Jump to Content

Textual Editing at the Mark Twain Project: A Brief Account

Mark Twa
in by Ernest Walter Histed, 6 July 1907, London.

Mark Twain by Ernest Walter Histed,
6 July 1907, London.

On June 22, 1862, Samuel Clemens wrote his brother Orion from Esmeralda, California, that “those Enterprise fellows make perfect nonsense of my letters,” explaining that “like all d—d fool printers, they can't follow the punctuation as it is in the manuscript. They have, by this means made a mass of senseless, d—d stupidity out of my last letter.”

Clemens was complaining about the text of a letter he had just published in the Virginia City (Nevada) Territorial Enterprise, which evidently punctuated it as the printer saw fit, not as Clemens wrote it. In setting up the manuscript, the printers (and their proofreader) probably made some careless errors, but they also exercised their time-honored prerogative to decide how the text should be spelled and punctuated throughout.

Typesetters had a strong economic incentive to punctuate and spell according to rules enforced by the proofreader, because departures from these rules were errors, and had to be corrected on their own time. The result was typically a text that differed in some small ways from the original text in the manuscript. Clemens's complaint that the printers “can't follow” the original punctuation was probably a soft way of saying that they wouldn't follow it, insisting instead on substituting their own punctuation, like “all d—d fool printers.” Clemens had of course been a typesetter himself.

It is safe to assume that their changes were not made maliciously: the printers thought they were helping the author punctuate correctly, helping him avoid the embarrassment of appearing in print with “incorrect” punctuation. Unfortunately, they seem to have misjudged the effect of their changes, which Clemens found unacceptable because they made “perfect nonsense” of what he had written. In fact, Mark Twain would continue to complain for the rest of his professional career about how the typesetters (and also typists, proofreaders, and editors) altered the text of documents he put before them, despite his instructions to follow his manuscript exactly.

Samuel Clemens to Orion Clemens, 22 June 1862, original owned by the University of Virginia.

Samuel Clemens to Orion Clemens, 22 June 1862, original owned by the University of Virginia.

As it happens, this Enterprise letter is completely lost: neither its manuscript nor any printing of it is known to survive. But if an original copy of the Enterprise or any reprinting of the Enterprise containing the letter were to be found, the Mark Twain Project would undertake to produce a critical text of it.

Critical Texts, Critical Editions, and Scholarly Critical Editions

A critical text reproduces an existing text while also making changes to it, so as to better conform to some desired goal. If you were to make a typed copy of your grandmother's diary and in the process corrected some of her spelling, or supplied missing but intended words (for the sake of legibility, say), you would have created a critical text.

If you went on to make several copies of the text and supplied each of them with notes, you would have created a critical edition (an edition containing a critical text).

Critical texts are everywhere around us, not just in the realm of scholarship. There are countless motives for changing texts in the process of reproducing them. (Think of your most recent letter to the editor—“improved” it, did they?)

For scholarly critical texts, the goal may be to correct what the transcriber regards as errors; or it might mean incorporating what are thought to be authorial revisions from a later edition; or it might mean rendering a Shakespearean play in modern spelling. Such texts are called “critical” because the changes made require the use of critical judgment about what is or is not an error, authorial revision, or modern equivalent of an archaic spelling.

Not all scholarly editions contain critical texts, and not all critical editions are scholarly. A scholarly critical edition must include a detailed report to the reader of exactly how the editor has changed a single documentary text, or constructed a text from several different documents, such as manuscripts, typescripts, and proofs. Scholarly critical editions invariably have what is called a textual apparatus, meaning various notes and lists or coded texts specifying exactly what documents were used and how they were altered.

The Way Mark Twain Wrote It

But why would scholarly editors want to change a text that Mark Twain had published, let alone one that he left unpublished in manuscript?

