In conversation at Zuerich a few years ago, Hans Gabler said "Of course, the Archive has to be completely redone". I have been stewing about that ever since. Was he right? What follows is entirely personal reflections from using it for a dozen years or so.

First, it's obvious that the Archive is a marvelous tool for research into the development of the Wake's text. When the Archive came into print, it became possible for us to start tracing the details of that development without having to go to Buffalo (for the Notebooks, which I know little about and will not deal with here) or to London (for the drafts, typescripts and proofs). No one would have said then, or would say now, that it became possible to do authoritative work with the documents without eventually staring them in the face, but still an enormous amount of the information in them can be gleaned from attention to their Archive reproductions in our own libraries or studies and on our own often necessarily fragmented schedules, in preparation for the final attentions to the things themselves-or, of course, in simple cases, not for preparation but for answers we needn't doubt. We are likely to ask "Can Joyce really have meant 'Wirrgelling and maries?' at FW: 88.33-4?", and it's easy to discover in the Archive that no, a typesetter's line-skipping error which never got caught cut the heart out of it: Joyce intended "Wirrgeling and boeuffickly bucephull. Wheateared, however, and with fallen mammaries?" (BL 47475-216; JJA 46: 208).

But it doesn't take much more than a glance to realize that the Archive isn't only a tool for research, but is in a crucial way research itself. If we did just go to London with the misguided idea of skipping the Archive stage entirely, as if the Archive had not yet been published and we were in A. Walton Litz's shoes preparing The Art of James Joyce or David Hayman's starting to work on the First-Draft Version, we would find at St. Pancras, as they did in Great Russell Street, the handsome black volumes such as BL Add. MS 47472 (and the two important "red-backed Notebooks", A and B of BL Add. MS 47471), in which are bound up together the multitude of pieces of paper which Harriet Shaw Weaver had arranged for the British Museum when she passed over to its final home the archive (with a small a) with which Joyce had entrusted her over the years. This arrangement is, understandably enough, what we might call documentcentric: Joyce revised, for instance, the typescript of his early drafts of chapters 2, 3 and 4 (as they then existed, of course), and sent it on to Miss Weaver when he was through with it, that is, after he had fair copied it so as to relieve the next typist of the intolerable burden of interpreting his chicken scratches (carefully indexed though they are). And this document is bound together as such, in BL 47472-139-62 (though two pages of unrelated typescript, fs. 142 and 149, also made their way into it).

But the Archive's genius was to realize that the document is not what we are most likely to be interested in (and if we are, we can trace it in Michael Groden's admirable Index, in which the documents in the BL are referred to placement(s) in the Archive), but rather the growth of the book Finnegans Wake. Rather than primarily a reproduction of the documents, the Archive is a proposition of the Wake's genetics! Its principle of organization is not "Put all the pages of a document together", but "Discriminate the stages and sub-stages into which Joyce, in his methodical and predominately linear fashion, should be seen as having divided his process of moving from first draft through multiple revisions of each of the chapters and sub-chapters [or "Sections", as the Archive calls them, though that is a trifle misleading since the printed book acknowledges no such subdivisions] of what would become his printed book." What the Archive correctly foresaw was that its ordinary users (if this is not an oxymoron) would not want analysis and interpretation of that revised typescript as such, but rather would want to trace the growth of, say, chapter 3 of the Wake-and therefore the pages of that typescript which constitute a distinct stage (revision level 3 in fact, of chapter 3) are reproduced in an Archive subsection along with the fair copy which underlay that typescript (level 2) and the fair copy made from its revisions (level 4). If document pages overlap the Wake chapters (as it so happens that BL 47472-156 does contain both the end of what would become chapter 3 and the beginning of chapter 4), why then BL 47472-156 is simply printed as well in the appropriate place for chapter 4. We find it both in the chapter 3 volume (at JJA 45: 199) and the chapter 4 volume (at JJA 46: 31). Just by the way it prints the document pages, then, the Archive is a kind of essay on the growth of the chapters of the Wake. And that, not just a bunch of photoreproductions of what's in the BL, was exactly what we needed when the Archive was printed and what we need now-though since reading that "essay" is so hard, given Joyce's handwriting, we also need transcription of the pages on which that "essay by organization" is based, a lack which is only gradually starting to be filled. When we go to St. Pancras for our final discoveries about the document which the Archive has left unrevealed, we can go armed with the organization the Archive has given us and into which those discoveries will fit. For this, Archive, much thanks!

Consistency of Designation

Like any such organization, the Archive's nice simple principles of discrimination do not just perfectly, or at least incontestably, fit the details of Joyce's revision process–the revision levels are not always quite as distinct as they must necessarily be made to seem in the process of giving them numbers and laying them out end to end. But let's be clear that this principle of neatly demarcating levels is a surprisingly good match for Joyce's practice, at least in the chapters, two through four, I know most about. Joyce usually does draft a unit (though often each unit does not wind up being a chapter, but either becomes a part of a chapter when it is later combined with other draft units, or is divided eventually into parts of several chapters) and then revise it by a series of expansions rather than by mining it for shuffleable parts. This enables the Archive strategy of labelling the units in terms of their final form: such and such a document is a first draft "of" such and such a "Section of" such and such a chapter (though of course it was not drafted as a chapter or chapter part, but simply as a unit whose specific coverage and limits would only come clear later). Joyce's basic (and astonishingly consistent, all in all) practice was to take the document embodying each unit, whatever its limits at a given point in the revision chain might be, and thoroughly "correct" it (as he always said, meaning to expand and enrich it) either by copying it as a new document (with its own revisions) or by modifying a copy (typescript or proof, made by someone else) of the revisions made to that unit up to that point. Each of the segments of the documents in which this work was done which correspond to chapters or parts of chapters of the finished book can thus be called a "revision level of" that chapter or parts of a chapter. These levels the Archive numbers, starting with zero for the first draft (since after all it is not a revision of anything). In chapters two through four, conveniently, these whole-numbered revision levels tend to apply to the same describable stages of the revision process. In the three chapters, Joyce revised through much the same stages: first drafts (level 0) got an initial revision (level 1) which got a fair copy (level 2) which Joyce caused to be typed and which he then revised (level 3) so extensively that he needed to make a fair copy (level 4) for a typist whose typescript he could revise (level 5) and then have retyped and rerevise (level 6) for the typesetter, whose proofs of a transition "Work in Progress" segment he revised (level 7) as he then many years later revised the published transition pages (level 8). The level 8 revisions were the basis for Finnegans Wake galley proofs, three sets of which he revised in turn (levels 9 through 11) and from which were struck page proofs (two sets, levels 12 and 13, only recently rediscovered) of the book itself, post-publication minor revisions to which Joyce made as level 14.

