University of Virginia Library

Transcription of Manuscripts: The Record of Variants
by
Fredson Bowers

The present discussion is limited to special problems of editing transcriptions made of manuscripts in a non-facsimile manner and of recording corrections and revisions in the texts. The material on which this study is based comprises a miscellaneous group of nineteenth- and early twentieth-century manuscripts of Nathaniel Hawthorne, Walt Whitman, Stephen Crane, and William James. No particular magic inheres to this period or to these authors; moreover, it is fortuitous that they are American. The manuscripts, mostly holographs together with a few typescripts, were greatly altered in the course of inscription and often in later rounds of independent revision. So far as can be determined, however, the methodology I propose is equally applicable to scribal versus holograph manuscripts and to revised typescripts. Morever, it may seem probable that the system proposed would be applicable to many manuscripts of other periods and languages. However, I prefer to write of what I know from recent firsthand experience; thus I draw my illustrations (when they are not invented for convenience, and noted as 000.00) from the autograph manuscripts of William James, which present a useful conspectus of problems owing to James's custom of rapid composition and extensive revision.

Page-line references are to the ACLS edition by the Harvard University Press (1975) of The Meaning of Truth. The earlier examples are given in comparatively full form so that the recommended abbreviations, which appear later, can be understood. For many situations there exist various optional ways of dealing with the problem, many of which are illustrated. Once an editor sees the principles behind the options he can make choices according to his own preferences for clarity or economy and construct his own consistent system of notation. In certain cases a prefixed asterisk indicates the present writer's own recommendations.

General methods of transcription divide neatly in two:

I. Clear-text Transcription and Apparatus

The finally revised form of the text is transcribed diplomatically; that is, the transcription exactly follows the forms of the manuscript in spelling, punctuation, word-division, italics (for underlining), and capitalization, but not in matters of spacing or in line-divison, nor is a facsimile visual presentation of alterations attempted. Only a few of the problems that are inevitably encountered in engaging oneself to relatively exact transcription can be considered here. For instance, it is an open question whether unusual idiosyncrasies of no possible textual interest should be followed when their reproduction would add to the difficulty of a reading text. Although in general a scholarly transcript should follow an author's abbreviations or informal shorthand, with or without following periods, some personal peculiarities might be modified. For example, I have seen an early manuscript by Edith Wharton in which a mark about the size of a hyphen customarily does duty for a period. In my own view, it ought to be possible for a transcriber to print these as periods, with notice. The early twentiethcentury philosopher John E. Russell, as in his letters to William James, not only often used such hyphens for periods and even for commas but sprinkled his text with them where no punctuation could possibly have been intended. Transcripts of these letters would have been close to unreadable if draconian editorial omissions of these latter, and various emendations of the former, had not been made.

Most printers may be unable, without expensive handsetting, to reproduce a manuscript's positioning of a period in an abbreviation directly under a superscript letter, in which case commonsense economy suggests that the period be silently moved to follow the letter, as in 'Mr.' or 'Co.' Similar difficulties may be encountered in placing acute or grave accents above a full capital, but here I believe the manuscript must be followed and handsetting resorted to if the use of an accent above a small capital is not possible to get around the difficulty. If one's edition is printed by photo-offset of type and not from type-metal,

then there is no problem, for all difficult characters and positioning can be drawn in before photographing.

It is impossible to discuss here the numerous problems and the occasional modifications required to deal with authorial orthographic idiosyncrasies and to assess the actual intentions. Some authors form a few minuscules (lower-case letters) very much like majuscules (capitals). Ordinarily the context and syntactical position in the sentence prevent confusion, and it would be most unwise to transcribe an obviously intended lower-case initial letter as a capital just because of its formation. However, problems may arise in distinguishing minuscules and majuscules when capitalized personifications or conceptual nouns are possible, or in dealing with variant usage in words of region like east or East, and so on. Some writers intend to make a single word in forms like anyone or someone and may often link the two but on occasion write them separately but with a shorter space between them than is usual between two words, in which case the forms should be transcribed as single words.

The position of punctuation in relation to quotation marks is particularly difficult to transcribe since some authors have no fixed principle and may place a comma, say, inside or outside a quotation mark almost at random on those occasions when it is not written directly under the mark. Not only would it be expensive to try to reproduce in type punctuation directly under quotation marks but the attempt would be in vain since such authors will also vary the positioning slightly so that a transcriber would need to make dozens of arbitrary decisions whether punctuation were by a hairline to the left or right, or under, all to no useful purpose in transcription. Arbitrary normalizing is the only answer. Ordinarily even a writer who is highly variable in his positioning of punctuation in relation to quotation marks will conform to some extent to the national tradition. In the United States, by the late nineteenth century at least (I am not historian enough to know when the shift was made) under the influence of the typographers a decision was made for arbitrary placement of commas and periods ending a quotation inside the quote marks but colons and semicolons outside. Question and exclamation marks were treated logically: inside or outside according to the meaning. In Great Britain, on the contrary, although the treatment of colons and semicolons has lately had a tendency to agree with the American custom,[1] commas and periods in strict usage are placed inside or outside quotation marks (like question and exclamation marks) according as they form part of the actual quoted matter or else as they belong, instead, with the syntactical pointing of the sentence. If a writer has some tendency to observe one or other system, a transcriber may be forced to impose it on him throughout, without record, although a general statement to this effect ought always to be made. But relatively consistent idiosyncrasies should be observed as the norm.

A more subtle problem may arise not only in the transcription of the correct font of punctuation but also in seeing that the modern printer observes the transcriber's intentions. When words are italicized in manuscripts by underlining, the author may on occasion carelessly extend the underline below a following punctuation mark, thus appearing to italicize it. Printing practice through the nineteenth and twentieth centuries has been highly variable in this matter of the font of punctuation to be used following an italic word when a roman word succeeds. Ordinarily there can be no question that within an italic passage all punctuation should conform to the italic text. Logically, in

all other circumstances the font of the punctuation should be roman or italic depending upon its syntactical relationship to the font of the sentence as a whole or else to the italicized matter. For instance, in the commonest cases when one or more words are italicized and a syntactical unit of the roman sentence requires a punctuation mark after the last italicized word, the sentence then continuing in roman, the font of the syntactical punctuation must be in roman since it applies not to the italicized matter but to the roman sentence, as in a sentence, 'What we know is this: the correlative . . . .', or 'If we write the word truth, we should mean . . . .' Quotation marks enclosing an italicized word or phrase in a roman sentence should be in roman, and so on.[2] In some cases the matter is not merely a typographical one, since meaning can be involved. Of course, the logical system here described has its basis in meaning, but vide Orwell some situations may be more meaningful than others. A roman interrogative sentence, for instance, may end with an italic word and an exclamation or question mark. If the mark is in roman, the reader is instantly made aware that the sentence as a whole is the exclamation or the query but not the italicized word(s). Similarly, if the roman sentence is declarative but the italic word (or words) ending the sentence constitutes a question, the italic question mark (or exclamation) satisfactorily transmits this sense.

For a time the best modern printers maintained this distinction about all roman and italic punctuation when their copy gave them any encouragement, but the present-day printer is growing careless and is as likely as not to set italic punctuation after any italic word regardless of the syntactical logic, and—worse—regardless of the exact underlining of his copy. A scholar transcribing manuscripts, therefore, is faced with a decision which is most easily solved by just such an arbitrary rule (after a general statement to the effect) as was made about the positioning of punctuation vis-à-vis quotation marks; that is, roman or italic punctuation should be marked in the transcript according to the font that is appropriate for meaning according to the logical relationship of the punctuation to the text.

The further problem of forcing one's modern printer to follow exactly the extent of the underline in one's transcript must then be met; usually general instructions are insufficient since they do not get transmitted to all the compositors who may work on a book, and in addition to specific instructions one must mark the typescript with warning signs about the font in any case (which is any case) when the printer is likely to go astray.