The manuscript of Clemens's letter to the Enterprise is unlikely ever to be found, but finding a copy of the Enterprise (or another newspaper reprinting from it) is very much within the realm of possibility. A Mark Twain Project critical edition based on either kind of source would have the goal of reproducing it exactly, except where the editors thought it contained identifiable, correctable (not just obvious) errors.

Since Clemens complained about punctuation errors that fatally altered his meaning, the editors would certainly try to identify (and correct) those particular marks. Lacking the original manuscript, they might not succeed in locating the offending punctuation, let alone correcting it. And certainly without the manuscript, they could not restore Clemens's punctuation where it had been altered but was not causing the text to be “nonsense.” In other words, the editors would judge the punctuation critically, and draw on whatever evidence they could to help the author say what he intended to say by applying his standards for spelling and punctuation, not theirs or the printers'.

Textual Apparatus Reports the Editors' Changes

Of course if there were an extant manuscript for this letter or for any text, the editors would certainly rely on it for primary evidence of what Clemens actually wrote. They would, in addition, eliminate the author's errors should these be detectable (supplying a necessary but inadvertently omitted word, for instance). It is changes of this kind that are created by exercising critical judgment—supplying intended authorial readings even when they were entirely absent from all the documents, or when they are presumed to have existed in the lost manuscript but can no longer be literally seen—clearly, a risky business.

It is because of that continuing risk that the editors would always provide the critical text with precise information about what they had changed, why, and on what evidence, thereby producing a textual apparatus. Also to that end they would supply notes explaining any obscurities in the text, any references to people, events, or places that were no longer self-explanatory nearly a century and a half after the letter was written—especially since the validity of any change they made to the text (or any change they declined to make) would depend very much on their understanding as fully as possible what Mark Twain meant to say. (The information in these notes is of course primarily intended to help the reader. It is not often mentioned that it is also essential to the editors as they decide how to construct the text.)

The difference between one critical text and another—in this case one by the Enterprise printers, one by the Project editors—often arises from the slightly different goals each had in mind, or different understandings of what the author intended. The printers wanted to reproduce at least the words and word order of the manuscript Clemens had written (barring anything they regarded as obscene or improper, like “devil” or “damn”). And to make the text better conform to what they regarded as correct punctuation, they felt wholly free to alter it. The Project editors also want to reproduce Clemens's words and word order, without regard to anyone's notions of propriety; likewise his spelling and punctuation, without regard to conventions, modern or otherwise, that he declined to honor. They would feel free to alter the text only where it contained an error, something not intended by the writer.

It is obvious that there could in fact be as many (slightly) different critical texts of the same work as there are editors or typesetters willing to undertake the task. That is because the critical judgments necessarily made by the transcriber are human judgments, and unlikely to be identical from person to person. “It were not best that we should all think alike; it is difference of opinion that makes horse-races.”

Determining the Author's Intentions

The most important difference between the Enterprise version and the Project's version would not be in any certainty that greater fidelity to the author's intentions had been achieved by one or the other, but rather in their accountability for the changes made. Printers typically make no record of what they change or why. Even the fact of change is often something obvious only to the writer, and usually then only when “nonsense” or other obvious errors are the result.

Scholarly editors, on the other hand, have an obligation to document and support all changes made to the source text. It is this record of decision, evidence, and reasoning combined with the willingness to make changes to the source text for the purpose of coming closer to what the author intended that distinguishes a scholarly critical edition from most other kinds of edition.

The primary purpose of the critical editions produced by the Mark Twain Project is to provide texts that embody, as exactly and fully as possible, within the limits of the surviving and available evidence, those words and marks of punctuation intended by the author at a specific time in the history of that text—often its “final” form, but not necessarily so. This goal may likewise include reincorporating any and all graphical elements (such as illustrations) intended by him, and even tables of contents and lists of illustrations he included in the first edition but did not himself write. The accompanying textual apparatus documents every textual choice made by the editors, so that those choices can be critically judged, not simply accepted blindly.

