A few of the typographical signs in these letters may seem a bit unfamiliar, if not wholly exotic. Others may seem familiar, even though they in fact no longer have the precise and accepted meaning they had when Clemens used them. Especially because some signs have fallen into disuse and (partly for that reason) been adapted by modern editors for their own purposes, it is the more necessary to emphasize that here they bear only the meaning given them by Clemens and his contemporaries. Purely editorial signs in the transcription are identified and defined in a section always accessible from the top of the note pane, and since they sometimes adapt typographical conventions, they must not be confused with authorial signs. They have, in fact, been chosen to help avoid such confusion, and especially to avoid usurping the normal, typographical equivalents for authorial signs.

Still, authorial signs present two related but distinct problems for successful transcription: (a) how to explicate those signs whose authorial meaning differed from the modern meaning, but can still be recovered, at least in part; and (b) how to represent authorial signs whose earlier typographical equivalent, if any, remains unknown—at least to the editors. The glossary of Special Sorts and table of Emphasis Equivalents which follow here are intended to solve these problems—to alert the reader to those changes in meaning which we can identify, and to describe the handwritten forms for which the typographical forms are taken to be equivalent—or, in a few cases, for which they have been made equivalent because we lack a better alternative.

The glossary includes signs that do not appear in every letter, of course, and it omits some signs that will be added only as they become relevant, as letters from later years are edited. Like the glossary, the table provides some information that was, and often still is, regarded as common knowledge, which may explain why the contemporary equivalent for some authorial signs has proved so elusive. That no table of comparable detail or completeness has so far been found in any grammar, printer’s handbook, dictionary, or encyclopedia, would appear to indicate that the system of emphasis was almost completely taken for granted, hence rarely made fully explicit or published, even by those who relied upon it. The particular meaning for Clemens of all such equivalents between manuscript and type, at any rate, has had to be deduced or inferred from the letters themselves, and from his numerous literary manuscripts, with his instructions for the typist and typesetter (sometimes with the further evidence of how they responded to his instructions), as well as from the consistent but usually partial evidence in a variety of printer’s handbooks, encyclopedias, manuals of forms, and other documents bearing on what we take to be the system of equivalents between handwriting and type.3explanatory note

Special Sorts

asterisks * * *  

Always called “stars” by Clemens and by printers generally, asterisks appear in his manuscript as simple “Xs” or crosses (X), or in a somewhat more elaborate variant of the cross (

X

), often when used singly. In letters (and elsewhere) he used the asterisk as a standard reference mark, either to signal his occasional footnotes, or to refer the reader from one part of a text to another part. (The conventional order of the standard reference marks was as follows: *, †, ‡, §, ‖, ¶, and, by the end of the century, ☞.) He also used asterisks for a kind of ellipsis that was then standard and is still recognizable, and for one now virtually obsolete—the “line of stars”—in which evenly spaced asterisks occupy a line by themselves to indicate a major omission of text, or—for Clemens, at any rate—the passage of time not otherwise represented in a narrative. For the standard ellipsis, we duplicate the number of asterisks in the source, thus: * * * * (see also ellipsis, below). In transcribing the line of stars, however, the exact number of asterisks in the original becomes irrelevant, since the device is intended to fill the line, which is rarely the same length in manuscript as it is in the transcription. The line of stars in the original is thus always transcribed by seven asterisks, evenly separated and indented from both margins, thus:

*      *      *      *      *      *      *

X

  

Clemens drew the brace as a wavy vertical line that did not much resemble the brace in type, except that it clearly grouped two or three lines of text together. He drew braces intended for three or more lines as straight (nonwavy) lines with squared corners, like a large bracket, usually in the margin. He occasionally used the two- and three-line braces in pairs, vertically and horizontally, to box or partly enclose one or more words, often on a single line. The one-line brace ({ }) was evidently not known to him, and would probably have seemed a contradiction in terms. It appears to be a modern invention, but has sometimes proved useful in the transcription when the original lineation could not be reproduced or readily simulated (see SLC to Orion Clemens, 9 June 62click to open link). Otherwise, the transcription always uses a two- or three-line brace and preserves, or at least simulates, the original lineation.