Diplomatic transcription and setting into type produce another particular problem in the word-division. Many authors are irregular in their division: the same word may be written as one word, as two, or as a hyphenated compound. Even if an author is scrupulous in his division, a word broken at the end of a line by a hyphen may create a problem of interpretation whether it should be transcribed as a single word or a hyphenated compound.[3] Unfortunately, the clearest way of

recording such words is impracticable; that is, it would be possible to transcribe them as broken, with a vertical stroke (or less preferably a slant) indicating the line-ending, as in child-|like. (Note that words broken at the end of a line, as marked, should always be set close-up.) The difficulty is that such scattered line-ending notations are obtrusive and could be subject to misinterpretation because of their use only on special occasions.[4] It may seem that this is another case where some editorial decision is needed to govern the form used in the transcription. Obviously, if there is no question that the word must or must not be a hyphenated form, one should transcribe it correctly and in silence. If the phrase vis-à-|vis or the number thirty-|four or a capitalized compound like neo-|Platonic were broken as marked, no problem exists, nor need there be any problem about broken compounds like eye-|sensations, world-|romance, or anti-|absolutist where a single word would be practically impossible. If an author can be shown to be consistent, a hyphenated form broken at the end of one of his lines can be transcribed silently according to his invariable custom. If he always writes to-day, then when to-|day is found it is unlikely to be today. Analogy can be misleading in determining authors' practice in such matters, however, and one must make sure of the invariable treatment of the word itself: with William James, for example, co-|exist can be transcribed coexist with confidence; on the other hand, pre-|exist or co-|ordinate might as easily be pre-exist and co-ordinate as preexist and coordinate and could not be transcribed as one or the other silently. Since it is not advisable to mark merely such problem words in the text with a line-ending stroke, an editor must transcribe one or other form according as he estimates the odds favor his choice, but the facts should be noted so that a user of the transcript may be warned that the forms of certain words have been editorially interpreted. Moreover, the printer setting the transcript into type will perforce break words that could cause a reader some perplexity about their form in the manuscript. These cases are easily enough handled by a word-division list in three sections as part of the apparatus.

In the first section it is convenient to state that all end-of-the-line hyphenated forms in the modern typesetting are to be taken as unbroken single words unless specifically noted otherwise. Thus a reader can readily recover the exact hyphenation or non-hyphenation of the manuscript when the form occurred within a line, and so was not subject to dispute, by knowing that all hyphenated compounds of

this kind that happen to be broken at the end of the line by the modern printer will be specified in the apparatus-list. The form there is best given visually as it occurs in the printed text:

• 000.00 pre-|empted
• 000.00 well-|being
• 000.00 co-|implicated
• 000.00 after-|thought

The presence of these words in the list guarantees that they do not come under the general rule that hyphenated broken words in the modern print are to be ignored; these would have read in the manuscript pre-empted, well-being, co-implicated, and after-thought.

A second list in the apparatus can treat doubtful words broken in the manuscript at the end of a line, with the form the transcriber has chosen to print in his text, either hyphenated or unhyphenated. The presence of these words indicates to the reader that an editorial decision has been made (unbroken words would have been broken in the manuscript only at the point where ambiguity about their form could exist):

• 000.00 pre-empted
• 000.00 precondition
• 000.00 coexist
• 000.00 re-compounding

A third list should contain the more uncommon occurrences of doubtful words broken at the end of a line in the manuscript and also in the modern print. In these cases the editor will give the forms he thinks should be transcribed if a reader were to copy out the passage containing the words:

• 000.00 pre-|empted (i.e., pre-empted)
• 000.00 co-|exist (i.e., coexist)

Needless to say an editor should exercise some discrimination in the selection of words for recording in this apparatus and need not list obvious cases where there could be no dispute about the author's intentions. If an editor prefers, these listed words could be included instead in the general apparatus. However, readers will find them most convenient to use if they have been set apart in a separate section for emphasis and not buried in a mass of other matter with which they have little in common. (For the forms of such lists and their headnotes, see the Centenary Edition of Nathaniel Hawthorne, the University of Virginia edition of Stephen Crane, and the ACLS edition of William James for the Harvard University Press.)

Whether letters or numbers are used, footnote apparatus indicators in a reading text are a nuisance for the majority of readers of a clear text, who do not at all times wish to consult the apparatus for a view of the growth and revision of the text. An editor is best advised to key his apparatus to the page-lines of his own printed text, with or without marginal line numbering.

Correspondingly, the apparatus of variants in a clear-text edition is best placed in the back of the book and not at the foot of the pages unless the text relies for its main interest on the account of the alterations so that readers will require them in the most convenient place for reference—certainly the foot of the page.[5] The simplest way of keying is to give the page number first, followed by a period and then the line number(s). For economy, similar entries within some unit like a chapter, or within a reasonable distance of each other, may be brought together: 243.18,20 will] interlined above deleted 'would'; or 243.18;244.20 will] above del. 'would'. In the second example the semicolon is required to distinguish separate pages, whereas occurrences on the same page are separated by commas: 243.18,20;244.31. It is economical of space to mark these numbers for close-up setting.

Formulas for describing alterations are an essential, and it is important that they be as descriptive as possible. The entry starts, after the page-line reference,[6] with the lemma to the left of the square bracket. This lemma is ordinarily the exact word or words found in the text. If anything longer than a word or two is being recorded in the lemma, space can be saved by the system of three dots (or four for a sentence end) to indicate omitted material that will be found in the text: 000.00 As . . . words]. The general rule is that punctuation following a word in the text is not noted in the lemma before dots or at the end of the lemma unless it is a part of the variant. For instance, in 000.00 list] final 's' deleted it is of no concern whether the text word read 'list,', 'list.', or 'list' with no punctuation so long as the punctuation was not affected by the alteration. (However, if a complete sentence forms the lemma, it is convenient for clarity to provide the closing period, with another in the description.) On the other hand, a comma after list would be required in the lemma in such cases as these and in any others that affect the punctuation:

In the above examples, the lemma is itself the alteration. However, some alterations—principally deletions—will not be present in the text and need to be keyed by reference to another word acting as the lemma: this] followed by deleted 'notion'; or: is] preceded by deleted 'notion'. Whether followed by (before) or preceded by (after) is used may be dependent on two factors. First, a word distinctive enough to be readily located in the text-line by a reader is more useful for positioning an entry than a neutral and commonplace word too easily overlooked. For this reason, a, the, is, and the like are less desirable as lemmata than other parts of speech. It is better to write: reflection] after deleted 'thought' (than) the] before deleted 'thought'. Moreover, it is simpler for a reader to follow a revising author's chain of thought if an alteration can be positioned by a lemma that illuminates the alteration. For instance, if an author first writes reflection but then deletes it and adds thought, the association is better shown by: thought] after deleted 'reflection' (than by) signifies] before deleted 'reflection'. In any clash of interests the second factor is subsidiary to the first, for quick identification is certainly of primary importance.

Occasionally identification of a lemma is obscured by the duplication of the identical word in the same line. The double occurrence must be taken account of in one's typescript of the apparatus made up from the temporary lining of the text before printing so that the exact word can be identified when the text is received in page proof, with different lineation. Thus in a case where the word is was inserted, and in one's copy-text (or in one's typescript) another is appears in the same line, one may adopt either of two means of identification. The most economical is to distinguish the lemma by a superscript figure depending upon whether it is the first, the second, etc., same word in the line: 1is] interlined (or) 2is] interlined. Especial care is needed later to check whether in the page proof is is in fact repeated in the

same line and if so whether the original superscript figure still remains accurate. If the repetition is not present in the proof of the text, then of course the figure is removed before the apparatus is typeset.[7] (Also, as a part of the rekeying of the typed apparatus to the different lining of the final text in page proof, any lemma must be checked to see that a repeated form has not crept in the proof [though not present in the copy-text] that will require identification by superscript figure.) Although more wasteful of space, positive identification of a lemma may also be made in certain circumstances by adding a word to the lemma or to the description that will distinguish it from the same word elsewhere in the line:

When an apparatus is being constructed not for a manuscript text that is being critically edited as in Stephen Crane's The O'Ruddy or The Red Badge of Courage but instead for a manuscript associated with an edition of a printed text, as in a revised book version that has been chosen as copy-text over the manuscript, some special problems arise about the lemmata. In William James's Meaning of Truth (1909), for instance, Chapters XIV and XV were heavily revised in the book from draft manuscripts that seem to have served as the basis for lost printer's-copy typescripts. Since in this case the further revised book was chosen as the copy-text for a critical edition (1975) in James's Works, the edited volume contains as apparatus a list of Emendations to the Copy-text, an Historical Collation of all rejected readings (accidental as well as substantive) in the manuscripts, and a list of Alterations in the Manuscripts.

The Historical Collation is made up from the finally revised readings of the manuscripts and (combined with the Emendations) enables a reader to reconstruct in every detail the manuscripts' revised texts. It is, then, the function of the Alterations list to complete the apparatus for the reconstruction of the manuscript by listing every alteration, whether of correction or revision, that James made during composition and review. Since the book-text often varies from the manuscript, a difficulty arises in the form of the lemmata of the list of Alterations which should key the readings of the two documents, for the lemma of a revised manuscript reading may not be present in the still further revised book-text.