The process of creating such an edition is called establishing a text—not because it is ever “established” once and for all, or is the only text possible, or is in any sense “perfect” or “ideal,” but because it is explicitly built up (established) from countless critical judgments about what to include and what to leave out, any of which may be wrong, or contradicted by new evidence, or by old evidence newly understood. Texts of this kind are also called “authoritative,” not because the editors are authorities, but because the author's wishes have been followed as closely as possible.

It surprises some readers to learn that any text, especially one that an author like Mark Twain has deliberately published, does not already read exactly as he wanted it to read. We tend to assume that a novel that has been printed and proofread and bound and issued as a book necessarily embodies “the” text intended, barring perhaps a few obvious “typos.” But the fact is that all texts are constructed (or reconstructed) by fallible human beings who must exercise judgment in doing so, whether they be the author and his typist, or scholarly editors who have studied the extant documents but cannot interview the author in order to decide exactly what he intended to write at every point.

All texts are therefore unstable in the sense of being subject to change from new or better evidence. And even though no two scholarly editors are likely to make exactly the same decisions about a text of any complexity, what they will do, if competent, is make choices guided by the clear purpose of realizing the author's intentions for it, not their own intentions and preferences, nor anyone else's. And they will record exactly what they have done.

As with most scholarly editions of American authors, the editorial policies of the Mark Twain Project grew out of editorial methods and theories developed originally for editing Shakespeare's texts. They are especially indebted to the meticulous reasoning and explication of G. Thomas Tanselle. This brief summary does not attempt to do full justice to the complexity of that evolution. For an excellent, much more thorough discussion of it, with emphasis on recent developments in the Project's policies, see Richard Bucci, “Tanselle's ‘Editing without a Copy-Text’: Genesis, Issues, Prospects,” Studies in Bibliography 56 (2003–4), pp. 1–44.

Correcting a Published Text: Adventures of Huckleberry Finn

Image of Chapter 1 of Huckleberry Finn.

In writing and publishing Huckleberry Finn, Mark Twain himself made small mistakes (giving one character two different names, for instance) which he failed to correct before or after publication. The typist copying the manuscript for him, and the typesetter copying the heavily revised typescript made still other mistakes, as well as thousands of well-intentioned changes designed to provide “correct” punctuation and spelling.

The typesetter and the typesetter's proofreader or editor made further changes not called for by the author, and the publisher suggested changes that the author accepted, but without due care for the consequences, a rather different kind of error. Most of these changes were departures from what the author submitted to the publisher in the form of a typed and heavily revised copy of the original manuscript, and they can be corrected, albeit not with absolute certainty, even though only the manuscript and not the revised typescript survive. Correction is possible because the manuscript and the first edition set from the lost intervening typescript can be closely compared, and the differences assigned to the author or to his several agents.

The First Edition In chapter 5, for example, where Huck is describing Judge Thatcher's past attempts to reform his drunken father, the text reads as follows in the edition Mark Twain first published, and in all subsequent editions derived from it. When Pap got out of jail, to which he had been briefly condemned for drunkenness,

the new judge said he was agoing to make a man of him. So he took him to his own house, and dressed him up clean and nice, and had him to breakfast and dinner and supper with the family, and was just old pie to him, so to speak. And after supper he talked to him about temperance and such things till the old man cried, and said he'd been a fool, and fooled away his life; but now he was agoing to turn over a new leaf and be a man nobody wouldn't be ashamed of, and he hoped the judge would help him and not look down on him. The judge said he could hug him for them words; so he cried, and his wife she cried again; pap said he'd been a man that had always been misunderstood before, and the judge said he believed it. (p. 42)

The Manuscript We know that this passage, like every other part of the text, was first written out in longhand by Mark Twain, then typed by a typist he hired for the purpose, and the typescript was then revised and corrected by the author and eventually handed to the printer, who set the first edition text from it, providing the author with a chance to correct the proofs before publication.

