurred. The published chronology was based on a private manuscript chronology, a document that Faulkner had used for the writing of the novel. We can imagine that when he drew it, he had forgotten that in TheSound and the Fury Quentin commits suicide in June 1910, so that the visit in the fall of 1910 would occur after his death. He later made the change in the novel, but did not go back to correct his working timeline, which he later used as a basis for the printed chronology that was appended to the first edition of Absalom, Absalom!, so that the inconsistency is caused by a discrepancy between different stages of the writing process.
This was an imperfection that seemed easy to mend. The editor considered that the novel had more authority than the chronology and adjusted the latter to the former. A similar improvement had, however, been rejected by Faulkner himself when he was made aware of the inconsistencies between TheSound and the Fury and the ‘Compson Appendix’ that he had written to help the reader through the complexities of that novel: ‘The inconsistencies in the appendix prove that to me the book is still alive after 15 years, and being still alive is growing, changing; the appendix was done in the same heat as the book, even though 15 years later, and so it is the book itself which is inconsistent, not the appendix’ (Cowley, 1966, p. 90).
In an article that I published a long time ago (Ferrer, 1989), I tried to show that the chronological oversight makes sense in a Faulknerian context: the fact that the story should be told from a perspective situated beyond the death of the narrator has precedents in Faulkner’s work and is suggested explicitly within Absalom, Absalom! – but my present point is more general.
There are two different kinds of imperfections involved here: diegetic inconsistencies and genetic inconsistencies. The aim of the corrections applied by the editor to the chronology seems to be purely pragmatic: helping the reader to get along in the maze of the novel. But the underlying assumptions are far-reaching: they imply that there is such a thing as a recoverable story, outside the text and somehow independent of the text. One might argue that such an assumption is inherent in the idea of writing a chronology. But this may be precisely what Faulkner didn’t like about the idea; this may be precisely why he built, consciously or unconsciously, a less than perfect external temporal structure (it is said that carpet weavers always introduce a deliberate imperfection in their design, for fear that their soul should become enmeshed in it). By introducing these imperfections, Faulkner was suggesting that the chronology was just presenting another version of the elusive story that had been trying to tell itself through the pages of the novel. Eliminating the imperfections, on the other hand, means that beyond Faulkner’s fancy narrative tricks, which finally backfired and confused the author himself, there are some hard facts for the lucid reader to grasp. It postulates a diegetic universe outside of the text, more important than the text and in relation to which the text can (and should) be emended.
Such a position can lead to absurdities, such as the one we find in Danis Rose’s edition of Ulysses, which goes so far as to correct Joyce’s facts in relation to geographic realities. In the ‘Ithaca’ episode, Bloom admires water, in particular ‘its unplumbed profundity in the Sundam trench of the Pacific exceeding 8000 fathoms’ (1984, episode 17, line 187). Rose corrects this to ‘its unplumbed profundity in the Marianne Trench of the Pacific, exceeding 6000 fathoms’ (1977, p. 585)! It is true that Joyce wished that the proofs of ‘Ithaca’ should be read by a competent scientist,1 but we don’t know how far Joyce would have accepted his suggestions, and we don’t know if the word ‘Sundam’, or the number 8000, were not important to him for reasons of sonority, of numerology or other motivations that we cannot fathom.
It is true, more generally, that, when he was writing Ulysses, Joyce took great pains to verify the conformity of his book to the reality of Dublin, but inside the book, he gives us a drastic warning against the referential illusion and the search for coherence: the narrator of half of the ‘Cyclops’ episode is an intradiegetic narrator, a very lively, foul-mouthed debt collector, but he never reappears elsewhere in the novel, outside of his own narration, except in the ‘Circe’ episode, which is a kind of recapitulation. When he reappears there, he is nameless and featureless:
(A panel of fog rolls back rapidly, revealing rapidly in the jurybox the faces of Martin Cunningham, foreman, silkhatted, Jack Power, Simon Dedalus, Tom Kernan, Ned Lambert, John Henry Menton, Myles Crawford, Lenehan, Paddy Leonard, Nosey Flynn, M’Coy and the featureless face of a Nameless One.)