These differences being of various kinds yield to various methods of treatment. In the first paragraph of Chapter XIV The Meaning of

Truth book-text reads but it entirely fails to hit the right point of view. In the manuscript hit is also the final reading, though revised from an earlier word, and so it is simple to write: 146.3 hit] interlined above deleted 'place himself at'. On the other hand, the book's it before entirely was originally he in the manuscript, and this he has been deleted without substitution. The notation is simple enough: 146.3 entirely] after deleted 'he'; nevertheless, there are elements of ambiguity present for a reader engaged in reconstructing the manuscript from references to the book-text in the Alterations: it is theoretically possible, of course, that MS could have read it before deleted he. True, if this had been the reading, the positioning of the entry for deleted he by the use of the key entirely would have been misguided on the part of the editor, for a normal entry would have read, more helpfully: it] before deleted 'he'. Moreover, if it had been squeezed in as an insertion before deleted he, or interlined as a substitute, the entries would have read respectively: it] inserted before deleted 'he' (or) it] above deleted 'he'. Thus an expert reader with some difficulty might figure out from the nature of the entry that it was not present in MS, and if he needed confirmation he could secure it from the Historical Collation where the absence of the word in MS would have been noted: it] omit MS. Still, even though it is obviously impossible to reconstruct a manuscript from an apparatus without the simultaneous use of the Historical Collation and the list of Alterations, an editor should be as helpful as he can in assisting a reader to find his way about; thus a dagger placed before the page-line reference informs the reader that in some respect the particular MS listing differs from the book reading in the lemma and he had better consult the Historical Collation (HC) to understand the difference. By this convention, the Alterations entry would read: †146.3 entirely] after deleted 'he'.

The virtue of the daggered entry is that it utilizes a lemma keyed to the book-text—one either identical with or readily identifiable in the Historical Collation—and yet conveys the essential information without elaborate description of manuscript variants that have no direct pertinence to the main point of the entry. For instance, in The Meaning of Truth at 155.7-8 (references are to the Works) the book copy-text reads But probably your intent is something different; so before I say . . . . In the manuscript this is an interlineation above deleted But before saying, but the manuscript inscription arrived at the book-text wording only after some revision of the original form of the interlineation. If the final form of the manuscript had been identical with that of the book, a regular entry could have been contrived without difficulty; however, there happens to be one accidental variant

in that the manuscript places a comma after so whereas no comma is present in the book. The essential fact a reader wants to know from the Alterations is that this passage was an interlined revision of a brief phrase; second, that it was also revised during the course of interlineation. That in the MS a comma is found that is not present in the book has nothing to do with the central fact of interlineation. The convention of the dagger deals with this situation very efficiently by conveying to the reader all the major information about the interlineation and its revision but with the warning that some detail, unessential to an understanding of the central facts of revision, has not been noticed in the manuscript description (since although part of the major alteration it was not itself an independent change that required listing). The reader can then identify this detail (or details) in the Historical Collation. The Historical Collation, to which the dagger entry directs the reader, provides the one point of variance between the book-text and the final readings of the MS: 155.8 so&c.rat;] ~, MS.

A substantive variant is present in the next example. The book reads proposition is one the whereas the MS interlined substitute for a deleted earlier reading is proposition means one the. A normal entry would read: 146.5 proposition . . . the] MS 'proposition means one the' above deleted 'idea is one whose'; but the daggered entry saves space: †146.5 proposition . . . the] above deleted 'idea is one whose'. By referring the reader to the Historical Collation (which would read: 146.5 is] means MS) for the variant, which is only a minor detail in the major fact of the interlined revision, the need to quote the passage in order to note the substantive variant is obviated.[9] A similar example is: †149.24-25 all . . . associates,] interlined. Here the book reads all such accidents as contents, implications, and associates whereas the MS has all such accidents as contents and implications, and associates.

These illustrations have concerned variants in the manuscript that were not represented in the words of the lemma; but so long as the situation is clear in the Historical Collation, a considerable flexibility is possible that saves valuable space even when variant words or accidentals are present between the lemma and the manuscript. In some cases, as will be shown below under the convention of the double dagger, it would be inefficient to refer the reader to the HC for a variant in which the manuscript encloses a word in single quotation marks but the book removes the quotes: †000.00 &c.rat;truth&c.rat;] interlined above deleted ''verity''. Such an expedient would prevent the necessity to quote, as in the normal entry: 000.000 &c.rat;truth&c.rat;] MS ''truth'' above deleted ''verity''. But, as will be suggested, this is the sort of case easily handled in another manner. On the other hand, in the next example the single-dagger entry is highly efficient. At 158.4-5 the book reads distinct from either the fact or whereas the MS interlined words are different from either the event or. The normal quoting entry would need to read: 158.4-5 distinct . . . or] MS 'different from either the event or' above deleted 'only another name for'; 'either' inserted. On the other hand, when the single dagger is used to refer the reader to the HC for the facts of verbal variance, a condensed entry can read †158.4-5 distinct . . . or] above deleted 'only another name for'; 'either' inserted. It is impossible to condense a quoting entry in the form 158.4-5 distinct . . . or] MS 'different . . . event or' above deleted 'only another name for'; 'either' inserted. The reason is that within single quotes all matter must be exactly transcribed, and literally such an entry would mean that in the manuscript the three dots were present between different and event. The case is altered in the Historical Collation, however, where the book-text is available to provide the full reading of the lemma and no single quotes limit the conventions of listing variants to the right of the bracket. Thus although two separate entries in the HC could be written, one for distinct[ different and the other for fact] event, an acceptable HC entry could pinpoint the only variants: distinct . . . fact] different . . . event MS. In some examples a line of apparatus might be saved, but not in all. This form would be especially useful in such variants as: is . . . is] was . . . was MS.

When an editor is concerned to save space, the single dagger referring the reader to the HC can be very useful if a somewhat

drastically reduced notation is thought to be acceptable. For example, a complete normal entry would read: 148.1 consequences] MS 'consequences that follow from their *nature [above deleted 'assumption']'. (For the use of the asterisk in this context, see below under the section on the transcription of variants without an apparatus.) In such a case the daggered entry †148.1 consequences] MS 'nature.' above deleted 'assumption.' by its reference to the HC would clarify the otherwise gnomic description, for the HC would read: 148.1 consequences] the consequences that follow from their nature MS. In a more complex example, with more widely divergent text, the book reads Perhaps the rising generation will grow up but the manuscript I must appeal to the rising general. Perhaps they may grow up, in which must was interlined above deleted shall. Ordinarily the Alterations entry would need to quote the whole in order to key in the single change from shall to must in an understandable context. However, as in the example from 148.1 above, this would unnecessarily duplicate the HC, which would read: 159.13-14 Perhaps . . . up] I must appeal to the rising general. Perhaps they may grow up MS. Thus it should be possible by the single dagger to save space in the Alterations by listing only the one change in the manuscript and referring the reader to the HC for the context: †159.13-14 Perhaps . . . generation] MS 'must' above deleted 'shall'. Special emphasis must be laid on the fact, however, that single-daggered entries of this nature in the Alterations must have the same lemma as that in the Historical Collation, else the necessary ready reference cannot be made.

In the convention of the single-dagger entries in the Alterations, the lemma in agreeing with or being easily recognizable in the Historical Collation must also agree with the edited text which the reader is using as the basis for following the changes made in the course of writing and revising the manuscript. When no dagger prefixes the page-line reference, the reader can be assured that the text of the book and of the final manuscript form is identical within the range covered by the Alterations lemma unless variants to the right of the bracket are specifically quoted and identified as MS readings not altered by revision. Thus it is the function of the single dagger to refer the user to the HC only for variant manuscript altered readings of words or accidentals (a) not printed in the Alterations lemma, or (b) not specified in the descriptive part of the entry as MS readings.[10] On the

other hand, a considerable amount of space devoted to quoting MS variants of a certain kind can be saved by the convention of the double dagger which is used to warn the reader that the lemma is not (as in every other circumstance) the reading of the book-text but instead that of the manuscript. Obviously, the two readings should be so similar that a person following the Alterations list in the book-text will be able to identify with certainty the reading that is intended. The simplest and most efficient use of this convention comes when only a single word or identical phrase is being noted but some accidental variant exists between the manuscript and book (the bracketed information in the examples below is not part of the apparatus but instead simply explains the situation):

• ††000.00 shall] above deleted 'does' [book read 'shall' roman]
• ††000.00 possesses] final 'es' added [book read 'possesses' italic]
• ††147.37 &c.rat;of course&c.rat;] interlined [book read', of course,']
• ††148.21 once equated&c.rat;] above deleted 'of the same value' [book read 'once equated,']
• ††154.12 Brother] ('B' over 'b'); after deleted 'dear' [book read 'brother']
• ††156.9 truth, and any] comma before deleted period; 'and any' above deleted 'Any' [book read 'truth. Any']

Some of these examples are more useful than others, but all join in revealing instantly to the reader of the book-text the nature of the variant without recourse to the Historical Collation. In this respect it is simpler to write italic shall as the lemma when the reader will instantly recognize the MS variant from the book-text roman without referring him to the HC, as in: †000.00 shall] above deleted 'does'. The normal entry 000.00 shall] MS 'shall' above deleted 'does' is certainly acceptable but takes more space and, when reference is being made to the book-text, is no more informative than the double-dagger entry. If 154.12 were to be written with a single dagger, it might be possible at small expense of space to note †154.12 brother] MS 'B' over 'b' etc. (The single dagger here is not strictly necessary but would be helpful since Brother has not been quoted as a word.) The punctuation variants are much more economically listed with the double dagger than with the single, which forces the reader to the further step of consulting the HC, and they are certainly more economical when listed in this manner than a normal entry with repetitive quoting that pinpoints the variance. Similarly, the limited substantive variant at 156.9 would be difficult to write with a single dagger without expanded quoting. Another simple example comes at 148.31 where following a question mark the book conventionally writes the first word of the next sentence as Likewise but the MS, in a more old-fashioned way for the start of still another question,[11] has likewise, which is an addition. The condensed entry ††148.31 likewise] inserted after deleted 'And' should cause a reader no difficulty at all.