Here is a line-for-line type facsimile of part of the last sentence quoted above. (For an image of this part of the manuscript itself, see p. 814 of Adventures of Huckleberry Finn, ed. Victor Fischer and Lin Salamo [Berkeley and Los Angeles: University of California Press, 2003].)


on him. The judge said
he could hug him for
them words; so he cried,
& his wife she cried, & the
old man cried
said he'd been a man that
had always been misun-
derstood, before, & the

The first edition text left out “& the old man cried” which is marked with gray background in this type facsimile.

As a result, the wife was said to cry “again” when she had in fact not cried before.

Also, where there are three bullets in the left margin, someone added the word “pap” so that the text read “pap said he'd been a man …”

What Happened? The problem for the critical editor is therefore to decide whose changes these were, and how they came about. Since they introduce an error or inconsistency into the text, it seems unlikely that they were all deliberate changes by Mark Twain, even though he had the opportunity to revise this passage both in the (now missing) typescript and on proofs of the first edition, also lost.

A better explanation is that the typist who copied the manuscript inadvertently skipped from “she cried” to the word following “man cried” (“again”) creating the reading “and his wife she cried again”—what is known technically as an eye-skip, something all typists do occasionally and that this particular typist was demonstrably prone to do. The omission not only had the wife crying again before she had cried at all, it also left the phrase “said he'd been a man …” with an ambiguous subject (“the old man” or “the wife”), and Mark Twain presumably fixed this problem by adding “pap” on the typescript (or possibly the proof).

Restoring the Original On the basis of this analysis, the editors of Huck Finn restored the original reading of the manuscript, omitting the word “pap” on the grounds that it was only added to correct an error introduced inadvertently by the typist and was now corrected by restoring what Mark Twain originally wrote.

One cannot claim for this decision anything like absolute certainty that it follows Mark Twain's intention better than the reading of the first edition. It is at least remotely possible that he himself made both changes and simply did not notice the slight problem he introduced (but if he did make that error it would itself call for correction, since an error is by definition not intended). On balance, it seems more likely that the reading published in the first edition was the result of inadvertent error by the typist and only partial correction by the author. The manuscript reading is Mark Twain's original intention, and is in any case all his, even if he did “correct” the flawed passage.

The argument is, therefore, that the reading of the manuscript is more authorial than the reading of the first edition, and it is adopted. But as with all such cruxes, different readers might come to different conclusions; so the decision to replace the first edition reading with the manuscript reading is supported in the technical apparatus by listing the change, and by an explanation of the editors' reasoning, as well as an image of the manuscript the typist is thought to have miscopied.

Instability: Why Published Texts May Change

But if the text produced by a critical editor is not certainly what the author intended, why bother to create it?

First, because it certainly can be and should be closer to what the author intended, if only because it can repair accidental damage inflicted, say, by the typist or typesetter.

Second, because the explicit, conscious, deliberate, and evidence-based decisions of a scholarly editor ought to be preferred to the inexplicit but no less crucial decisions of typist, typesetter, and proofreader, usually made without access to all the evidence, or even consultation with the author.

A critical edition, if properly done, gives us a more authoritative text—even if it remains the case that still better judgment or evidence, or both, would produce an even better text. It is for this reason that we never refer to our critical editions as “definitive.” All texts, including those produced at great expense and effort by the Project, are permanently unstable, always subject to the discovery of new evidence or the more skillful interpretation of old evidence.

It is worth mentioning here that this process of finding and removing changes made by others also creates a record that should be of interest to editors who have recently argued for the “socialized” text, i.e., one that recognizes and celebrates the collaborative nature of all published texts. Mark Twain himself was not of this mind; indeed, he repeatedly denounced such interference in what he had written. The Mark Twain Project editors are therefore interested in giving Mark Twain the text he wanted but never got. But their record of readings emended out of the text is also a record of what the various actors in the social situation added to it.