In a stroke of luck for the Archive user's memory, exceptions to this whole-number/description abstraction cancel each other out in chapters two through four, though it required an inconsistency in the Archive numbering (almost certainly not with this intention) to manage it. Chapter 4 is the only one of the three in which Joyce's retypings and recombinations of level 5 warrant a separate level 6 designation-or may warrant one, as we will see below. And since there were perhaps at least two successive transition galleys for him to work on for each chapter it would seem the numbers would fall out of phase: in chapters 2 and 3 Joyce made only a few corrections on a first set of galleys and then extensive revisions on a second set (though the second set for chapter 2 is missing, its content inferrable from transition 2 as set). In chapter 4 the only surviving set of galleys is heavily revised, like the second sets of chapters 2 and 3, and there is no clear evidence that the equivalent to the first set of light corrections was made, though a letter of Joyce's to Harriet Shaw Weaver of 20 May 1927 (Letters , 253) discusses with her what should be done with two different copies of the proofs (not necessarily successive sets but perhaps just different strikings of the same set). The Archive Preface (JJA 46: xii) does claim that "Another, cleaner set (missing) with a slightly different overlay was sent to the printer", though I myself think there is little evidence for this claim and a lot against it. You would think, then, that the Archive should, since the level 6 of chapter 4 is a second typescript which chapters 2 and 3 lack, call its supposedly missing second galley level 8, just as it labels the missing second galley of chapter 2 as level 7, and that we would thus be out of phase: the revised transition pages of chapters 2 and 3 should be level 8 as they are, but the same pages for chapter 4 should then be level 9. But although the Preface to volume 46 makes the claim for existence of this missing set of galleys, the "Specifics" page (JJA 46: 129) for the section of documents in the volume where this would ordinarily be given its number (as is the missing level 6 of chapter 2 in JJA 45: 87) makes no mention of it. So we lucky users can continue with a fairly good conscience to think of levels 8 through 14 as consistent across all three chapters, and need only bear in mind that level 6 is different in chapter 4 from the other two.

This is not the only instance of discrepancy between claims made in the Prefaces to the volumes and those made in the specific descriptions of the Archive sections in each volume. One reason is that different people often did the different writing tasks, and disagreement on inferences may have been either unnoticed or unresolvable. But then, too, those tasks were different, and in one aspect these inconsistencies might be considered desirable and useful consequences of that difference. Even though we can say that the Archive organization itself is a sort of essay on the Wake's genetics, it is embodied in an organization of reproductions of pieces of paper whose differences, each from the prior ones in the revision chain, constitute that genetics. It necessarily focusses on the positive, so to speak, emphasizing what we still have that we can take pictures of, and only including reference to what we lack when there is clear evidence of extensive revision on a document we have lost (that is, no effort is made to guess how many carbons of typescripts may have existed upon which no revisions were made, or how many unrevised copies of proofs were struck). The Prefaces, on the other hand, really are essays, discussions of what there is and (to a limited extent) what there may have been, allowing more extensive speculation about the possibly once existent pieces of paper and even telephone calls and conversations which might, if we could reconstruct them, explain some (though surely not all) of the odd rifts and inconsistencies in the concrete set of differences among documents. The Prefaces talk about the Wake's genetics, the Archive reproductions organize and represent it.

But that is not to go back on the initial claim that the Archive ordering is an abstraction of the genetics. As we said, this is not just a concatenation of pictures of pieces of paper, but a discriminating catalog of their relationships to each other and to the process of making the Wake. And what we can recapture of the inevitable messiness of a process of revision involving the shuffling of various pieces of paper sometimes cuts against the specification of sets of those pieces of paper as discrete successive levels. To look again at the troublesome levels 5 and 6 of chapter 4, the typings and retypings of the level 4 fair copy in preparation for the transition 4 typesetter: A good typescript (level 5) was made and much revised in – as the Archive Preface puts it – "three inks and a number of hands (including that of at least one amanuensis)" who is almost certainly Giorgio Joyce. As it got to seem less and less imaginable that the scribbled-over pages could be models for a typesetter, many of them were retyped and the retypings were also much revised, with the retypings themselves sometimes retyped and revised in turn. As the Archive says, "[i]n the end Joyce was obliged to assemble a new typescript from retyped and original pages," and this the Archive labels level 6 and prints together, original level 5 pages re-reproduced in their appropriate places among the level 6 retypings. This is, I agree, the best way to lead the Archive reader through the revision levels, but its abstraction does mislead the reader about the linearity of the process and it does violate the Archive's own admirably specific (and therefore critiqueable!) definitions of its level and sublevel assignments: in other words, the abstraction of the analysis into levels 5 and 6 is undercut by the bundling of representations of the pieces of paper involved (the achievement connoted by the word "archive").

First, as to linearity: very often in these pages we will find on the level 6 retypings overlays in Joyce's hand (usually in red ink but sometimes in black, re-emphasizing how the revisions were made in multiple passes) which seem to be copyings of overlay we can "already" observe on the level 5 pages but which were not, like most of the level 5 overlays, incorporated in the level 6 retyping. But many (though not all) of these overlays on level 5, which sometimes differ significantly from their "copies" on level 6, are in green ink and Giorgio's hand, to which we have been signalled by the HSW letter. Not hard: after the retypings Joyce went over them and, incapable of leaving well enough alone, made revisions on the retypings which he asked Giorgio to "write back" to the level 5 pages (and sometimes which he "wrote back" himself), to try to keep the two sets squared up. Therefore, treating the reproduction of pages as a proposition about the discreteness and time-directional linearity of revision levels necessitates, in this very unusual set of cases, falsification of the labelling. We find the actual later revision, as drafted by Joyce on the "level 6" page, copied on the supposedly earlier level 5 typescript.