How far to carry the convention of the double dagger beyond the relatively simple examples given is difficult to adjudicate, for its chief

usefulness certainly depends upon its simplicity. Nothing much is gained if to enable a reader to compare the manuscript reading in the lemma with the book-text in order to understand the variant being listed, the lemma must be expanded beyond normal. For instance, it seems easiest to use the single dagger, as illustrated, at 146.5 where the variant in an interlineation was MS means but book-text is (as in: †146.5 proposition . . . the] above deleted 'idea is one whose') than ††146.5 proposition means one the] above deleted 'idea is one whose'. (Note that dots for space-saving cannot be used in double-dagger entries when the MS variant would be omitted by this device. The lemma must contain the variant.) Allied with this difficulty is a more serious matter when the manuscript alteration to be noted is only part of a larger variant between MS and book. For instance, at 157.19 the book reads, in italic, in case he existed, which is roman in MS, with roman in case interlined above deleted roman if. If in case had been the only variant between the two documents, the double-dagger convention would handle the situation admirably; but if one were to write for the actual example ††157.19 in case] above deleted 'if', a reader following the book-text might well be puzzled about the status of the remaining italic. It is true that he could ascertain the facts by reference to the HC; but since the reference would be required contrary to the purpose of the double dagger, it may be thought simpler to refer to the HC for the whole entry by the single-dagger entry: †157.19 in case] above deleted 'if'. If the lemma is short, a moderate expansion required by the double dagger may still create the most efficient entry. At 148.32 the book reads in italic nicht wahr followed by a comma and roman etc., etc. In the manuscript the German words are in roman and by an alteration the original comma has been changed to a question mark and following etc., etc, deleted. If the entry were only 148.32 wahr] before question mark over comma and deleted 'etc., etc,', the problem of book italic nicht would remain, as in the example above at 157.19. But rather than solve the problem by the use of a single-dagger entry requiring consultation of the HC, it is perfectly simple to use the lemma ††148.32 nicht wahr] and one's troubles are over.

So long as the reader can easily identify the manuscript variant by comparing the double-daggered lemma against the book-text, uncomplicated variants can be noted. At 147.27 the book-text has reader but the manuscript reader himself inserted as an interlineation that is part of a larger variant. The single-dagger convention cannot handle this situation, but the double-dagger entry is efficient: ††147.27 reader himself] interlined. A normal entry is also acceptable though longer:

147.27 reader] MS 'reader himself' interlined. Another example is book prove the but MS interlined prove his, readily handled as: ††148.12 prove his] above deleted 'clinch the'. A borderline case where the altered word in MS is part of a more extensive variant is illustrated by 146.8-10 where the book reads an obvious absurdity, for that fact is the deliverance of a new proposition whereas for the whole passage MS has only which obviously is quite a new proposition. The MS alteration of obviously from original notoriously must be listed. Because of the peculiar nature of James's alteration, the description is a bit complex, but if we chose the generalized form of entry, one could read, normally: 146.9 obvious] MS 'obviously' revised from 'notoriously'. The need to quote would be removed, however, by the alternative: †† 146.9 obviously] revised from 'notoriously'. This use of the double dagger violates the general understanding that its main purpose is to save the reader consultation of the Historical Collation; but since with either entry the reader cannot understand the alteration without seeing what the entire MS reading was as recorded in the HC, the convention of the double dagger here would seem to be fairly acceptable, although—to preserve the principle of the self-explanatory lemma when the double dagger is employed—an editor might well prefer the normal quoted entry.

Certainly, there are inherent dangers in straining this convention that permits the shortcut of the double dagger to indicate that the lemma is drawn from the manuscript instead of the book-text. Only when substituted words are so close that there is no strain, should an exception be made. In the discussion of the usefulness of the single-dagger entry, an example was drawn from 158.4-5 in which the book-text read distinct from either the fact or and an MS interlineation different from either the event or. Because of the double variant distinct . . . fact and different . . . event the convention of the double dagger cannot be employed without a too lengthy lemma; but if MS had actually varied only in different for book distinct, then the doubledagger entry might well have been used since the reader would have had no difficulty in identifying in the book-text the interlined phrase of the manuscript and so could have been made aware of the alteration and its variant without referring to the HC: ††158.4-5 different . . . or] above deleted 'only another name for'; 'either' inserted. However, in the actual case, because of the presence of the added variant in MS event, this entry must not be utilized.

Alterations in manuscripts resolve themselves into only a few general categories: (1)Simple insertions; (2)Simple deletions; (3)Insertions

accompanied by deletions; and (4) Alterations made by superimposing other letters or words.

(1) Simple Insertion.

The purpose of an insertion is to add material not present in the original inscription, either during the course of composition or on review. Wherever possible, authors generally interline such additions, with or without a caret or guideline. If the editor always specifies the use of a caret,[12] the reader has a guarantee that the position of the interlineation has been fixed by the author and has not been subject to editorial interpretation; thus it may be that an editor will wish to distinguish those interlineations authorially marked by a caret from those that are unmarked.

• habit] interlined with a caret
• habit] interl. w. caret
• habit] interlined
• habit] interl.

Abbreviations of the sort illustrated are necessary to conserve space and may be used so long as they do not proliferate past the point of easy recognition of their meaning. Readers ought not to be forced to memorize more than a few conventions to be able to use an apparatus freely.

On the other hand, experience suggests that there is seldom any significance in whether or not a caret has been marked. In theory a writer might more frequently omit a caret (or add one) depending upon whether he were interlining during the course of composition or revising on review—in which rare case an editor should observe the distinction—but otherwise when the positions of interlineations are not subject to dispute, perhaps the specification may be thought not worth the space; hence if an editor chooses ultimate condensation, all mention of carets or guidelines may be dispensed with except when the editor thinks the information useful, as when a caret has been moved to another place, or has been misplaced, or a guideline has been extended to accommodate an addition, or a caret has intentionally deleted some mark of punctuation. If carets are not to be noted except in special cases, a general statement to the effect should be made.

Not all insertions are interlined. William James, for example, often added a word or more in the left margin before the word to follow instead of interlining the insertion at the end of the line above after the immediately preceding word. If there were no room for an

interlineation on the last line of a page, he might place the addition below the line in the bottom margin, or tailspace. Unless a specific value exists in noting the exact position of such additions that are not simple interlineations, there seems to be little point in specifying marginal addition (except for footnotes or sidenotes or lengthy passages) whether to the left or the right of the line, so long as the position in the text is not in question; and the same with additions placed below a line—especially since a reader has no ready way of knowing the lineation of the manuscript that produced the divergences. They are best lumped, then, under the general head of insertions, although additions is as acceptable: either word—like interlineation—signifies that the item was not written currently with the text on the line. Additions may be made in manuscript to simple insertions with or without guidelines. When a guideline is extended, the evidence can be described: very great] 'great' interl. w. caret and guideline extended to included added 'very'. However, since the fact of addition is the only important one, a condensed form without evidence can be written: very great] interl.; 'very' added. Other forms of insertions may be noted as:

• naturally] final 'ly' interl.
• nothing&c.rat;] comma inserted
• these] final 'se' added
• nothing&c.rat;] comma inserted after MS 'naught'

(2) Simple Deletion.

Simple deletion is an excision made currently before inscribing the next word, or in some cases it is deletion of an excess word made at a later time, or the alteration of one form of a word to another by deletion. Sometimes the difference in time may be established by the use of another medium, like pencil in an ink text; but if the ink is like that of the text, often no positive evidence can exist unless the nature of the alteration reveals that it must have been performed currently.