Instability in Huckleberry Finn Our editions of Huckleberry Finn are a vivid example of this instability. In 1988, the Mark Twain Project published what turned out to be our first critical edition of this work. The editors established the text by relying on the only extant and then available part of the author's manuscript (it consisted solely of the last half of the book), exercising all due care and judgment to bring the whole text into more perfect alignment with the author's intentions. But in 1990, the first half of the author's manuscript, missing for more than a hundred years, was discovered in a Hollywood attic. By 2003, the Project had re-edited the whole novel based on this now-complete manuscript. Good and carefully constructed as it was, the 1988 edition was in no way “definitive.”

Only four of Mark Twain's major books—The Innocents Abroad, Roughing It, The Gilded Age, and A Tramp Abroad—are still lacking their complete original manuscripts. And dozens of shorter works published in newspapers and journals have still not had their original manuscripts (or in some cases, their revised typescripts) located. But some or all of these could be found at any time. In all such cases, the instability of the texts as they are presently constituted is greater than it would be if they had been critically edited using their manuscripts.

Instability in Other Works Various kinds of instability are evident when we look at the many, many short works—essays, newspaper letters, journalism of all sorts—that we know, or suspect, Mark Twain wrote, for far less than 80 percent of these texts were ever collected or reprinted by him.

The majority exist in authoritative form only in their first publication, or in some few cases, also in their original manuscript; some also exist in manuscript deliberately left unpublished at his death. Outstanding problems for works of this kind include:

  1. Finding the first publication or, better still, the manuscript that can serve as the basis of the edited text;
  2. Lacking either, finding one or more re-printings of the original book or journal printing, from which the editors can construct a reliable text;
  3. Verifying in some way that Mark Twain actually wrote the work in question, which usually means locating where it first appeared and establishing that Mark Twain did in fact contribute work to the journal in question; and
  4. Restoring, where possible, the original words and punctuation to physically damaged texts when they survive only in a single copy of the first printing or only in a unique but damaged manuscript.

All of these problems are part of the larger challenge of historical research aimed at recovering and preserving what Mark Twain actually wrote. The results of this research apply not just to the words and punctuation of works he is known to have written, but also to the question of whether he did in fact write the work attributed to him.

Although the explicit purpose of a critical edition is to offer a text more faithful to the author's intentions than any existing text, the pursuit of that guiding purpose invariably generates information about how and when the text was written, revised, and published—information which simply cannot be found in any other way. For in order to make the kinds of choices the editors must make in creating a critical text, they inevitably generate a more detailed reconstruction than ever existed of how the text was changed, from point to point, by the author as well as his various agents in the publishing process.

In the case of Huckleberry Finn, for example, it was discovered that more than half the revision Mark Twain performed on his draft before publication took place on the typescript copy of the manuscript which eventually served as printer's copy. The typescript itself has not been found; it was almost certainly thrown away by the typesetter. Yet exactly what Mark Twain did on that typescript, how he changed its words, how he deleted draft passages from it, must be figured out by the editors in order to make their editorial choices for the critical text. The result is a text closer in thousands of details to what Mark Twain intended to publish in 1884, but perhaps just as important is the detailed record of how he must have revised his masterpiece, something that one cannot now simply “see” on the revised (lost) typescript.

Texts that Mark Twain published, or that he intended to publish, require the editors to construct the critical text, comma by comma, word by word, relying on all the documentary evidence they can find. One can think of their goal as republishing the text as he would have published it, given the time and resources now available to us. Broadly speaking, errors in the source documents would be corrected because they were by definition not intended, and changes to the text that can be reasonably attributed to him would be adopted, or not, depending on what point in the evolution of a text we were aiming for. (For instance, changes made to sketches for several different sketchbooks he published over a decade would be clearly identified and reported, but if the editors were aiming at the text of the earliest printing, would not be incorporated.)