dashes –  —
——  ———
=  ===  

Clemens used the dash in all four of its most common typographical forms (one-en, one-em, two-em, and three-em), as well as a parallel dash, usually but not invariably shorter than an em dash. The parallel dash appears to be used interchangeably with the much more frequently used em dash, but almost always at the end of a line (often a short line, such as the greeting). Its special meaning, if any, remains unknown. Clemens occasionally used dashes visibly longer than his em dash, presumably to indicate a longer pause: these are transcribed as two-, three-, or (more)-em dashes, by relying on the length of em dashes in the manuscript as the basic unit. That Clemens thought in terms of ems at all is suggested by his occasional sign for a dash that he has interlined as a correction or revision (|—|), which was then the standard proofreader’s mark for an em dash. Clemens used the dash as terminal punctuation only to indicate abrupt cessation or suspension, almost never combining it with a terminal period. Exceptions do occur (SLC to Olivia L. Langdon, 26 and 27 Jan 69click to open link), but most departures from this rule are only apparent or inadvertent. For instance, Clemens frequently used period and dash together in the standard typographical method for connecting sideheads with their proper text (‘P.S.—They have . . .’), a recognized decorative use of period-dash that does not indicate a pause. The em, two-em and, more rarely, the en and the parallel dash were also used for various kinds of ellipsis: contraction (‘d—n’); suspension (‘Wash=’); and ellipsis of a full word or more (‘until ——.’). Despite some appearance to the contrary, terminal punctuation here again consists solely in the period. On the other hand, Clemens often did use the period and dash combined when the sentence period fell at the end of a slightly short line in his manuscript (‘period.— | New line’), a practice copied from the typographical practice of justifying short lines with an em dash. These dashes likewise do not indicate a pause and, because their function at line ends cannot be reproduced in a transcription that does not reproduce original lineation, are always emended, never transcribed. Clemens used en dashes in their familiar role with numerals to signify “through” (‘Matt. xxv, 44–45’). And he used both the em dash and varying lengths and thicknesses of plain rule—in lists, to signify “ditto” or “the same” for the name or word above, and in tables to express a blank. See also ellipsis and rules, below.

ellipsis ----- ...... **** – – – – – – – — — — —  

Nineteenth-century typography recognized an enviably large variety of ellipses (or leaders, depending on the use to which the device was being put). Clemens himself demonstrably used hyphens, periods, asterisks, en dashes, and em dashes to form ellipses or leaders, in his letters and literary manuscripts. The ellipsis using a dash of an em or more is also called a “blank” and may stand for characters (‘Mr. C—’s bones’) or a full word left unexpressed. In the second case, the dash is always separated by normal word space from the next word on both sides (‘by — Reilly’), thereby distinguishing it from the dash used as punctuation (‘now—Next’), which is closed up with the word on at least one side, and usually on both (‘evening—or’). When any of these marks are used as leaders, the transcription does not necessarily duplicate the number in the manuscript, using instead only what is needed to connect the two elements linked by the leaders. But for any kind of ellipsis except the “line of stars” (see asterisks), the transcription duplicates exactly the number of characters used in the original.

fist
☞ ☜  

Clemens used the “fist,” as it was called by printers (also “hand,” “index,” “index-mark,” “mutton-fist,” and doubtless other names), not as the seventh of the standard reference marks, but for its much commoner purpose of calling special attention to some point in a text. As late as 1871 the American Encyclopaedia of Printing characterized the device as used “chiefly in handbills, posters, direction placards, and in newspaper work,”4explanatory note but Clemens used it often—and without apology—in his letters. We transcribe it by a standard typographical device, either right- or left-pointing, as appropriate, except in special circumstances. The following case, for instance, requires facsimile of the original, since Clemens clearly meant to play upon the term “fist” by drawing the device as a distinctly open hand:

X

“Put it there, Charlie!” (SLC to Olivia L. Langdon, 12 Dec 68click to open link)

paragraph ¶  

The paragraph sign is both a mark of emphasis and the sixth of the reference marks. It is actually “P” reversed (left for right, and white for black) to distinguish it from that character. Clemens, however, commonly miswrote it as a “P,” drawing the hollow stem with large, flat feet, but not the left/right or white/black reversal in the loop. Whenever the sign is used in a letter, we transcribe it by the standard typographical device, with a record of emendation when it has been misdrawn. Clemens used the paragraph sign as a reference mark and as shorthand for the word “paragraph,” but most commonly in letters to indicate a change of subject within a passage, one of its original meanings. When he inserted the paragraph sign in text intended for a typesetter, he was doubtless specifying paragraph indention. But when he used it in a letter, he was usually invoking that original meaning. The transcription always uses the sign itself, even when it was inserted (¶) or was manifestly an instruction to a typesetter. In the textual commentary, however, the paragraph sign in brackets ¶ is editorial shorthand for “paragraph indention.”

X

  

Double rules (a), parallel rules (b), and plain rules (c), or rule dashes, in manuscript are usually, but not invariably, centered on a line by themselves, serving to separate sections of the text. When used within a line of text, they are positioned like an ordinary em dash and may serve as a common form of ellipsis, or to mean “ditto,” or simply to fill blank space in a line. This last function may be compared with the original purpose of the eighteenth-century flourish, namely to prevent forged additions in otherwise blank space. But as with the flourish, this function had in Clemens’s day long since dissolved into a mainly decorative one. Rules appear in Clemens’s manuscript in three distinguishable species, each with two variant forms. We construe wavy lines in manuscript as “thick” rules, and straight lines as “thin” rules, regularizing length as necessary. (a) Double rules appear in manuscript as two parallel lines, one wavy and the other straight, in either order. (b) Parallel rules appear in manuscript as two parallel lines, either both wavy or both straight (thick or thin). (c) Plain rules appear as single lines, either wavy or straight (thick or thin).

The transcription always represents authorial revisions where they occur in the text, just as it does all but the most ephemeral kinds of self-correction. Either kind of change is wholly given in the transcription, except when giving all details of an individual occurrence or all cases of a particular phenomenon would destroy the legibility of the transcription. For revisions, the transcription always includes at least the initial and the final reading, with intermediate stages (if any) described in the record of emendations. But in letters, revisions are rarely so complicated as to require this supplemental report.

Self-corrections are sometimes omitted by emendation, and are more frequently simplified by emendation than are revisions, chiefly because if fully transcribed in place they often could not be distinguished from revisions, except by consulting the textual commentary—even though the distinction is perfectly intelligible in the original letter. This limitation comes about in part because causal evidence of errors, such as a line ending (“misspel- | ling”) or a physical defect in the pen or paper, cannot be represented in the transcription without adding a heavy burden of arbitrary editorial signs. Thus a word miswritten, then canceled and reinscribed because of such a defect would, in the transcription, look like a revision, or at least like hesitation in the choice of words. In part, however, the problem with transcribing self-corrections lies in the sheer number that typically occur in manuscript. Self-corrections internal to a word, for example, are so frequent that more than one kind of emendation has had to be invoked to bring their presence in the transcription within manageable, which is to say readable, limits.

Another limitation of the present system is that the transcription does not distinguish between simple deletions, which may have been made either before or after writing further, and deletions by superimposition, in which the writer deleted one word by writing another on top of it, hence almost always before writing further. Because we have as yet no way to make this distinction legible in the transcription, we represent all deletions as simple deletions. In 1988 (with the first printed volume of Mark Twain’s Letters) we supplemented the transcription in this respect by recording as emendations “each instance of deletion by superimposition” ( L1 , xxxiv), thereby enabling anyone to ascertain the method of cancellation used, since deletions by superimposition were always described (‘x’ over ‘y’), whereas simple deletions were not. The advantage of this procedure was that, while clumsy and expensive, it meant the transcription with its apparatus could always be relied on to indicate the method of cancellation, whether or not the editors thought this information was useful in any given case. Its great disadvantage was that it caused the record of emendations to be nearly overwhelmed by reports of superimposition, only a small percentage of which were of any interest.