• simple] deleted
• is&c.rat;] comma deleted
• be] preceded by del. 'pr'
• may] followed by del. 'pr'
• in] foll. by del. 'ha'
• natural] final 'ly' del.[13]
• natural] initial 'un' del.
• a hurry] prec. by del. 'ha'

The terms followed by and preceded by (with their abbreviations) are perhaps the most precise one can employ. The simpler follows or precedes in their full form are equally precise but could be ambiguous if abbreviated, especially since in a description it may be necessary to write phrases like following comma deleted. Thus the best alternatives to preceded by and followed by are the words before and after, which could be further reduced to bef. and aft. These may well be adopted as the preferred forms.

• in] before deleted 'ha'
• in] before del. 'ha'
• in] bef. del. 'ha'
• a hurry] after deleted 'ha'
• a hurry] after del. 'ha'
• a hurry] aft. del. 'ha'

An editor is advised to select one set of terms capable of serving all purposes and to use them exclusively (except in cases of real ambiguity); random variation has no intrinsic value and could be confusing to a reader who might suspect a different meaning when none exists.

Deleted false starts are as important to record as complete words, especially when a change of intention is shown.

• 000.00 great] aft. del. 've'
• 000.00 end] aft. del. 'concl'
• 000.00 very] aft. del. 'gr'
• 000.00 usual] aft. del. 'cust'

Unless there is contrary specification, two adjacent deletions are to be assumed as made at the same time, since independent deletion at different times shows a change of intention:

• 000.00 decision] aft. del. 'very great' (but) 000.00 decision] aft. indep. del.
  'very great' (or) aft. indep. del. 'very' and 'great'

However, reference to another formula in the section on descriptive transcription will illustrate what may be the better method in cases of ambiguity bearing on the order of deletion. Although there is none here, one could write: If great had been excised earlier than very, the formula would be: Either by reason of a change in the medium or of a different system of deleting strokes, the fact that separate deletion has occurred at different times can sometimes be determined. If the case is ambiguous, no indication need be given, or the queried statement can be made: decision] aft. 'very great' indep. (?) del. Sometimes a false start will be deleted but its substitute also deleted. That the odds so strongly favor each to have been independently excised makes it needless to state the fact: 000.00 decision] aft. del. 've great'.

When a reading is quoted in the description to the right of the bracket, absolute clarity is gained only by enclosing the reading in single quotation marks, the standard bibliographical means for denoting exact quotation. (Double quotes are less exact in their common usage.) The device of italicizing all descriptive words also serves to bring the quoted reading into relief. If it were not for punctuation concluding or beginning a quoted reading, the roman type without quotes might serve as a sufficient distinction from this italic description; but the more complicated descriptions of alterations in which the description must use its own punctuation become thoroughly ambiguous as to whether the punctuation is part of the reading or of the description. Hence experience dictates, unfortunately, the invariable enclosure of all quoted readings within single quotation marks. Single quotation marks in the text within the quoting marks are kept as single since the quotation must be exact.

(3) Insertions Accompanied by Deletions.

Complex insertions and deletions may take a number of forms and offer a series of problems. The basic formula covers the usual case of the deletion of a reading and the interlined (or added) substitution of another.

The word above must always mean 'in a position higher than', in short, an interlineation; thus there is no need to specify interlined above when there is also deletion. An interlineation may itself be deleted in whole or in part, in which case the formulas must be constructed with particular care to avoid ambiguity while achieving economy. For example, 146.8 be in] bef. del. interl. 'the truth of' requires the reader to understand that be in only keys the position in the text of the deleted simple interlineation the truth of and is itself on the line and no part of the alteration. If the whole phrase be in the truth of had been a deleted interlineation that had not acted as an originally intended substitute for some deleted text, then a suitable key in the text needs to be found to position the deletion: To] bef. del. interl. 'be in the truth of'. On the contrary, if be in is itself part of an interlined alteration not wholly deleted, two distinguishing entries may be written to cover the different circumstances: By its terms the first entry indicates that there have been two interlineations: first the truth of and then its deletion when 'be in' was inserted before it on the interline as a substitute. That be in was not part of the original interlineation may sometimes be determined by the context, by an extension of a guideline-caret to include it, by evidence that it was squeezed in or not normally written, or by its inscription in a different medium.[16] The specification that both parts of the described alteration were interlined is necessary to emphasize the distinction between the time of their inscription and to avoid ambiguity.

The second entry solves a nasty problem by the use of a special transcriptional formula that will be described later under problems

of transcription. With no ambiguity it describes the interlineation of the complete phrase be in the truth of and then the deletion of the truth of. No unambiguous alternative to this formula is available without a lengthy descriptive entry. The problem is the same whether the undeleted part of the interlineation precedes or follows the deleted section: In the first, on occasion is a later substitute following deleted often; in the second the original interlineation was often on occasion in which often was then deleted; in the third the original interlineation was often it can be proved on occasion, deleted except for it can be proved; in the fourth the original interlineation was often on occasion, which was deleted as a whole or in two sections at different times (such a matter is not always to be determined) and it can be proved interlined above the interlineation. If evidence were available for separate deletion, the formula could read: interl. '['often' indep. del.] *it can be proved [ab. del. 'on occasion']', etc. Other typical problems arise:

Interlineations with substituted additions to replace deletions are readily handled as above. However, a problem in compression arises when the deletion, or the deletion and substitution, is within the interlineation. In the first example, not as the was interlined but then as deleted. This can be handled in several ways:

A bold use of the double dagger (to be employed with discretion) would permit such an economical entry as: Internal brackets of this nature should be set in a smaller size.

Some complexity is introduced when the interlineation is a substitute for a deleted reading:

If one supposes that not the is interlined, but as has been added by further interlineation, we have:

Further changes can be rung if the interlined as was later deleted:

It is sometimes clearer and more economical to associate consequential changes, as in

than to make two entries, as

On the other hand, if an editor prefers two entries, he may find that in double-column apparatus two lines would be needed in any event by the combined entry and that two condensed separate entries may require fewer ens of typesetting. Sometimes it is worth counting off the length of an entry before deciding on its form. Two entries, as in those above, may prove to be acceptable, if not preferable.

Not all complex changes can be broken down into a number of single entries, however. In the next example, several stages of revision appear. What happened is this. James first wrote different from from (by dittography between lines) the one originally believed in, and is an idea . . . . Perhaps the initial revision was to delete in and its comma and to insert a comma after believed. At some point he then deleted originally believed and interlined whose truth is in question as a substitute. Then he deleted different and interlined idea, but deleted it and interlined before it a new proposition. Later he added a comma after proposition and deleted from from the one whose truth is in question, and is an idea, interlining one above the deleted final idea. Last, he appears to have deleted this one and interlined and one. The final reading was a new proposition, and one. In some respects a formulaic approach to this problem would be the best one to adopt (see below) but in descriptive terms, although lengthy, an editor might write the entry:

How far to expand an entry by a true chronological description, and how far to require a reader to reconstruct the sequence from the facts themselves, is not always an easy question to answer. In the following passage James first wrote: The great shifting of universes in this discussion comes from making the word truth . . . . His first revision was to delete comes from and to interline is the, accompanied by the deletion of making and the interlineation of carrying. He then deleted is the and prefixed to it interlined occurs when we, at the same time deleting the ing of carrying. The revised text was, The great shifting of universes in this discussion occurs when we carry the word truth . . . . A chronologically contrived entry could read: A more concise note, merely listing the facts, might read: The condensation of the lemma in a case like this would seem to be sufficiently clear when the entry is read against the text; since the main purpose of the lemma is to accompany the page-line reference for identification of that part of the text being noted, there would seem to be sufficient justification to remove the repetitive quoting of text in the description. (For a truly formulaic approach to this entry, see the description of the folio at the end of this paper.)

(4) Alterations by Superimposition.

Frequently an author may start to write one word and then change his mind. Sometimes he merely deletes the false start and continues on the line with the different word, but sometimes he may choose merely to write the new word over the letters of the false start. Changes of mind like this—whether occurring during composition or in later revision—must be recorded, for often an acute critic can guess from the context what the original word would have been if it had been completed, and more information is gained about the author's style. In the description the word over means, literally, this superimposition and it must always be distinguished from above which describes an interlineation. True revisions of this nature are to be distinguished from mere mendings in which a writer touches up an ill-formed letter for

clarity—these are not worth recording.[18] Allied to such mendings as not worth recording are the repairs of inadvertent slips, such as letters transposed in haste or an anticipated letter written too soon, or occasionally what may be a genuine but inadvertent misspelling. Mechanical slips created during the haste of composition add little or nothing to a critical view of an author, and their value would seem to be nil. When a repaired genuine misspelling can be distinguished from a slip, it may be thought of sufficient interest to record. Discretion is needed, of course, even at the expense of uniformity of procedure. For example, if one's author is a notably good speller but he hastily writes ocult and then squeezes in the second c, the expense of noting the slip is wasted effort. On the other hand, if occasionally he shows a little weakness about the ei and ie words, and writes reciept, only to alter it by ei over ie, an editor may choose to record the fact. But suppose a good speller starts to write propi, stops, alters the i to o, and then continues sition, the odds favor not a misspelling but a mind outrunning the pen. The value of such information, given the expense of notation, cannot ordinarily be justified.

