1911 Encyclopædia Britannica/Textual Criticism

by John Percival Postgate

TEXTUAL CRITICISM, a general term given to the skilled and methodical application of human judgment to the settlement of texts. By a “text” is to be understood a document written in a language known, more or less, to the inquirer, and assumed to have a meaning which has been or can be ascertained. The aim of the “textual critic” may then be defined as the restoration of the text, as far as possible, to its original form, if by “original form” we understand the form intended by its author.

Texts may be either autographs or they may be transmitted texts; the latter, again, being immediate copies of autographs or copies of copies in any degree.

Autographs (which may be taken to include whatever, though not actually in the writing of its author, has been revised and attested by him) are not exempt from the operations of textual criticism. Editors of journals remove the slips of the pens of their contributors; editors of books, nowadays usually in footnotes, the similar lapses of their authors. With this branch of textual criticism, however, modern scholarship is not largely concerned. Not so with immediate copies. Textual criticism is called upon to repair the mischief done to inscriptions (texts inscribed upon stones) by weathering, maltreatment or the errors of the stone-cutter. Any great collection, such as the Corpus of Latin inscriptions or the similar Corpus of Greek, will show at once its activity and ability in this direction.

The chief field of textual criticism is elsewhere. The texts of the older authors which have come down to us were written for the most part not on stone but on papyrus, parchment or other perishable material. Of these several copies had to be made, both by way of prevention against the wear and tear of use and as a means of satisfying the desire of other persons than the original possessor to be acquainted with their contents. Had the copies made of ancient writings been mechanical reproductions of the originals, such as the photographic facsimiles of modern times, there would have been little here for textual criticism to do. The ancient texts have not come to us in this way, but through copies made by the human hand directed more or less by the human intelligence. Now a copy made thus can in no circumstances be a quite exact rendering of that from which it is copied or its exemplar. A copy, qua copy, can never be the equal of the exemplar, and it may be much its inferior. This deterioration increases with the number of successive copyings. Let us suppose that from a text which we will call A a copy has been made which we will call B, and from this again a copy which we will call C. If the copyist of B goes wrong once and the copyist of C twice in a hundred times, then, assuming that there is no coincidence or cancelling of errors, the relative correctness of the three texts A, B, C will be 100 (absolute correctness), 99 and 97.02. If C had made his copy direct from A, his percentage would have been 98. The importance of this must be borne in mind when we are dealing with transmitted texts, which have passed through many stages of copying.

In the Epidicus of Plautus, 1. 1. 10, the right reading habitior, “more portly,” has been preserved to us by Donatus, an ancient commentator on Terence (Eunuchus, 2. 2. 11). It was corrupted to abilior by omission of the h and confusion of t and l, and this corruption, which is attested by the oldest extant copy, the Ambrosian palimpsest, was still further corrupted in the other copies to agilior.

The first step towards the restoration of a text is the examination of the evidence upon which it is or is to be based. This begins with the investigation of its traditional or transmitted form. For this we have usually to rely upon manuscripts (MSS.). By manuscripts (q.v.) we understand copies of the text made before the art of printing came into general use. These may be either extant or non-extant. The evidence of extant manuscripts must be ascertained by collation. To collate a manuscript is to observe and record everything in it which may be of use towards determining what stood in the source or the sources from which it is derived. A manuscript is not usually a clean or single piece of writing; it is commonly found to contain alterations by erasure, addition or substitution. Such alterations may be due to the writer or writers of the MS., called the scribe or scribes, or to some other person or persons (for there may be several) called correctors. The relative importance of these corrections, it is obvious, may be very different. It is therefore necessary to distinguish the different hands which have been at work on the manuscript. Account must also be taken of the number of lines in each page, the number of pages in each quire, of gaps or lacunae in the manuscript, and so forth. The work cannot be considered complete till all the extant manuscripts have been collated or at least examined.

When this is done we shall have the materials for pronouncing a judgment upon the text as directly transmitted. Perhaps there is only one extant MS. of the text, as in the case of the Mimes of Herodas and the Annals and Histories of Tacitus. Then this part of our work is done.