Editing Texts of Private Papers

Texts that Mark Twain did not publish and may never have intended to publish—chiefly letters and notebooks—pose a different set of problems and require a different approach to errors, inconsistencies, and revisions. Transcriptions of letters, for instance, differ from critical texts of works intended for publication because their purpose is to provide as complete and accurate a text as possible of what the writer actually sent, not some polished and corrected version of it.

When editing private documents, the editor aims to reproduce the document as exactly and fully as possible, but not to introduce the author's later corrections and second thoughts, or to correct misspelled words and other errors, or to omit such things as cancellations and insertions made at the time of composition. A transcription which did undertake to correct and polish the text would be almost useless to a historian, since the value of letters as evidence arises from their having been created, sent, and received at a particular time and in a particular condition. A letter can be thought of as a snapshot in time, and the wisest editorial course is to resist the temptation to “clean it up.”

Picking and Choosing But scholarly editions of letters and notebooks—even those that claim to be literal, verbatim, or “diplomatic” transcriptions of the original, or that consist of photographic reproductions of them—of necessity pick and choose which classes or kinds of detail to copy or transcribe and which to leave out, even though the overall goal is to copy the original as closely as possible. Photographs and digital scans, for example, notoriously leave out, or heighten the clarity of, certain details in the original. Which details are included and how they are affected are choices made by the editor.

Should the facsimile show the faint blue rules on which the author wrote, even though these are close to invisible in the original, and showing them would make other elements of the manuscript more difficult to read? Did Mark Twain intend to write “water-wheel” or “waterwheel” when he broke the word at the end of a line? And all transcriptions that decline to reproduce the text line-for-line and page-for-page are unable to replicate certain occurrences in the manuscript which are intelligible only in connection with the original lineation.

So even “literal transcriptions” of documents ought, in our view, to be considered critical texts, even though the decisions being made by the editor are of a different order of magnitude from those involved in editing a literary work. But in both cases the editor is exercising critical judgment to ensure a text that is as faithful as possible either to a text as it was written and sent, or a text as it was inscribed or altered on a series of documents, like typescripts and proofs. Critical judgment is inescapable in both kinds of texts.

Plain Text Our letter transcriptions are designed to change only as much as necessary, and to make the transcription as legible and intelligible as the original, or more so. In contrast, the editor of a literary work usually cannot transcribe a single document and still get the text that the author intended. Instead, the editor wants to construct a text as inscribed on several different documents whose various readings must be brought together and sorted out if the author's intentions are to be followed. Of course, it is also sometimes the case that the original letter documents do not survive, but only typewritten or printed copies of them. In that case the text is still a critical text in the sense that the editors aim, not to change as little as possible in the source copies, but rather as much as necessary in order to approximate the readings of the lost original.

For this purpose the Project uses a system of transcription invented by the editors, called plain text. Plain text is designed to maximize the kinds and amount of manuscript detail that can be legibly and intelligibly included in typographical transcriptions, and it is fair to say that there are now no competing systems of transcription which can include as much detail while also making the text highly legible.

Plain text is a frankly critical method of transcription, and purposely omits certain details that it could include, but only at the expense of legibility and without adequately increasing meaningful detail to offset that loss of legibility. For instance, most of the original lineation in a letter is the result of chance—where and how the words must be divided into lines to fit on the paper being used—and therefore cannot ordinarily have any meaning. If one were to reproduce letters line-for-line in type, all written on many different kinds and sizes of paper, the result would often be harder to read than the original. (Obviously, where lineation is not random, but is deliberate, as in an address or in the opening lines of a letter, it is meaningful, and plain text does reproduce it.)

There are also some few details in manuscript letters for which we have no legible method of transcription (places where the writer's pen or paper has caused a problem, and he must as a consequence cancel and then rewrite a few characters). But in these cases the actual reading of the manuscript is always instantly accessible in this electronic edition, recorded as an emendation and hyperlinked to the edited text. The plain text system, and this approach to editing private papers, is discussed in more detail in the Guide to Editorial Practice.

October 2007