However, an editor needs to be on the watch for alterations of letters that probably mean a change to a different form of the word. For instance, if the manuscript reads dialectic with the penultimate i written over an a, the probability that dialectal had at one moment been in the author's mind is sufficiently strong to merit the entry: dialectic] second 'i' over 'a'. An editor seeking some rule of thumb might propose to ignore mended slips of non-words or impossible spellings, but if an actual word has been originally formed, he might decide to record its alteration in order to prevent any mistake in editorial judgment from concealing what might have been momentarily an authorial intention. Moreover, special reasons may attend more scrupulous than usual notation.

Typical entries for revision and correction are:

• explanation] first 'n' over 'i'
• clearness] 'c' over 'g'
• hole] 'ho' over 'sa'
• truth] 't' over 'T'
• True] 'T' over 't'
• Dr.] 'D' over 'M'

In the first column the examples represent either obscure slips or else the start of words that may be guessed at but not demonstrated; hence they would not be subject to the formula altered from. In the second column an editor if he chose could write, for instance: Dr.] alt. fr. 'Mr.', and so with: True] alt. fr. 'true'. It would seem that an editor need not be consistent but could choose whichever form seemed to him either immediately clearer or else more economical: the exact method by which the alteration was accomplished is of little specific interest here.

Some superimposition revisions occur in connection with other alterations:

• take] ab. del. 'hear'; 'e' over 'en'
• *take] ('e' over 'en'); ab. del. 'hear'

The first entry illustrates a good general rule that the position of the word specified by the lemma may seem more important than alterations in its form and should ordinarily come first. However, another method for writing the same information might be: take] alt. fr. 'taken' ab. del. 'hear', which is only slightly longer. It would be possible, of course, to write: take] ab. del. 'hear'; alt. fr. 'taken' but the form has less coherence than alt. fr. 'taken' ab. del. 'hear'. When only a single lemma word is being noted, even though two alterations are described there is seldom ambiguity no matter which form (as above) is chosen. Problems begin to arise when the lemma must be two or more words. For instance, the following entry is improper because its note of the alteration could apply either to 'take' or to 'take in': take in] ab. del. 'hear'; alt. fr. 'taken'. The expansion created by quoting would solve the ambiguity, of course: take in] ab. del. 'hear'; 'e' of 'take' over 'en' (or) 'take' alt. fr. 'taken', (or, in this case) *take in] ('e' over 'n'); ab. del. 'hear'. [If the alteration had been in hear one could write, take in] ab. del. 'hear' ('e' over 'a').] Indeed, quoting is necessary only to refer to a specific word in a lengthy lemma, or to a lemma that does not list the altered word, or to a lemma which has more than one example of the altered letter: take in more] ab. del. 'hear greater'; 'take' alt. fr. 'taken' (or) 'e' of 'take' over 'en' (or) we . . . sense] ab. del. 'we derive more meaning'; 'e' of 'take' over 'en'. In the first case, the bold use of the double dagger, if favored, would condense the entry: ††000.00 take ['e' ov. 'en'] in more] abov. del. 'hear'. Psychologically, the deeper the information of a letter alteration within the description, the more annoying it is for a reader to refer back to the lemma when the word is not quoted, and thus a clash may result with the general principle of the usefulness of providing the information about position first: take in difficult] ab. del. 'distinguish more complex'; 'e' ov. en'. One way of getting around this difficulty, without quoting, is to place the alteration of the letters first but with a special sign that indicates that the usual order has been broken: take in difficult] ('e' over 'en'); ab. del. 'distinguish more complex'. Although this parenthesis is more useful for lemmas of two or more words than for a single word, an editor alert to any possible misunderstanding of his meaning can employ it flexibly. For instance, he might feel that: take] ab. del. 'know'; 'e' ov. 'en' could scarcely be misread, whereas: take] ab. del. 'hear'; 'e' ov. 'en', despite the semicolon, could trouble a reader whether take or hear were in question, and hence that: take] ('e' ov. 'en'); ab. del. 'hear' closes all possible avenues for error. On the whole, it may be thought that, without quoting, the use of parentheses in this manner coming before the position provides the happiest solution. Without quoting, unless the description is so long as to puzzle a reader about the quoted word when he finally comes to it, priority may ordinarily be given to position. Alternate forms would be:

In some forms of entries the use of altered from may condense the description when the exact means of alteration is of no significance.

It is the normal interpretation of the above that the change was not made during the original inscription but as an afterthought, and so with: What] 'W' over 'We', 'hat' interl.[20] On the contrary, an entry

using altered from would in this case be ambiguous whether the change were made currently or on review: What] alt. fr. 'We'. The compression of the latter must be balanced against the value of the information of the precise description. One must also consider that not all precise descriptions do in fact distinguish the time of alteration. The following entry could represent a change made during inscription or on review: What] 'Wh' over 'We' (which could be further condensed under the circumstances to: What] over 'We'). If an editor felt the distinction were important, he could always expand: What] 'Wh' over 'We', 'at' squeezed in (or the reverse: 'What' currently over 'We').

In the examples so far, the alteration of letters by superimposition has affected words in the lemma that constitute a unit of description. Sometimes alteration of letters in one word can be combined with changes that need recording in an adjacent word or words; but in such cases the problems of position versus clarity may be intensified. For instance, the question of distance arises in such an entry as: When we discover] 'discover' ab. del. 'find'; 'W' over 'w', which can be solved by quoting: 'discover' ab. del. 'find'; 'W' of 'When' over 'w' (or) by the parenthesis: *('W' over 'w'); 'discover' ab. del. 'find'. On the other hand, since in double-column typesetting two lines would need to be devoted to this combined description, the question arises whether in the long run the reader is not better served by two brief separate entries, also taking up two lines:

• When] 'W' over 'w'
• discover] ab. del. 'find'

II. Descriptive Transcription

The transcription of a manuscript may be required: first, in an edition of a manuscript in which the editor wishes to present what has been called a 'genetic' text; that is, one that by a special construction provides within the text itself all the evidence for its growth and thus is without an apparatus; second, within an apparatus itself when another document provides the copy-text but an unreproduced underlying manuscript must be reconstructed by reference to the Historical Collation and a special list of Alterations in the Manuscript. In this second connection two main uses appear: (a) to transcribe passages of deleted or of variant matter of some length as part of an Alterations entry or else in a separate section of the apparatus devoted to transcripts of all deleted draft leaves and extraneous matter; (b) to solve the problem of a more condensed

form of description in a complex regular entry than can be managed by the usual verbal description.

Heretofore, 'genetic' texts have been produced by transcribing the original form of the manuscript and then indicating the series of its revisions directly after the revised word(s) by means of a number of arbitrary symbols like pointed brackets to the left or right, arrows, bars, and so on. This method was brought to a high state of refinement by the publication in 1962 of Melville's Billy Budd, edited by Harrison Hayford and Merton M. Sealts, Jr., which has since served as a general model for other manuscript texts. Despite the maximum compression gained by these symbols, the form of transcription is subject to certain difficulties: the symbols take considerable acquaintance before they can be read with any sort of ease and understanding, and this difficulty is compounded by the lack of any agreed standard for the use of these symbols; as a result, each subsequently edited manuscript has introduced different purposes for the same marks and freshly minted symbols preferred by its editor in an attempt at improvement on his predecessors, much to the confusion of the reader. Moreover, a genetic text of this sort cannot be read with ease (the difficulty of symbols apart), principally because the basic transcript is by necessity of the original unrevised inscription; hence the successive revisions that form the final text are encountered only as one plunges deep into the thicket of the arbitrary markings. It is difficult to read the original text consecutively; it is impossible to read the revised text at all in a coherent sequence. As a consequence, the editors of Billy Budd found it necessary to accompany the genetic reference text with a reading text that presented in usable form the finally revised version. A genetic text, in short, cannot be a reading edition except of a very lightly revised manuscript. It follows that an editor must make the basic decision, often, whether as in a genetic text his edition will be confined to the status of a reference work on which other editions will be built, or whether he will edit a reading text of the finally revised manuscript and in an apparatus detail by description and quoting, as necessary, the development of the final text from its original form. Since in either case a reader must work hard to reconstruct the other form, the matter may boil down to the general usefulness of having to hand the final revised text and to labor over the original, or vice versa.