Pending the invention of an affordable, reliable way to signal this distinction in the transcription, subsequent volumes record deletion by superimposition only when it is judged to be useful information—chiefly where the timing of a cancellation can be established as immediate from this evidence in the manuscript, although in the transcription the timing appears indeterminate. For example, where the transcription reads ‘Dont you’, the manuscript might show either (a) that ‘you’ followed ‘Dont’ (a simple deletion, hence indeterminate), or (b) that ‘you’ was superimposed on ‘nt’ (deletion by superimposition, hence certainly immediate). Since the record of emendation gives only those cases where cancellation by superimposition establishes the immediate timing of a change, and only where this fact is deemed relevant, readers are entitled to infer from the absence of an entry for two such words (‘& at’, for example) that simple deletion has occurred. Where it is deemed irrelevant (‘in order that so that’), the method of deletion is neither transcribed, nor recorded as an emendation.

All transcribed deletions are, with minor exceptions, fully legible to the editors, and were therefore arguably so to the original recipient of the letter. But Clemens clearly did make some deletions easier or more difficult to read than others. His occasional addition of false ascenders and descenders to his normal deletion marks, for instance, has the intended effect of making it quite difficult to read what was canceled, at least so long as the presence of these false clues remains undetected. Such cases show, in fact, that Clemens must have known that his normal methods of cancellation did not prevent most readers from reading what he crossed out. Indeed, we know from certain letters that he enjoyed teasing his fiancée about her practice of reading his cancellations—even going so far as to challenge her to read a deletion he made “impenetrable” (SLC to Olivia L. Langdon, 28 Feb 69click to open link).

It is clear that Clemens experimented with a wide variety of cancellations more or less actively intended to be read, but even apart from these deliberate and relatively rare cases, his methods of cancellation in letters ranged across a full spectrum of difficulty. The transcription does not, however, attempt routinely to discriminate among these, simply because we lack any conventional means for representing the differences legibly. Cancellations thus actively intended to be read—or not, as the case may be—are identified in the notes when their special character is not otherwise apparent from the transcription. But deletions accomplished by unusual methods are simulated whenever possible, for the methods themselves often convey some such intention (see, for example, SLC to Olivia L. Langdon, 6 and 7 Sept 69click to open link). And in some letters, Clemens used two methods of cancellation which occupy opposite ends of the spectrum of difficulty.

Mock, or pretended, cancellations are words crossed out so lightly that they are easily read, visibly distinct from normal deletions, as well as being (for the most part) deletions of words still necessary to the sense. Clemens used various methods for creating mock cancellations, but the current rendering does not distinguish them. Clemens also deleted parts of some letters by tearing away portions of the manuscript page, which he then sent, visibly mutilated. By their very nature, such deletions are unlikely to be read by anyone, but occasionally Clemens left enough evidence in the torn page to permit as much as the first or last line of the suppressed passage to be reconstructed. Yet even if the entire excision somehow survived, it would not be included in the transcription, simply because it was not part of the letter he sent. When text canceled in this fashion can be reconstructed, therefore, it may be transcribed with wholly missing characters as diamonds and partly missing characters as normal alphabetical characters, bracketed as interpolations: ‘I j◇◇◇ ros◇ ◇p &’ (SLC to Olivia L. Langdon, 27 Feb 69click to open link). The result is not, in the ordinary sense, readable—any more than the original manuscript at this point was, except where one or two characters or words left standing could still be read, out of context. The fully legible reconstructed reading is, therefore, given only in the notes.

It may be added here that some deletions in manuscript, especially of punctuation, were indicated there only by methods not themselves transcribable. For instance, when Clemens added a word or more to a sentence already completed, he rarely struck out the original period. Instead, he signaled his intention simply by leaving only the usual word space between the original last word and the first word of his addition, rather than the larger space always left following a sentence period. Whenever someone reading the manuscript would have understood something as canceled, even though it was not literally struck out, the transcription represents it as if it had been deleted in the normal fashion, and the record of emendations reports the fact as an implied deletion.

We emend original documents as little as possible, and nonoriginal documents as much as necessary, but we emend both kinds of copy-text for two fundamental reasons: to avoid including an error, ambiguity, or puzzle that (a) is not in the original, or (b) is in the original, but cannot be intelligibly transcribed without altering, correcting, resolving, or simplifying it.