But the (again, quite helpful) Archive effort to abstract into two levels the concrete pieces of paper which took as input the fair copy level 4 and gave as output the model for the transition 4 typesetter entails a basic contradiction to the Archive's definitions of levels and sublevels. The "Draft Analysis: Reader's Guide" in each Archive volume says that "Each draft stage [what I prefer to call "revision level"] is given a numerical index or codenumber (0, 1, 2, 3, and so on) to correspond to its relative position in the total range of drafts which together carry the 'continuous copy' of a Section [of a chapter]", and specifies concerning the "further segmentation of the major draft stages into their secondary stages" that "draft stage n+ indicates redrafted pages of draft n (where all of draft n is redrafted, this stage is given as n+1). . . . By redrafted pages is meant pages either retyped to allow for further overlay or reinscribed to permit a clearer fair copy." (Further redrafts of n+ are indicated by n‡, n triple dagger, etc.) Under these definitions, then, the retypings of level 5 ought to have been called level 6 only if the whole of level 5 had been retyped, rather than selected pages being retyped (often spilling over into a second retyped page, since there were additions to increase the length beyond the page limit). Those retyped pages should not be level 6 but level 5+ (and/or 5‡, etc.), and indeed David Hayman's earlier "Draft Catalogue" in his First-Draft Version (a table of correspondences to the appropriate section of which is given in each Archive volume) does consider the totality of these pages as one level, designated in his system (which begins with 1 rather than 0) "I, iv [6]", though also describing it as including "Second typescript, a complex unit prepared for the printer of Transition [sic] 4".

I would prefer the Archive to be consistent with its own definitions, but aside from that minor twitch toward good order I can appreciate the decision to articulate two levels. Insofar as the Archive is an implicit essay on genetics, it makes sense to regularize a messy situation like the multiple retypings of level 5 in the direction of the Archive's main, and mainly correct, implicit assertion: that Joyce worked by firming up the sections of each chapter in a series of distinct shapings by revision, moving to the next level only when the prior level was essentially complete and ready to be bundled as free-standing input to the next stage which was to copy (and then revise) it. It is always astonishing how few loose ends he left at every level (though there are usually a few-spaces left in which a needed word could be crafted later, etc.), and there is no doubt that level 5 of chapter 4 was whole, was revised piecemeal, and then was scavenged for pages to become part of a new whole, which it makes every sense to call level 6 since it is a whole version of a whole chapter which was to be the basis for a new stage of exactly the same thing.

One particularly problematic case shows the dangers of departure from the proper designation of retyped single pages, though. This involves what was at levels 5 and 6 the last page of chapter 4, on which now appears for the first time the poem , "Sold him her lease of ninenineninetee," which thereafter until level 8 ended the chapter (as the Rann ends chapter 2). What happened was that as usual Joyce importantly revised the level 5 page (BL 47472-289; JJA 46:111), adding to the appeal to ALP for "mercy of thy balmheartsyheat" the passage making clearer that we are about to turn our attention to this "little lady waiting and her name is A.L.P." This short page was retyped almost perfectly – the only variations are the silent correction of a punctuation error introduced by level 5 indexing and the dropping of the second hyphen from "Dropping-with-Sweat" – but though two carbons of this new level 5+ survive (47472-290, 291; JJA 46: 126-7) the ribbon copy is lost. The Archive asserts in its "Specifics" page (JJA 46: 87) that the missing ribbon of level 5+ also "includ[ed] as an addition the holograph version of the concluding poem", and whether that is true, or the poem was drafted elsewhere, a new page ending with the poem was typed, in ribbon (BL 47472-314; JJA 46: 159) and (at least) one carbon (BL 47472-292; JJA 46: 128). The ribbon copy, by the basic Archive rule for retypings, should be called level 5‡ (though it is bundled into the packet labelled level 6), and it differs from its model level 5+ only in the inclusion of the poem and the minor changes that that "areyou looking for Pearlfor" becomes "areyou-lookingfor-Pearlfor" and "Wery wee wight" becomes "Wery weeny wight", the last at least surely by Joyce's instruction in the typing process. As we have seen often happened, Joyce drafted further overlay on the ribbon copy. He added the passage "He spenth his strenth amok haremscarems . . .agelong pine tomauranna" (FW: 102.25-29) which somewhat awkwardly interrupts the smoothly flowing change of attention toward ALP. He made (at a different time and in a different hand) a little rearrangement of the end of the poem so that it does not end with "Attabom, attabom, attabombomboom!" as it first appears. And he also made two minor corrections, "Sold" for the mistyping "Told", and just the different word choice of "Then" for "For". Joyce himself copied the "Attabom" change to the carbon as well (in fact, we cannot say he did not draft it on the carbon), but Giorgio in the characteristic "writing back" process copied the others (perfectly accurately, for once, though of course omitting Joyce's intradraft rejections and alterations), in his distinctive green ink, to the carbon, which by the Archive rules should be 5‡, since "draft stage n indicates pages with identical textual substratum to those used at draft stage n (such as two copies of the same issue of transition) but with different overlay" (JJA 46: xxiii). Then, after Giorgio's copy back to the carbon, Joyce made on the ribbon copy two changes in the names of the haremscarems: "Yalva" to "Marinka" and "Indra" to "Anileen" (though there's a little more to be said about the second shortly).

But with what actually happened in mind, let's return to the Archive arrangement. There we find the carbon first, labelled as 5‡ as if it were itself a retyping of its model 5+ rather than a carbon copy of the ribbon which actually is such a retyping, and we find that ribbon copy labelled as level 6 and printed with all the other re-representations of re-used level 5 pages and representations of the ultimate retypings of previous typings (whether ribbon or carbon) which went on to the transition typesetter. But despite the extraordinary oddity, in terms of the production process and the definitions specifying it, of calling a ribbon typescript a retyping of its own carbon, as the sequence 5‡-6 here demands, it must be that the Archive is primarily concerned to track the sequence of revisions which are passed on (or not) in chain-i.e., the genetics! And despite the oddity, stemming from the writing back, of its not being Giorgio's cleaned-up copy on the 5‡ page but Joyce's messier draft (same text both times, though) which gets passed on as to the transition typesetter, the revisions on the level 6 page are a later stage than those on 5‡, if only in the fact that it was after the copy to 5‡ that Joyce changed "Yalva" to "Marinka", which we can observe to have been followed properly by the level 7 typesetter .