THE NAMELESS ONE
Bareback riding. Weight for age. Gob, he organised her.THE JURORS
(all their heads turned to his voice)
Really?THE NAMELESS ONE
(snarls)
Arse over tip. Hundred shillings to five. [1984, episode 15, lines 1139–49]
We recognise his way of speaking, but he has no name and he has no face, and this state of fact is perfectly logical in the context of the book. He has never been presented from the outside, therefore, narratively speaking, he has no features! He has not been called by his name; therefore, he has no name! This character is abruptly reduced to a ghostly presence by this sudden refusal to go on playing the usual game of inference: a discourse implies a voice, a subject, a human being… We are brutally reminded that characters are not all-round living beings but literary artefacts, temporary clusters of linguistic elements, structurally incomplete.
Should an editor erase this glaring imperfection and provide this entity with a name and a face, under the pretext that there was no such thing as nameless and faceless characters in Dublin? So far, no one has suggested that we should.
Even a naturalistic writer like Émile Zola does not hesitate to sacrifice the diegetic coherence of his Rougon-Maquard saga in the interest of narrative construction: whereas Gervaise Coupeau in L’Assommoir has two sons, suddenly in La Bête humaine she has three, because a new character is needed for the development of the story. Zola corrected retrospectively the printed genealogical tree of the saga (equivalent to Faulkner’s appendicular material), but not the text of L’Assommoir, so that Faulkner’s phrase is literally true in this case: it is the original novel that is inconsistent (that has become inconsistent with the general design of the cycle).
There is another kind of imperfection that an editor feels perfectly legitimate to correct: when it can be proved that circumstances have thwarted the author’s intentions so that the text does not reflect those intentions.
But the problem is that when you look closely into the genesis of any work, you find that intentions are multiple and contradictory. Let us take one very peculiar example: in the course of writing Ulysses, Joyce inserted a passage on a set of proofs that arrived too late at the printer’s (when the printer received the proofs, the type for this section of the text had already been reset). But Joyce did not want to waste that passage, so he inserted it, with a few modifications, in another place (episode 15, lines 4506–09).
In the 1984 corrected edition, the addition was incorporated in the place where it had originally been intended (episode 15, lines 1914–17).2 But this reinsertion created an interesting but completely unintended echo with the almost identical passage at the end of the chapter. In the 1986 revised version of the corrected edition, the editors considered that the mid-January 1922 insertion cancelled the indication on the mid-December 1921 proofs, so that only the second passage was kept. Another, drastic, solution would have been to reinstate the passage in its original place and to excise the other one, on the grounds that it was only inserted at the end of the chapter as a second-best choice. This would not be a worse compromise than the other solutions, except that we would lose the changes introduced in the second incarnation of the piece, and, of course, the coherence of that part of the text would be destroyed.
Admittedly, this is an exceptional case, but my point is that, because no work is done instantaneously, in one single moment, imperfections (corresponding to the interference of different layers of intention) are built into the fabric of the text. According to Michael Riffaterre, the text always implies the absent intertext in its structure (for him, this incompleteness is the defining characteristic of the literary text). The interference between the two systems causes a perturbation in the text (which Riffaterre calls ‘agrammaticality’ or ‘catachresis’) introduced by the presence in the text of the ‘connector’, the extraneous element issuing from the intertext (1980; 1981). These agrammaticalities can be anything that is somewhat out of place or unforeseeable in the new context. The connector belongs both to the text and to the intertext: it is grammatical in the intertext, but when it is imported into the text, it stands out as agrammatical. In the same way, each state of the text is built with fossil remnants of the previous states. They are reinterpreted in the context of the current state, but they remain in some respects agrammatical in this new context, in the same way that the date of Quentin Compson’s visit to Sutpen’s Hundred is inconsistent in the context of the revised chronology. So that no text (except perhaps if it is created instantaneously, like an extemporised haiku) can be immune from that form of genetic imperfection.