For the purposes of the present paper, however, the problem centers chiefly on the method of transcription to employ when a manuscript must be reconstructed from the apparatus in cases where extensive transcription of rejected text or of text in the process of

revision is required. (Also in question is the method for transcribing, apart from the Alterations apparatus, deleted or discarded trials, drafts, and the like, that may accompany the final manuscript text.) The methodology of transcription does not differ in the system here proposed, no matter to what purpose it may be put, however: it is the same whether it is utilized for condensing an entry in an ordinary descriptive list of Alterations or for transcribing a revised document for reading and reference. In the latter case it may often serve as an acceptable compromise between a reading and a reference text.

The central feature of the system is that it transcribes the finally revised form of the text and within brackets placed after the appropriate words describes the changes made in this text from the original inscription. (The method of this description agrees with that illustrated above for use in an apparatus list of Alterations.) In a genetic text with arbitrary symbols one is forced to transcribe the original version since the symbols must be placed immediately after the word affected by revision and such placement is possible only with the original word, to which the revisions may then be appended. In descriptive transcription the convention of the asterisk frees the transcript of this necessity and effectively delimits even a considerable area of final text to be described. For instance, as illustrated, the first page of the manuscript of Chapter XIV of William James's Meaning of Truth (1909) is transcribable as follows:

This is complicated, but the revised text itself can be read simply by skipping all brackets; indeed, at some extra cost the legibility of the revised text could be increased by reducing the size of the type in all bracketed items, although this expedient might be imprudent in transcriptions within the apparatus list where ordinarily the type has been made as small as can be read without strain.

Several points need comment. First, all text intervening between the asterisk and the applicable bracketed description is covered by the description. Bracketed entries containing quoted words, as in notation of deletions or interlinings, may intervene without ambiguity between the start of the asterisked text and the applicable description, as in: *When, ['we' del.] for instance, we [ab. del. 'If we']. If an editor feels uneasy, he might, as in this example, reduce the size of these intervening brackets as is done in bracketed entries within bracketed entries. Morever, an asterisked passage may itself have an internal asterisked entry (which should be double asterisked), in which case the first small-bracketed description to appear applies to the internal double asterisk (and so on) and the final full-bracketed description to the whole of the material present between it and the original single asterisk. No example occurs in the transcript, but if we hear you were interlined above a deletion and that then interlined after hear, the text would read: *we hear **that [interl.] you [ab. del. 'it has come

to our attention']. If, instead, that had been a part of the original interlineation but had been deleted, then no asterisk would be needed: *we hear ['that' del.] you [ab. del. 'it has come to our attention']. In the transcript an example does appear, however, of double asterisked words within a bracketed description to an asterisked passage: *a new proposition, ['idea' del.] [ab. del. 'different from | from the one **whose truth is in question' [ab. del. 'originally believed, ['in,' del.]']].

It will be remarked in this example that idea (deleted) was the original interlineation and that a new proposition is the revised interlineation. There should also be no difficulty in perceiving that in the first inscription the deleted reading was originally believed in, and that a comma was added after believed when in, was first deleted before the decision was made to revise the whole phrase. No asterisk is needed if the description applies only to a single word following a bracketed description, since ambiguity could not then be present: was [interl.] he [ab. del. 'she'] aware. Independently deleted words may be given separate brackets: this ['great' del.] ['good' del.] noble man. But, as in the transcript's independently deleted idea proposition, the two can be included in the same bracketed notice: this ['great' del. 'good' del.] noble man (or in a descriptive entry): noble] aft. del. 'great ['good' del.]' (or, in another situation): this *noble [ab. del. 'great ['good' del.]'] man.

On the other hand, since there could scarcely be ambiguity about the deletion of a false start before a word that was later itself deleted, no very useful purpose is served by marking them as separately deleted, and the simple form would appear to be acceptable: All *of life's [ab. del. 'ro mental'] roads. Of course, if one wished to be scrupulous, one could write: All *of life's [ab. del. 'ro' del. 'mental'] roads (or preferably) [ab. del. '['ro' del.] mental'] roads.

As remarked, simple deletion is best handled by: is ['something' del.]. But when description must be added which applies to the deletion, the simplest order is: is ['something' del.] ['a' del. intrl.]. Another sequence might be used to clue the reader in earlier on the position when the deletion is a long one, and thus to prepare him for the final notation of deletion: [del. intrl. 'the four day working week has been tried in certain industries']. Preliminary warning is even more necessary in such cases as: [del. intrl. 'the four day working week has been tried in certain industries' ab. del. 'various experiments to promote leisure have been tried'], which is clearer than [intrl. 'the four day working week has been tried in certain industries' del. ab. del. 'various experiments to promote leisure have been tried'] and more satisfactory for indicating the position at the earliest moment than the conventional [del. 'the four day working week has been tried in certain industries' ab. del. 'various experiments to promote leisure have been tried'].

In an ordinary apparatus entry if the description quotes text that passes from one page to another, usually no purpose is served by noting the end of one page and the beginning of the next so long as the pages are in order: the reader has no opportunity to identify the page endings and beginnings in the edited text, anyway; thus in normal diplomatic transcripts (and always in edited text) no more significance inheres to passing from one page to another in the original document than from one line to the next. The case is altered, however, in the transcription in appendixes of discarded trials and the like, not as entries in the apparatus, or in the edition-text itself in a diplomatic transcript. Moreover, if the transcription in an apparatus entry continues not on the next page but on another that is not in order, although the text is continuous, notation of the difference in the page is necessary. When on folio 17 of the manuscript for Chapter XIV of The Meaning of Truth James deletes a sentence that begins on the last line and continues over onto folio 18, the distinction between the

deleted text on the two pages is superfluous. However, deleted text at the foot of folio 4 is continuous with deleted text beginning folio 10 as a consequence of a major revision and expansion of the deleted material that takes up folios 5-9. It would be most unhelpful to a reader to stop the transcription in the apparatus of this deletion with the end of folio 4, continue with normal entries for a considerable amount of text on fols. 5-9, and then when folio 10 was reached begin the broken-off deleted text as a fresh entry. Obviously, the whole passage must be transcribed as a unit since it is continuous text, no matter how separated, but an indication must be made at the proper point that the deleted text jumps from folio 4 to 10.

When the beginning of pages is being indicated within transcribed text, it is usually best to enclose the folio number for the new folio (or page) in square brackets after a vertical stroke at the point where the transition occurs: as Mr. Russell justly says, 'to | [fol. 2] settle the plain question . . . . In the example mentioned of a skip, the transcript

might read: He is a logician and a mathematician, accus-[end fol. 4 | begin fol. 10]tomed to think in complicated chains of identities. (Broken words on either side of a stroke are set close up; separate words are spaced.) In transcribing a manuscript when a sentence begins a new folio, the bracketed folio number may be placed on a separate line and the vertical stroke at the end of the text of the preceding sentence and page. An attempt to set the brackets before the sentence could be ambiguous, whether or not a paragraph were also present. Space may be saved by including the folio number within the text itself and running it on: to settle the plain question. | [fol. 92] This is a matter etc. If the text on the new folio begins with a paragraph, one may write in this form of continuous transcription: to settle the plain question. | [fol. 92] [¶] This is a matter etc. However, when a page begins with a new paragraph, or with something not continuous, like another item in a list written in separate lines, it is consistent with the rest of the transcript if a new page is indicated visually by the setting of the bracketed folio number in a separate line, the text then continuing below it: Of course, in transcription within an apparatus entry, the run-on notatation must be employed.

Descriptive transcription of underlined (italic) passages should put the bracketed description in roman instead of the customary italic for differentiation: *of believing which [intrl.] are good he [intrl. 'for instance' moved fr. aft. 'us' and del.] assumes us to mean . . . . Similarly, bracketed folio numbers intervening between italic text may be printed in roman.

In the apparatus to a complicated revision, descriptive transcription is sometimes necessary to reproduce a revised passage that, say, has been deleted or is a substitute for a deletion. The simplest case is exemplified by lengthy deletions to shorten text, or deletions at the heads or feet of pages accompanying major revision. As a consequence of the rewriting of folio 5 by a substitute leaf in James's Chapter XIV,

folio 6 begins with deleted text that has no antecedent. The record of this passage in the apparatus reads: Since this passage must appear in the apparatus, no separate listing of the variants can be managed, and descriptive transcription is a necessity.