But imagine our surprise, then, to find that the other difference, "Indra" becoming "Anileen", is not followed by that typesetter, but that that change is overlayed on the level 7 galley (JJA46:177). But again, not hard: this is in its turn a "writing back"-some time during the proofreading process Joyce made the change on the galley (it is in his same hand as all the other changes on that page) and then he or someone else pencilled it back in to the level 6 page, though not quite perfectly since on the galley it is "Indra and" which is crossed out and ",Anileen," which is substituted, whereas the copy incorrrectly leaves the "and" in place. Again, then, the later change, what is properly passed on to the correcting typesetter for the transition issue itself, appears first in the Archive on the earlier Archive page, with this time the apparently earlier version not, as with Giorgio's copy of the level 6 draft to level 5‡, more perfect than its draft but apparently corrected by it.

There are other difficulties with the many typings and retypings of levels 5 and 6 and their sublevels. But if the Archive seems to have sacrificed with level 6 some of the logic of designating retypings with the series +, ‡, etc., for the sake of holding together as full revision stages the pages in which " all of draft n is redrafted, . . . [which] stage is given as n+1", quite the opposite solution is reached with the revisions Joyce made to printed pages of transition and passed on to the Wake typesetter. Here the priority seems to have been to emphasize the close connection of the pages in the production sequence, at the cost of losing the fact that " all of draft n is redrafted" and should therefore be called "n+1".

Joyce in the 1930's heavily revised the transition pages published in 1927 (and which after publication had had just a few revisions inscribed on them already), so heavily in fact that even more than with the level 5 revisions and retypings it was perfectly evidently necessary to transcribe these to another set of transition pages. The pages of the first set of all three chapters can, we have seen, quite properly be designated as revision level 8. The transcription to the second set was done a little differently from the retypings of (some of) the level 5 pages, in that rather than writing the overlays from level 8 onto the new transition pages, on those pages index marks are put which key to pages of typescript (manifestly made with Joyce "at the typist's shoulder," so to speak, dictating differences from and solutions to the difficulties of the level 8 inscriptions). Of course, Joyce being Joyce, new overlays are also drafted right on the new transition pages and indeed also on the typescript pages, and these sets of "tandem pages", as I think of them, are sent to the typesetter, who follows the indices on the transition pages where they lead, whether to the typescript (and some holograph) lists of overlays or to the margins with new holograph overlay.

Now it makes every sense to print, as the Archive does, the typescript pages and their indexing pages together as a single unit, though I think that the revisions on the typescript pages should be distinguished by a "T" designation: 8+T. But by the logic followed with levels 5 and 6, and indeed by the definitions of the labelling system, surely the unit 8+-8+T should get an "n+1" designation, and should be revision level 9, no? But what it is called is level 8+. I think this decision is a mistake in judgment, unlike the decision for a separate level 6. But again I can imagine the reasoning. Unlike the relation between, say, the first and second galley proofs (levels 6 and 7) of transition 3, or between the two different settings of the Wake page proofs, the relation between the two sets of transition pages is not that of a resetting-the level 8+ pages are just another copy of the level 8 pages. Partly because of this, the connection among the level 8 revisions, their indices on the second set, and their indexed transcriptions are so tight that to think of them as comprising one major stage is appealing. But then, there are no two sequential revision stages which don't show that same tight connection-a full set of revisions is by one method or another transcribed (often with changes) to a set of pages on which much new revision is made. That will do as well as a description of the relation between a typescript and its fair copy, say levels 3 and 4, as of the reindexing and typescript relations of what is called 8+.

The case with the different sets (levels 9, 10 and 11) of Wake galley proofs, which are also "tandem documents" consisting of overlay on the galley transcribed to accompanying typescript pages (which I think are also best designated with a "T"), is a little different. Though it is true that on a 9T page you will find the transcription of level 9 overlay, you do not (unlike the 8T pages) find it in a ribbon copy but in carbon (a second carbon, usually), and on a 10T page you do not find simply the level 10 overlay but rather (in the overwhelming majority of cases) you find the ribbon copy of the level 9 overlays (which are not reproduced on the level 10 page itself) immediately followed further down the page by the (first) carbon of the level 10 overlays. All of the typed overlays on the 10T page are indexed on level 10 with (usually) letters in longhand (probably Paul Léon's), the level 9 overlays with matching indices on the level 10 page but the level 10 overlays with their own indices different from those on 10T. Then on the last set of galleys, level 11, you find (as well as a very very very few new overlays) the index numbers for both major revisions, level 9 and 10, which are those also on the 9T and 10T pages. On 11T one finds in sequence down the page the first carbon of the level 9 overlays and the ribbon of the level 10 overlays. There are exceptions such as cases where a given overlay appears in ribbon on all three "T" pages or where the order of ribbon and carbon as between levels 9 and 10 overlay are reversed, or where alterations of the ribbon/carbon order for different passes through the typewriter show that Joyce sometimes added overlays after the first typing and had these added to more than one copy with no great attention to the general rule. But still, Joyce's basic procedure is clear enough, and both simple enough to have been workable and complicated enough to keep all his difficult material in order.

That procedure must usually have been this, for any given galley page (let us take galley 40 as a random example). Joyce adds overlay on the first set's page (level 9)(JJA49:95). With galley 40 on top of the pile of first set galleys on the table (and almost surely with two other piles next to it, the second and third sets with galley 40 also on top), the typist puts in the typewriter a packet of paper consisting of: the verso of the preceding galley (galley 39) from the second set (what will become level 10 typescript, i.e. 10T); a piece of carbon paper, the verso of galley 39 from the third set (what will become 11T); another carbon; and the verso of galley 39 from the first set (what will become 9T). This will achieve for level 9 overlay: ribbon typing on the 10T galley 39 verso (JJA49:388) of the overlays which wind up lettered "a-k", though we should note that "g-k" are done in a second pass through the typewriter to match the (at least) second pass through the galley, since overlays "a" through "f" are in black ink on the level 9 galley, while "g" and "i" are in red ink and "h" and "k" in blue; first carbon on 11T (JJA50:100; and second carbon on 9T (JJA49:94). The typist then transcribes the overlay from level 9, in the process adding a light place marker (usually a pencilled vertical line with a rightways hook at bottom) to the level 10 and 11 galleys where Joyce's original index for each overlay is on level 9-note that no index letters or numbers are typed before the members of the list of typed overlays.