Errors made by the writer are not emended so long as they can be intelligibly transcribed. Some few errors of omission may be corrected by editorial interpolation—that is, by supplying an intended but omitted character, word, or words within editorial square brackets, ‘thus’ or ‘“thus”’—but only when the editor is confident that the writer has inadvertently omitted what is thus supplied. Interpolated corrections may be necessary to construe the text at all, let alone to read it easily, and would therefore be supplied by any reader if not supplied by the editor. They are thus a logical extension of the decision to transcribe errors when (and only when) they can be intelligibly transcribed. Interpolated corrections do not conceal the existence of error in the original, and are therefore not emendations of it: like editorial description, or superscript numbers for the notes, they are always recognizably editorial, even when they enclose a conjecture for what the writer meant to but did not, for whatever reason, include in the letter sent. Interpolations are therefore not normally recorded in the textual commentaries. Interpolations are not always supplied, even if what is missing seems beyond serious doubt, nor could they be used to correct all authorial errors of omission: mistaken ‘is’ for ‘it’, for example, or a missing close parenthesis that must remain missing because it might belong equally well in either of two places.

Most errors in a nonoriginal copy-text, such as a contemporary newspaper, are attributable not to the writer, but to the typesetter, and are therefore emended. Yet even here, certain grammatical errors and misspellings may be recognizably authorial, and therefore not emended. On the whole, however, Clemens’s precise and meticulous habits, which were well known in editorial offices even before he left the West, make it more rather than less likely that errors in such a printing are the typesetter’s—especially because editors and typesetters were typically committed by their professions not to a literal transcription, but to a “correct” form of any document they published. Typesetting errors are self-evident in such things as transposition (‘strated’ for ‘started’), wrong font (‘carriEd’), and some kinds of misspelling (‘pouud’). In addition, we know that by 1867 Clemens consistently wrote ‘&’ for ‘and’ in his letters—except where the word needed to be capitalized, or the occasion was somewhat more formal than usual (for example, see SLC to Charles Warren Stoddard, 27 April 67click to open link). In any nonoriginal copy-text, therefore, ‘and’ is sure to be a form imposed by the typesetter, who had good professional reasons for excluding ‘&’ as an unacceptable abbreviation. The word is therefore always emended as an error in nonoriginal copy-texts.

But if authorial errors are preserved uncorrected, it may well be asked why it is ever necessary to emend originals to avoid including them, not to mention how this can be done without changing the meaning of the original letter, and therefore the reliability of the transcription.5explanatory note The general answer to these questions is that in a transcription which does not reproduce the text line for line with the original, some forms in the original must be changed if they are not to assume a different meaning in the transcription—in other words, if they are not to become errors in it. Clemens’s characteristic period-dash combination at the end of a line is a classic example of something that must be emended because it would become an error if literally transcribed. The period-dash apparently combined as terminal punctuation in Clemens’s manuscripts virtually always occurs at a line end, at least until about the mid-1880s, when he seems to have trained himself not to use the dash there, probably because contemporary typesetters so often misinterpreted his manuscript by including it in the type, where it would appear as an intralinear dash between sentences. The typographical origin of this device was probably as an inexpensive way to justify a line of type (especially in narrow measure, as for a newspaper), but Clemens would certainly have agreed with the majority view, which frowned upon the practice.6explanatory note As already suggested (see ‘dashes’ above), when Clemens used a dash following his period, he indicated simply that the slightly short line was nevertheless full, and did not portend a new paragraph. The device may owe something to the eighteenth-century flourish used to prevent forged additions in otherwise blank space, since it sometimes occurs at the end of short lines that are followed by a new paragraph. At any rate, he never intended these dashes to be construed as punctuation. Yet that is precisely what happens if the typesetter or the reader does not recognize the convention and reads it as a normal dash, signifying a pause. Any dash following terminal punctuation at the end of a line is therefore not transcribed, but emended. When ‘period.— | New line’ occurs in a newspaper or other transcription of a lost letter, it doubtless reflects the typesetter’s own use of this method for right justification, and is necessarily emended. And when ‘period.—Dash’ occurs within a line in such a printing, it is almost certainly the result of the typesetter’s misunderstanding the convention in Clemens’s manuscript, and is likewise emended.