Descriptive transcription can also clarify as well as condense descriptive entries of revisions. In the transcript above of folio 1 of James's Chapter XIV, the book-text reads a new proposition, quite but the final MS text is quite a new proposition,. A combination of description and quoting could be contrived to deal with the complex revision:

but this compromise, useful as it sometimes is, in this case yields in clarity and compression to straight descriptive transcription: In the first entry it would have been possible to refer the reader by a single dagger to the Historical Collation for the variant MS text, but the quoting of the MS text is here a preferred device to make an extended and complicated description immediately understandable within itself. Of course, in the second, full-fledged descriptive transcription no reference to the Historical Collation need be contemplated since the transcription automatically provides the MS text.

The book-text continues shortly and is, moreover, a fact usually very hard to verify, it being "far easier," as Mr. Russell justly says, "to settle the plain question of fact:. A series of separate entries could be contrived to deal with the MS variants in this stretch:

If an editor chose, these separate facts could be combined into one descriptive entry pretty much as they are in separate entries; however, descriptive transcription would be much clearer and shorter: This example may serve as a good illustration of the way in which it is possible to do away with a series of short entries in favor of one transcript.

At 151.4 James wrote we must but then deleted must before continuing need to stay. Later, he deleted we and interlined one above it, and added a final s to original need. The need for flexibility in the choice of constructing an entry is shown by the following four options:

The first (straight descriptive transcription) takes about 65 ens to set; the second, a mixture of description and descriptive transcription, requires about 64 ens; the third and fourth, similar mixtures, take about 56 and 52 ens respectively. The compression of the last is gained by the omission of needs from the lemma as not strictly required. Economy suggests a choice either of the third or fourth.

In the next sentence James first wrote comes from making (the m of making written over some letter that just possibly may be an a). First he deleted making and interlined carrying; he then deleted comes from and interlined is the but deleted this is the and added before it occurs when we, necessarily then deleting the final ing of carrying. Two separate entries are possible:

However, in this case the two entries can profitably be combined into a single descriptive transcript: The rest of the alterations for this page (see the illustration) can be given as examples of fairly typical entries in a list of Alterations:

III. Historical Collation

The present paper has concerned itself with the transcription of a manuscript that forms the copy-text for an edition and also with the list of Alterations appended to such an edited text. It has also surveyed the problems that arise in this list of Alterations when the copy-text is, instead, a printed document and the revisions created during the course of the inscription of the manuscript may lead to considerable differences between its final text and that of the copy-text. In this latter situation a special form of notation involving the use of the double dagger has been suggested for use in the Alterations list that will enable a reader to reconstruct the variant manuscript text and its revisions with reference only to the copy-text as printed. On the other hand, the suggested convention of the single dagger requires reference to an Historical Collation for a full understanding of variants involving a rejected manuscript reading.

In the special list of emendations to a printed-document copy-text no need ordinarily arises to record the inscriptional variants within an adopted manuscript reading since these are not usually pertinent to the essential matter at hand—the simple record of the emendation of the copy-text from another source. It is understood that all rejected readings of the copy-text that result from emendation will also form a part of the Historical Collation when they are substantive, and that variants created during the inscription of the final manuscript reading used as an emendation will be listed in the record of Alterations in the Manuscript.

This exclusion of inscriptional alteration within the records of the list of Emendations is admittedly arbitrary and is suggested only because the number of substantive items in such a list is usually so small as to place little burden on the user: the facts about the final manuscript readings adopted as emendations are provided in the Emendations list and if a scholar is concerned to see whether any of these readings developed in the manuscript from others, he can readily consult the list of Alterations for full information about accidentals as well as substantives. On the other hand, if an editor wished he could record changes according to the system suggested below for adding manuscript alterations to the Historical Collation in certain circumstances: Once engaged, an editor must be consistent, however, whether the information is pertinent as in the first illustration or of no immediate concern as in the second.

The purpose of an Historical Collation is to list all rejected readings in the authoritative documents collated, whether substantives only, or in some editions, according to the circumstances, accidentals as well as substantives. By convention the lemma repeats the reading of the edited text and to the right of the bracket appear the rejected readings with their sigla. Also by convention any collated document that agrees with the lemma is not listed to the right of the bracket but instead only the variants. Thus when the sigil for a collated document is not present, the reader must take it that its reading—in contrast to the others recorded—is that of the edited text. The illustrations below assume that the collated documents comprise a manuscript and three printed editions, I, II, and III.

• right] wrong MS right] wrong MS, I
• right] wrong II-III right] wrong MS, I-III

In the first, all three printed editions read right but MS reads wrong; in the second, MS and I read right and the second and third editions wrong; in the third, the editor chose from the second edition the emendation right as a substitute in his critical text for the reading wrong in MS and the first edition; in the fourth, right is an editorial emendation, unsupported by documentary evidence, for the reading wrong in all collated texts.

When a manuscript is present but a printed document has been selected as copy-text, a certain number of the rejected manuscript variants will represent the original and unaltered inscription; but others will be drawn from the final version of the manuscript text altered during the course of inscription or review. In such a situation an editor has two choices:

Not all situations are susceptible of improvement or clarity, however, by such combination. In the following example the book text read the gradual wearing away by internal friction of portions that originally interfered. The MS agreed with the book through friction of. at which point James first wrote interfering tendencies., inserted originally before it, but then deleted the augmented phrase and continued his inscription with forces that originally interfered. so that in the MS final text the only variant is forces for portions. Some editors might well prefer to break up the record conventionally into HC ‡000.00 portions] forces MS (using the arbitrary sign ‡ to let the reader know that further information would be found in the Alterations) †000.00 portions . . . interfered.] aft. del. 'originally [insrtd.] interfering tendencies.' The combined HC entry would be lengthy and the real point of difference between lemma and MS reading would be somewhat obscured by 000.00 portions . . . interfered.] ['originally [insrtd.] interfering tendencies.' del.] forces that originally interfered. MS. In another example, when the only variant appropriate for notice in the HC is also part of a lengthy MS alteration that otherwise agrees with the book text, the attempt to combine entries in the HC may

come close to destroying the HC's usefulness. For example, the book text reads 'monistic species in the pantheism'. In MS 'pantheism' is in roman as part of an interlined substitution that otherwise agrees with the copy-text. In such a case it would be simple to separate the entries as HC ‡000.00 pantheism] pantheism MS, and Alt. ††000.00 monistic . . . pantheism] ab. del. 'other is commonly' whereas the combined HC entry may be thought only marginally acceptable because of the obscuration of the single point for HC recording, the roman in the MS for the book pantheism in italic: 000.00 monistic . . . pantheism] *monistic . . . pantheism [ab. del. 'other is commonly']. Still, the entry is possible. On the other hand, the following example, quite obviously, must be handled separately since a combined HC entry would be a monstrosity: HC 000.00 clean] pure MS and Alt. ††000.00 something pure and] ab. pencil del. 'an inner structure as *pure and classical and temple-like [ink del.]'

The occurrence of such complex cases very difficult indeed to combine into any sort of clear HC entry suggests that not every occurrence of textual variation accompanied by alteration can be included, with the alteration, in the Historical Collation in a suitable manner and thus that full consistency is almost impossible to achieve. This is not necessarily a disqualification, however, for the use of the arbitrary signal ‡ readily steers the reader to the Alterations list when for particular reasons the account of alteration has been omitted from the HC.

One may sum up by remarking that under either system the Historical Collation contains the full record of differences between the book copy-text and the final text of the manuscript. Some editors may feel that the retention of this integrity of record in the Historical Collation, with the economy (and occasional illumination) of the association of MS alterations when practicable with this notation outweighs the disadvantage of the fact that the list of Alterations is thereby denuded, in part, of information proper to its full record. Experience suggests, indeed, that sometimes a moiety, if not a majority, of the cases where alteration accompanies textual variation can be transferred to the HC by a combined entry. One needs to consider, thus, that the difficulty may be more apparent than real, for in any event a reader must consult both lists to be able to reconstruct every detail of MS variation from the copy-text, including all alterations. And even if a scholar is for the moment intent only on reconstructing the alterations made in the course of the manuscript's inscription, reference to the HC for single dagger entries in the Alterations and to the text for

double dagger entries will be required, for some alterations can scarcely be understood without the necessary information of their textual variation in final form from the copy-text. Under these circumstances, perhaps, the splitting of the record of alteration between the two parts of the apparatus—depending upon whether it is or is not associated with textual variation—may not seem to be a serious defect in view of the gains that may accrue, particularly since the system is not rigid and alterations too far dissociated from the exact point of variance which it is the business of the HC to record can be left in the Alterations. In this connection the signal ‡ is very useful for directing the scholar's attention to such cases. Nevertheless, if an editor dislikes this suggested system, or if his problem of recording variants is so simple as to make it of small value, he can readily construct the two independent and complete sections of apparatus with no more overlap than is required in the Alterations formulas of notation with daggers for cross reference.

Notes