With this done, Joyce can then take the second set (level 10) galley 40 page (JJA49:389) and inscribe new overlay on it (knowing from the placemarker indices where the level 9 overlay already is, and from the 10T verso of galley 38 what it is) with new indices in the text to the overlay in the margins, and when the typist gets it much the same procedure is followed. Into the typewriter goes: the 11T verso of galley 38 (on which is already the first carbon of the level 9 overlay); with under it a piece of carbon paper; and then the 10T galley 38 verso (on which is the ribbon of level 9 overlay). The first set, level 9, is set aside at this point, so that on the 9T pages we always find only the second carbon of the level 9 overlay (see JJA49:94), though it is not set aside until the index numbers are applied to all three sets. The level 10 overlay is then typed right below the level 9 overlay, so that on the 10T page (see JJA49:388 ) we find finally the ribbon copy of level 9 overlay and the first (and only) carbon of level 10 overlay, and on the 11T page (see JJA50:100 ) we find the first carbon of level 9 overlay followed by the ribbon copy of level 10 overlay. Again, however, the fact that there must have been two passes through the level 10 galley itself is indicated by the fact that on it the overlay "Salary Grab ...", indexed "1" in the right margin of JJA49:389 and "p" in the typescript, is in red ink where all the other overlays are in black, and to match this that "p" typing is in ribbon on the 10T page and carbon on the 11T page-in typing up that last overlay the two sheets were reversed.

Placeholder index marks for the level 10 overlay must be made on the level 11 galley 39 page, where there are already placeholder indices for level 9 overlay, and when the level 10 overlay is all transcribed (with whatever backings, fillings and variations from this pattern inevitably happened) the helper (Paul Léon, I am quite sure) wrote in the final index letters on all three "T" pages-since the hands and inks (usually red) are exactly the same, the level 9 galleys, rectos and versos of course, must not have been set aside until this point-in the sequence to match and regularize Joyce's index system unique to each of the level 9 and level 10 pages. That is, Léon did not write any index letters on the level 9 page where Joyce's own indices marked all the level 9 overlay (which thus does not match the letters we find on the 9T page). On the level 10 page he added what would become the final numbers (in order going down the page, for the most part) to the placeholder marks already on the page for level 9 overlay (no doubt checking to make sure the new numbers matched the proper overlays in the typed lists). He no more changed the Joyce indices already given to the level 10 overlay to match the new numbering system than he had changed the level 9 Joyce indices on that page, but on the level 11 page he at what must have been the same time both copied the final level 9 index letters from the level 10 page and gave the appropriate new set of letters to the placeholders put at typing-time on the page in the spots of Joyce's own overlay indexing of the level 10 overlay.

Now something of the same logical dilemma about designation we've confronted already appears here too, as a result of this procedure. As we said, the three sets of Wake galley pages are not sequential resettings, but simply different proof copies, just as the the second set of transition pages are simply different copies of transition. And if that was what led the Archive to call them 8 and 8+, why are the three galley sets not 9, 9+ and 9‡? It would be the more defensible in that exactly the same procedure of indexing is followed: index marks on the text pages of the second set match (usually, for of course there are errors) the positions of the first set overlays, and key to and are the same as index marks on the typescript, just as with the index marks on 8+ vis à vis 8+T. And the same for levels 10 and 11. The only difference is that because of carbon copies the "T" pages contain not just the overlay from the preceding level but that from "their own" level.

Now it is perhaps for that reason that the decision was made to designate them as different levels, 9, 10 and 11, and again it is a good reason for the user: no doubt about it, if you look at a level 9 page and across at its 9T match you can check the overlays against each other right there even though what you find on 9T is a carbon copy, whereas if you look at level 8 you have to flip the pages to a new place, find the 8+ index and look across at 8+T. And the same for level 10 and 10T, though of course when you look at 10T you also find, confusingly till you get used to it, the ribbon copy of the level 9 overlay. Unlike "the 8s", in the case of the galleys the overlays and their typed transcriptions are right there together, and calling them a single revision level, an "n", is at least rhetorically appropriate. And it is not perhaps even inappropriate in terms of the Archive's general propensity, which we have noted before, to give whole number levels, "n"s, to sets of document pages which are done as a whole and passed on to the procedure of creating another equally integral set, n+1. If it is not perfectly accurate, it is still a sentence we are, as the Archive implicitly is, happy to take away from all this complexity: "Joyce made a number of passes through each of two discrete sets of Wake galley proofs, and blended the results in an only lightly newly revised third set."

Finally, we might note that the oddities of calling the three interlocked copies of "the same" Wake galleys by different whole "n"s have only recently come into sharper focus with Luca Crispi's rediscovery of the two sets of Wake page proofs in the Paul and Lucie Léon/James Joyce Collection at the McFarlin Library of the University of Tulsa ("rediscovery" in the sense that though published as part of the Libraries Special Collections, they were unknown to the vast majority of Wake geneticists, to the Wake scholiast, we might say). These really are different "n"s, no two ways about it: one set is revised and then reset, and the resetting is then itself (lightly) revised. Nothing could be more clear than that they deserve separate designation as revision levels 12 and 13. We simply must bite the bullet and acknowledge that this requires re-designating the post-publication Corrections from the Archive's level 13 to level 14. Beyond that, given the unlikelihood of a complete reworking of the Archive and the relative scarcity of logical problems of the sort I have been pointing out here, I think we should be happy for what we have and that we can stay with the Archive designations, warts and all. The Archive does not need to be redone on their account.

Adequacy of the Reproductions

The adequacy of the Archive's designations of revision levels is only one aspect of the question of our comfort level in relying on it as the basic, pre-direct-inspection research tool into Wake genetics. In our daily practice, the biggest question is of its adequacy as a set of reproductions of the documents: just what are the limitations in its ability to give us information about them? In my experience, the answer is different for different document types, and the limitations will seem more or less severe depending on what use we need to make of the information.