Ambiguities left by the writer are also not emended if they can be intelligibly transcribed. But both original and nonoriginal copy-texts will inevitably contain ambiguous forms that, because the transcription is not line for line, must be resolved, not literally copied. Ambiguously hyphenated compounds (‘water-|wheel’), for example, cannot be transcribed literally: they must be transcribed unambiguously (‘waterwheel’ or ‘water-wheel’), since their division at a line end cannot be duplicated. Using the editorial rule ( | ) to show line end would introduce a very large number of editorial signs into the text, since consistency would oblige the editor to use the symbol wherever line endings affected the form in the transcription. Even noncompound words divided at the end of a line may sometimes be ambiguous in ways that cannot be legibly preserved in the transcription: ‘wit-|ness’ in the copy-text must be either ‘witness’ or ‘witness.’ Dittography (of words as well as punctuation) likewise occurs most frequently at line ends—physical evidence that makes it readily intelligible as an error in the source, but that is lost in a transcription which abandons the original lineation. Dittography becomes more difficult to construe readily when it is simply copied, because the result is at least momentarily ambiguous. It is therefore emended, even in intralinear cases, in order not to give a distorted impression of this overall class of error. The general category of manuscript forms affected by their original position at line ends, however, is even larger than can be indicated here.

Puzzles created by the writer are likewise preserved if they can be intelligibly transcribed. On the other hand, we have already described several aspects of the author’s alterations in manuscript which would, if transcribed, introduce puzzles in the transcription: the method of cancellation, errors with a physical cause, implied deletions, self-corrections that would masquerade as revisions, and changes internal to a word which the editor may simplify. These alone show that holograph manuscripts invariably contain many small details which we simply have no adequate means to transcribe.

The question posed by such details is not simply whether including them would make the text more reliable or more complete (it would), but whether they can be intelligibly and consistently included without creating a series of trivial puzzles, destroying legibility, while not adding significantly to information about the writer’s choice of words or ability to spell. There are, in fact, a nearly infinite variety of manuscript occurrences which, if transcribed, would simply present the reader with a puzzle that has no existence in the original. For instance, a carelessly placed caret, inserting a phrase to the left instead of to the right of the intended word, may be readily understood in the original, but can be transcribed literally only at the cost of complete confusion.

Exceptional Copy-Texts. When the original documents are lost, and the text is therefore based on a nonoriginal transcription of one kind or another, the normal rules of evidence for copy-text editing apply. When, however, two transcriptions descend independently from a common source (not necessarily the lost original itself, but a single document nearer to the original than any other document in the line of descent from it), each might preserve readings from the original which are not preserved in the other, and these cannot be properly excluded from any text that attempts the fullest possible fidelity to the original. In such cases, no copy-text is designated; all texts judged to have derived independently from the lost original are identified; and the text is established by selecting the most persuasively authorial readings from among all variants, substantive and accidental. Before this alternative method is followed, however, we require that the independence of the variant texts be demonstrated by at least one persuasively authorial variant occurring uniquely in each, thereby excluding the possibility that either text actually derives from the other. If independent descent is suspected, even likely but not demonstrable in this way, the fact is made clear, but whichever text has the preponderance of persuasively authorial readings is designated copy-text, and the others are treated as if they simply derived from it, whether or not their variants are published.

Damaged texts (usually, but not necessarily, the original letters) are emended as much as possible to restore the original, though now invisible, parts of the text that was in fact sent. This treatment of an original document may seem to be an exception to the general rule about emending originals as little as possible, but a damaged manuscript is perhaps best thought of in this context as an imperfect copy of the original. And despite some appearance to the contrary, emendation in such cases is still based on documentary evidence: sometimes a copy of the original made before it was damaged, or damaged to its present extent—more commonly, evidence still in the original documents but requiring interpretation, such as fragments of the original characters, the size and shape of the missing pieces, the regularity of inscribed characters (or type) and of margin formation, the grammar and syntax of a partly missing sentence, and, more generally, Clemens’s documented habits of spelling, punctuation, and diction. This kind of evidence cannot establish beyond a reasonable doubt how the text originally read. Its strength lies instead in its ability to rule out possible readings, often doing this so successfully and completely that any conjecture which survives may warrant some confidence. At any rate, we undertake such emendations even though they are inevitably conjectural, in part because the alternative is to render the text even less complete and more puzzling than it is in the damaged original (since sentence fragments are unintelligible without some conjecture, however tentative), and in part because only a specific, albeit uncertain, conjecture is likely to elicit any effort to improve upon what the editors have been able to perform. For this same reason, a facsimile of any seriously damaged document is always provided.