Of course the most trustworthy information is to be had from reproduction of documents which were themselves originally produced mechanically-typescripts and proofs or printed pages. Except in cases (and there are many) where overlay obscures letters or words or (most commonly) punctuation, we can confidently check them for evidence of typist's or typesetter's error or of change possibly bespeaking missing instruction. Even here a distinction is to be made, however. Reproduction of printed text (whether in proof or as published) is seldom ambiguous-we usually learn little in London that we didn't already know from Archive work at home. But with typescript, especially cases of strikeover are often opaque in the Archive but can be made out on the original. For example, in the typescript of level 9 overlay of galley 39, at what is finally indexed "f" (JJA50:98), the typist added, incorrectly, the level 10 overlay ", franksman, to make a heart of glass," to the "Titus," already in place on 9T (JJA49:92), but then exed it all out and retyped "Titus," with the correct level 10 overlay, "on a burgley's clan march": the incorrect overlay is not legible under the exes in the Archive but is legible in the original. (The attentive reader may notice that we have here another exception to the ordering rules, since in this case both level 9 and level 10 overlays are in ribbon on the 11T page and carbon on the 10T, except for one level 10 overlay, in green ink while all the rest are red, which is, conversely, in ribbon on 10T and carbon on 11T.).

And when our intent is to infer matters of order of revision it is often useful to tell ribbon typescript from carbon, and of first carbon from second. You can make some good guesses from the appearance of Archive pages about which typing is ribbon and which is carbon (see how on galley 39 level 10T the final overlay typescript is so obviously ribbon as against the carbon which precedes it, JJA49:386), but that they remain guesses has come home to me in the number of times direct inspection has set me straight-look how much harder it is to tell the converse on the 11T page, JJA50:98. There is just no doubt, looking at the originals, about what is ribbon and what carbon, though the degree of carbon can remain doubtful just because a situation, for instance, of first carbon with an overused piece of carbon paper and second carbon with a fresh piece can seem to reverse priority. I suspect that science could (if one had the money) give objective grounds for claiming ribbon or carbon, but I doubt that even machines can solve that last problem better than the eye and brain.

But there are a few other sorts of information which can only, or at least can best be gleaned from the originals of typescripts. One type which only direct inspection can reveal stems from limits not in Archive reproduction but in the British Museum packaging of the originals. This is the situation where the careful notches which the Museum technicians cut in the gluestrips with which they were binding the document folios into the handsome black volumes, in order to reveal text on a page close to its edge-our illustrations show a good number of examples-sometimes do not quite manage to reveal it all. In the carbon of level 10 overlay of galley 49, JJA49:410, we can see the notch but not the period under it after the word "Vuncouverers", which however can be revealed by shining the bright fibre optic light wand in the British Library through the original page and gluestrip from beneath. Missing from the book, that period throws the grammar of the passage seriously off.(1) And while not simply missing from the reproduction like the information under a gluestrip, some visual information about a typescript stands out more clearly in the original: the wildly under-researched issue of just what machines were used (including their makes and models, a question I have found less easy to answer than the published charts of typewriter characteristics imply that it should be) will yield better results from a microscopic analysis of an original than a copy. And, to look again at galley 39, some information stands out better in the original though a hint of it is surely present in the copy. In lines 4 and 5 of the "o" overlay typescript (the ribbon one, JJA49:386), note first how it is impossible (as it is when looking at the document) to tell if the question mark after "Armen" overstrikes the also-present period or vice versa (though that the period seems more firmly struck gives a hint), but also that another odd characteristic can help with the answer. Looking also at the period after "aves" right before "Armen" (and also at the other periods in the passage) you will see that the typewriter used usually (for it does not do so after "Asterr" or "Diseased", though it does in each other instance) not only doubles itself in the line below, from some flaw in the mechanism, just a little bit more than one full line down and hair to the left of the intended period, but the intended period strikes not at the base of the line but in its middle. This typewriter is infrequently but occasionally used elsewhere in the galleys (for instance, this double period is the same one as the occluded period after "Vuncouverers" we just saw, or did not see), flagging for us how much more complicated than the procedure outlined above the galley revision production methods actually were, and I must say that although I spotted the oddity of the extra marks in the Archive copies I only saw what they were (and it took the help of one of the British Library's curators to make the proper inference even so!) in the originals, in which it is clear they can't just be dirt specks or artifacts of reproduction. And note: the doubled period (above and below) after "Armen" is just a bit lower with regard to the base line than that after "aves" (and the others). It seems likely then that the question mark was inscribed during the typing, and then, perhaps when the typescript was checked against the overlay itself (JJA49:387), it was noticed that the question mark was incorrect. As usual, when the carriage was turned back to the spot the registration was not quite perfect with the original line, and the period (and its twin) wound up lower, as we see. A pretty good inference, then, is that the period overstrikes the question mark and not the reverse, though, since the question mark is not more clearly struck out, the Wake typesetter took it, the more noticeable mark, as what was intended and the question mark, surely erroneously, prevails in the text. A counter-example can be seen in the second line of the typescript overlay, though it too can perhaps be solved by inference if not by direct inspection. Inspection of neither the 10T ribbon nor the 11T carbon can reveal whether "Mosses" is overstruck to "Mooses" or "Mooses" is overstruck to "Mosses". Again to the draft (JJA49:387): in the level 10 overlay Joyce first writes "Missas" and then, highly unusually, revises it by striking out the "i" and the "a" and writing "o" and then "e" below it in the margin next to the word. The typist sees the intended correction, but not (forgiveably) that the strikeout line for the "i" is not intended to delete the first "s" as well, and so (stretching a little) seems to decide that the marginal "o" was to replace both letters, yielding "Mooses". Mature reflection (or perhaps Joyce's instruction) probably led to reversing the judgment, and the second "o" is overstruck by the "s" it should all along have been, and as it is in the book.

Though we have been able to supplement arguments about the galleys with information about overlay colors, none of these cases absolutely turn on the fact that for financial reasons the Archive was restricted to reproducing (virtually) all documents in black and white, and that is why I say that typescripts and print, black and white to start with, more predictably give up their secrets than other document types. On the other hand, typed and printed documents almost always have handwritten material on them, and in all such cases, as in pure holographs, ink colors and even the differences among ink, pencil and crayon can be enormously informative, both as to fine-tuning the analysis of Joyce's revision process and making out the intended text. If there is one strongest argument for the re-doing of the Archive, it is (as so often in the history of technology) the desirability of the move from black and white to color. From attending to ink colors and nibs (and the hands with which these were employed) we can tell a lot about the order of Joyce's revisions even on pages, like those of level 5 and its sublevels and level 6, which were not entirely linear-that is, where there were further revisions made on earlier pages after the later pages were typed.

As we've seen, some level 5 pages were retyped for level 6 (sometimes more than once), and others (five in number) were not. On the retyped level 6 pages Joyce made fewer revisions than he did on level 5, and we've seen that most of those which were in red were "copied back" by Giorgio in green. On the level 6 pages we do also find, however, a good number of simple corrections in a fine-nibbed red Joyce hand, and in the same hand copies of some, but not all, of the large number of revisions which we find on the level 5 pages also in red but in a broader hand-the rest of those revisions appear typed on the level 6 pages. So we can see that Joyce made some red revisions on level 5, had the pages retyped, made some further red revisions on the originals and copied them forward to the new pages. For instance (JJA46:97and140), at level 5 the red "within the black of your nail of being" and "kayoed" are typed on level 6, p. 140, but "hyougono" (also in red though one can see even from the Archive reproduction that it looks a little different) is not typed on 6 but was added later and copied forward (along with a big red inkblot) in the finer hand.

Most of the changes to level 5 are in black, however, and these are all typed in the new pages except for a couple missed by the re-typist and restored in black-that is, Joyce did not make further black-ink revisions to the level 5 pages after they had been retyped. On the retyped level 6 pages, however, he made some important new revisions in black ink, and most of these we also find on the level 5 pages in green ink and in Giorgio's hand (one such on the pages we've just seen is "and five hundred and one muscles"): it must have been after this black-ink pass through the level 6 pages that Joyce asked Giorgio to do his best to square things up, explaining his note to HSW. With the five level 5 pages which were simply bundled in with the retypes it is impossible to tell, of course, at what time the various revisions were made, whether before or after the retypings of the other pages, but since we find on them black ink revisions characteristic of the first pass through the others, and red revisions only in the broad nib, we can propose that Joyce did not make changes on these pages after the retypings of the others.

A good example of the interlockings of black, red and green overlays from which inferences can be made only if we have access to the facts of the colors, is on p. 110. On level 5 Joyce drafted "O me and O ye, the hungray and anngreen", which was indexed as "W" to be tucked in after "one hundred again", and to precede the "F" extension "(and if she is older...my dear!)"-all these are in red and are typed for level 6, as are the several surely earlier black ink level 5 additions: ", da! da!," "caddishly", "crookedrib", "mouthless", "bondwomen". Then in red on level 6 he added ", cad and prim," which Giorgio wrote back, in green ink, tucking it into the right spot in the red overlay and striking out the exclamation point after "ye" to fit the comma which Joyce's draft requested, (2) as he also wrote back "stinkingplaster seal" (which however was a black draft on level 6) and "some" (a red draft on 6), and struck off the "er" from Joyce's own draft on 5, "murrmurrer", which he had copied in the fine-nib red to level 6 as "murrmurr". All of this could perhaps be recaptured, without the redoing of the Archive in color, by some published equivalent to what I am using myself, notes in themargin of my own copy of the Archive volumes. Accurate reproduction of colors would tell much more, though, and my method, at least in my hands, is vulnerable: I do not show "murrmurrer" on level 5 as red but black, which it just cannot be!

Knowing the colors is useful for levels 5 and 6, but is almost imperative for accurate interpretation of the galley proofs, levels 9-11T. In fact, the galley originals seem less like the reproductions than any other documents, partly because the Archive reproductions are so much more reduced in size than the other documents that the originals are surprisingly easy to read, but more because it isn't just order that we can infer from the colors, but the words and connections of the overlays themselves. We've glanced at the fact that Joyce and his helpers used a variety of colors (and of media-pen, crayon, pencil) for the several passes yielding many overlays on most galleys, and very often impenetrable tangles in black and white snap clear when you see the colors of each. Initial work with the Archive galleys, before checking against the originals, must of necessity be so much more doubtful than with any other revision levels (including the first drafts, which for all their density don't show up badly in reproduction since they are themselves mostly monochrome) that it is almost (though not quite) not worth doing.

But for an example of how the original galleys are necessary for adequate readings let us look not at color but at pencilled revision. Consider galley 24 at level 9 (Galley24), for the overlay to be inserted at index "g", after "flute", with the following "which Delaney" deleted. First note that Joyce after drafting "unscrewments" in black ink, in pencil changed the "u" to an "i" and the "c" to a "t", creating "instrewments", which the typist only partly read, giving "inscrewments" on the 9T page (of which the ribbon this time is on 11T, as shown!). Next, note that the black ink insertion caret for the secondary overlay ", Piggott's purest . . .", is clearly before "that onecrooned king...." As often on this galley, though, after the final intended "instrewments" there is a pencil line (very faint in the Archive) to the "Piggott's purest . . ." overlay, by which the position is changed by Joyce or by the typist under his instruction, since in these galleys pencil lines seem often to be added to make more explicit the connections between overlays and index marks, as a typist might well do. More complexly, then, Joyce drafted "horn, piped," as the end of his overlay, to lead back into the galley text's "anticipating . . ." after the scratched "which Delaney". But "piped" is boxed in pencil and a line (this time entirely invisible in the Archive reproduction) runs from it to where "drew" is crossed out in the text, effectively moving "piped" to replace "drew". The content of the overlay on 9T is correct (as we would expect if the pencil line is a result of Joyce's instruction to the typist), with the main overlay ending "horn", and "piped" as a separate overlay. But in indexing the 9T overlay (on the level 10 galley) a helper misunderstands the pencilled change: "piped" is carefully placed (with a new index letter, "l") before the comma after "rapsods", and "drew" is not deleted but remains. The grammar is not Joyce's but with "piped" as a participle is not absurd: "anticipating a perfect downpour of plaudits among the rapsods piped, drew . . .". Unlike many, though, in this case Joyce must have remembered his intention while revising the page proofs (which, being unpublished, I cannot illustrate), where he changes "rapsods piped, drew" by adding a comma after "rapsods" and deleting ", drew", restoring what he had intended in revising level 9 overlay: "rapsods, piped". Finally, we must note one more fact: above the gap between "Delaney," and "horn" at level 9 is a lightly but clearly (though not clearly in the Archive) pencilled "(Mr Delacey?)," which the typist takes as second overlay, inserting it (with the period after "Mr" which this typist characteristically adds) after "Mr. Delaney" . It could be in Joyce's hand, though it lacks an insertion caret (out of his practice). It could, however, be a query whether "Mr Delaney" is correct, as if the typist were marking the text in Joyce's absence to try to be true to Joyce's intention which was here not clear to him or her. But all things considered I judge not--the comma at its end is overlay-like, and "Mr Delaney" is clearer than one would think required a query.