Letters and, more frequently, parts of letters that survive in the original but cannot be successfully transcribed constitute another exception and will be published in facsimile. For example, four letters that Clemens typed in 1874 (joking the while about his difficulties with the typewriter) clearly exceed the capacity of transcription to capture all their significant details, particularly the typing errors to which he alludes in them. Partly because they were typed, however, the original documents are relatively easy to read and therefore can be published in photographic facsimile, which preserves most of their details without at the same time making them any harder to read than the originals. These are true exceptions in the sense that most of Clemens’s typed letters can and will be transcribed. But it is generally the case that facsimile cannot provide an optimally reliable and readable text, even of Clemens’s very legible holograph letters, which comprise at least eight thousand of the approximately ten thousand known letters.

Yet by the same principles which justify transcription of most letters into type, facsimile should serve to represent within a transcription most elements of a manuscript which would (a) not be rendered more clearly, or (b) not be rendered as faithfully by being transcribed (newspaper clippings, for instance)—or that simply cannot be faithfully transcribed, redrawn, or simulated (drawings, maps, rebuses, to name just a few of the possibilities). It follows that if an original newspaper clipping enclosed with a letter cannot, for any reason, be reproduced in legible facsimile, it will be transcribed line for line in what approximates a type facsimile of the original typesetting. If the text of an enclosed clipping survives, but no example of the original printing is known, it is transcribed without simulating the newspaper format. And long or otherwise unwieldy, or doubtful, enclosures may be reproduced in a separate document.

Letters which survive only in the author’s draft, or in someone else’s paraphrase of the original, are also exceptions. In the first case, the source line of the editorial heading always alerts the reader that the text is a draft. In such cases, emendation is confined to those adjustments required for any original manuscript, and is not designed to recover the text of the document actually sent, but to reproduce the draft faithfully as a draft. Likewise, if a letter survives only in a paraphrase, summary, or description, it is included in the volume only if the nonoriginal source is judged to preserve at least some words of the original. And like the author’s draft, it is not necessarily emended to approximate the original letter text more closely, since its nonauthorial words usually provide a necessary context for the authorial words it has, in part, preserved. When it is necessary to interlard paraphrase with transcription, the paraphrase appears in italic type and within editorial brackets, labeled as a paraphrase, in order to guarantee that there will be no confusion between text which transcribes the letter and text which does not pretend to. When nonoriginal sources use typographical conventions that are never found in Clemens’s manuscripts—word space on both sides of the em dash, for instance, or italic type for roman and vice versa—the normal forms of the lost manuscript are silently restored.

Silent Emendations. In addition to the method of cancellation, which is usually omitted from the transcription and the record of emendations, several other matters may involve at least an element of unitemized, which is to say silent, change. To save space, we transcribe only routine addresses on envelopes by using the vertical rule ( | ) to signify line end; nonroutine text on envelopes is transcribed by the same principles used elsewhere. The text of preprinted letterhead is reproduced in extra-small small capitals, usually in its entirety, but as fully as possible even when unusually verbose, and never to an extent less than what Clemens may be said to adopt or refer to (‘I’m down here at the office’). Only substantive omissions from letterhead are reported as emendations, since the decorative variations of job type are literally indescribable. Postmarks are also transcribed in extra-small small capitals, but only unusual postage stamps are transcribed or described. Whenever Clemens used any of the following typographical conventions in his original letter (hence also whenever they occur in nonoriginal copy-texts and are deemed authorial), the transcription reproduces or simulates them, even when it is necessary to narrow the measure of the line temporarily, which is done silently: diagonal indention; hanging indention; half-diamond indention; squared indention; the flush-left paragraph and the half line of extra space between paragraphs, which is its collateral convention; text centered on a line, positioned flush right, or flush left; and quotations set off by quotation marks, indention, reduced space between lines (reduced leading in type), extra space above or below (or both), smaller characters in manuscript (smaller type in nonoriginals), or any combination of these conventions.