As well as showing that good readings of the documents depend on reproductions which will show color and the differences among pencil, crayon and ink, these examples also show that the reproductions just need to capture more detail than does the Archive facsimile technology. In some cases, especially among first drafts in ink, the added contrast of the facsimile makes the Archive reproduction more clear than the originals rather than less. But far more often marks on the original are just invisible. For example, at level 3 of FW: 86.15 (BL 47472-160; JJA 46: 37), only a caret after "hyacinth" (which has no period) seems, in the Archive reproduction, to index the overlay "The gathering" in the top margin. But in the original it is clear both that there is a caret before "The gathering" and that Joyce tried to draw a line between the carets, though his ink did not take for most of it. We can see hints of the line down through "disguising", and then the paper is scored, though inkless, through "with" and "ratepayer", but then the scoring disappears under that word. Though the line is not headed in the direction of "hyacinth" at that point, Joyce's index lines often curve around in this way and it seems virtually certain that the overlay was to be inserted after "hyacinth", as it is at level 4.

I do not know if any photographic reproduction could be expected to show the scoring in this case unless specifically intended to do so, e.g., using a raking light. Even if possible, it would necessitate knowing each case in advance, whereas the whole point of good reproductions is to enable researchers to discover what is not already known. And in many places the documents are probably not going to be made legible by no matter how good the photography. Joyce's overlays, across the printed text, on the transition 4 page BL 47475-36v (JJA 46: 184), were scrawled to start with and are now so badly faded that even the ultra-violet light at the British Library reveals only a few of their mysteries: you can make out "Sacred avatar" and "By the Lord" and a few other words. The variable frequency lamp might do better (I have not had the money to try), and so a spare-no-expense redoing of the Archive might manage to give its readers, for those pages which are particularly impenetrable, what their naked eyes on the originals could not, much as the Archive reproductions of certain Buffalo Notebook pages included both plain photography and the use of filters to reveal note text under crayoned strikeouts.

Such a major high-tech project is nowhere on today's horizon, though I don't know that anyone has tried seriously to fund it. What we could hope for would be more complex than but not unlike the Hengwrt Chaucer Digital Facsimile published by Scholarly Digital Editions which Peter Robinson showed the attenders of the Genetic Joyce Studies conference in Antwerp in 2001: good pictures of the document pages linked to their transcriptions (full or split screen for each), which would in our case be keyed to synoptic representation of the whole document chain. A user could for instance, starting from the full genetics of a passage which showed where the interesting changes are made, call up the document and transcription of each relevant revision level. Display of the pictures of the documents themselves would be equally easy-that is, presentation without, what the Archive had to do, splitting documents into portions applicable to different chapters with repeats of the crossover pages. And integration with the Notebook Edition could also be easily achieved (technically, at least), with buttons on the genetics giving pictures and transcriptions of the notebook entries underlying the selected draft passages.

The bar to such a project is more financial than technical-the computing would be simple (one hard part, the reduction to chosen levels of the genetic representation, is done already), and the photography laborious but possible. But that financial bar is real, and yet the present situation is undesirable. The Archive is out of print. Most American scholars, and many elsewhere as well, have access to copies in major libraries, while a few by luck or good management have copies of their own. A few of us had the idea a while ago of asking Garland Press if they were interested in a simple reprinting of the existing Archive on CD-ROM or DVD, but they were not: it seems it is not as simple as it looks to photograph the photographs, and the possible return seemed to them not to warrant it.

To work with the existing Archive is far more difficult than we would like it to be, then, and it is frustrating to think that it is not possible for every student and/or scholar to pop a CD in the computer and check out the revision chain on an interesting passage: the Wake would be better read if it were. But for all such limits in its production and distribution as those we've been looking at, I think the Archive remains a marvelous and indispensable research tool with still the capacity of advancing Wake scholarship beyond the adolescence it finds itself in. We must simply remember to use it with care (like any tool) and that we must always check our finding against theoriginals-as long, that is, as the British Library lets us continue to see them, which is another story.

So it may be that I have simply had to learn to find my way around within the Archive's limitations well enough that they have become tolerable if not invisible to me. But if they are not greater than what I have found-if the Archive does not make more mistakes elsewhere than in chapters 2-4 and if I have not missed many of them there-I think Hans Gabler's remark that the Archive "has to be" redone is not true. It would be far better if it were, and we can easily envision a tool for the ages, far better in appearance, richer in information, and more perfect in its designations. But given the unlikelihood of the funding for such a project, for the ages immediately before us I think the Archive has plenty of life in it yet.

Notes

1. Why the period should not be on the ribbon copy, the 11T page, gluestrip or no, is another story, for which see my "The Development of the 'Eye, Ear, Nose and Throat Witness' Testimony in I.4", in Probes: Genetic Studies in Joyce (Amsterdam: Rodopi, 1995), 203-54, pages 222-23.

2. On level 6, the exclamation point remains which Joyce had not struck off, though Giorgio adjusted punctuation on level 5. The typesetter at level 7, apparently on his own initiative, dropped the comma instead, leaving the unJoycean punctuation we have in the book.