Normal paragraph indention is standardized at two ems, with variations of one em and three ems often occurring in the same letter. We silently eliminate minor, presumably unintended variation in the size of all indentions, and we place datelines, complimentary closings, and signatures in a default position, unless this position is contradicted by the manuscript, as when extra space below the closing and signature show that Clemens intended them to appear on the same line. But unmistakably large variation in the size of indention is treated as deliberate, or as an error, and reproduced or simulated, not corrected or made uniform. Notes which Clemens specifically did not insert within the letter text but wrote instead in its margin are nevertheless transcribed at the most appropriate place within the text, and identified by editorial description: ‘in margin: All well’, or ‘in bottom margin over ’. The editorial brackets in these cases may enclose just the editorial description, or both the description and the text described, depending on which conveys the original most economically. The only alternative to transcribing these notes where they are deemed “appropriate” is to transcribe them in a completely arbitrary location, such as the end of the letter. We likewise transcribe postscripts in the order they occur, even if this differs from the order they were intended to be read, so long as the intended order remains clear. Thus a marginal ‘P. P. S.’ can intelligibly precede a ‘P. S.’, just as a ‘P. S.’ inserted at the top of a letter can precede the letter proper, whether or not it was actually intended to be read first. But if, for example, a postscript inserted at the top is written across or at right angles to the main text—a sign it was not intended to be read before or with the text it crosses—the intended order must prevail over the physical order, and the postscript is therefore moved to the end of the letter. Only changes in writing media are noted where they occur in the text, as in ‘postscript in pencil’, from which it may also be reliably inferred that all preceding text was in ink. Line endings, page endings, and page numbers are silently omitted from the transcription, but where they affect the text or its emendation, they are given in the record of emendation.

Each commentary usually has four sections but may have as many as five (sections are omitted when there are no facts to report). □ Copy-text identifies the document or documents that serve as the basis for the transcription, and from which the editors depart only in the specific ways listed as emendations. □ Previous publication cites, in chronological order, all published forms of the letter known to the editors, indicating which of these (if any) are notably incomplete or erroneous. The record of publication given here is not necessarily exhaustive, since the aim is to identify where and when a letter was first made generally accessible. Publications frequently referred to are described in the Description of Textsclick to open link. □ Provenance reports what the editors have been able to learn about an original letter’s history of ownership. Under this heading the reader may often be referred to the history of the relevant collection of manuscripts, as given in Description of Provenanceclick to open link. □ Emendations and textual notes lists all deliberate departures from the copy-text (barring only changes categorized here as silent emendations). But the list may also include (a) editorial refusals to emend, identified by ‘sic’, and (b) textual notes (always italicized and within square brackets) which explain the reasoning behind particular editorial decisions to emend, or not, as the case may be. When no copy-text has been designated because two or more documents descend independently from the lost original, all variants are recorded and identified by abbreviations defined under Copy-text, and this section is renamed Emendations, adopted readings, and textual notes, to signify that variant readings have been chosen on their merits from two or more authoritative texts. □ Historical collation lists variants between two or more nonoriginal documents that may have descended independently from a common source, but have yielded no conclusively authorial variants.

All entries in these lists begin with the reading exactly as it stands in the transcription, except where indention ¶, line ending ( | ), or abbreviation (‘Write . . . is’) is necessary. As far as possible, entries are confined to the words and punctuation being documented. Line numbers include every line of letter text on a page, even when the page contains text for more than one letter, including all rules, all enclosures, and all lines that are wholly editorial, such as ‘about one page (150 words) missing’, ‘in pencil’, the editorial ellipsis ( . . . . ), or the full-measure envelope rule. Line numbers exclude all editorial matter in the letter headings and in the notes. Each reading is separated by a centered bullet ( • ) from the corresponding reading of the copy-text, transcribed without change or emendation, insofar as our notation permits, or described within brackets and in italic type, as necessary.