But often we have to take account of a number, and it may be a large number, of manuscripts, whose respective claims to attention we must determine. In the first place we shall discard all manuscripts which are derived by copying from other extant manuscripts. If a MS. is immediately or ultimately derived by copying from another MS., it cannot, qua copy, tell us anything that we do not know already if the latter MS. is known to us. But how can we tell that a MS. is so derived? It must be later than the other MS., and the similarity between them must be such as to permit of no other explanation. In the absence of explicit dates the relative age of MSS. is often hard to determine, and hence the criterion of unmistakable resemblance is one of special importance. If the MSS. agree in singular though trivial mistakes, if they omit, apparently without motive, words and passages which other MSS. preserve, we shall be safe in pronouncing that there exists a close bond of connexion between them, and if one of them shows errors which, though strange in themselves, are quite intelligible when we see what stands in the other, then we shall be justified in concluding that the second is that from which the first is derived. For the proper consideration of such points a personal examination, autopsy, of the MSS. or of facsimiles of them, is very often indispensable. It was thought at one time that a MS. of the Latin poet Propertius at Naples (Neap. 268) might have independent value as an authority for the text. But its claims were disposed of when (amongst other facts) it was observed that at book iv. 8, 3, the MS. with which it most closely agreed (F, No. 36, 49 in the Laurentian library) had a gap at the beginning of the line and only the end words “uetus est tutela draconis,” with the marginal note “non potuit legi in exemplari hoc quod deficit,” and that Neap. 268 gives the line as follows, “non potuit legi uetus est tutela draconis.”

Accident apart, identity of reading implies identity of sources. The source of a transmitted reading may undoubtedly be the author's autograph: but if not, then it is some MS. in the line of transmission.

The peculiar resemblances of two MSS., though not sufficient to warrant the derivation of either from the other, may be sufficient to establish some connexion between them. From the axiom which has just been cited it follows that this connexion can be due only to community of source, and we thus arrive at the idea of families of MSS. Suppose that a text is preserved in seven MSS., A, B, C, D, E, F, G. If we find that of these A stands apart, showing no great similarity to any of the other six, while B, C, D on the one side, and E, F, G on the other, much resemble each other though differing considerably from the rest, we may express this by saying that B, C, D form a “family” descended from a hypothetical common “ancestor” which we may call X, and E, F, G another “family” descended from a hypothetical “ancestor” which we may call Y. The readings of X which can be deduced from considering the agreements in B, C, D will be of higher antiquity and of greater external authority than any of the readings in B, C, D taken singly. And similarly for the readings of Y and those of E, F, G. Nor shall we stop here: but we shall further compare the readings of X and Y with each other and with those of A, and thus deduce the readings of a still more remote ancestor which we may call Z. Z will be the archetype of all our existing MSS., and we may embody our results in a pedigree of manuscripts or stemma codicum as follows:—

If we have done our work properly, the texts that we arrive at for X and for Y will be freer from error than the texts of the separate members of the families B, C and D, and E, F, G respectively, and that of Z freer from error than that authenticated by any existing MS.

The procedure, however, is by no means always so simple. That a text may be improved by the comparison of different MSS. is not a modern discovery. It has long been known, and the knowledge has led to the production of what are known as conflated manuscripts or Misch-codices. These are MSS. produced by “crossing” or “intermixture.” In the following stemma M and N are “mixed” or “conflated” MSS., being formed by the blending of readings from the “pure” or “unmixed” codices A, B and D, E respectively.

Intermixture may take place to any extent, and the more of it there has been the more difficult does it become to trace the transmission of a text.

Whether crossing improves a given text or not depends ultimately on the knowledge and the judgment of the crosser, and these will vary indefinitely. On the whole it is probable that it does, provided it is not accompanied by other attempts at improvement. If it be, as may very well be the case, the text will probably suffer. For but a small proportion of scholars' corrections are really amendments, and a far smaller proportion of scribes'.

The “genealogical” method, as we may call it, cannot in strictness be applied to conflated MSS., as their mutual relations can rarely be with certainty disentangled. But it is often possible to detect in such MSS. a common strain, shown by their agreement in peculiar corruptions or in probable readings when these latter would have been hard to discover by conjecture. This is practically an application of the method to a portion of such manuscripts.

A special value attaches to a conflated codex when one of the MSS. from which it has been compounded has perished and its readings are thus otherwise irrecoverable. This is exemplified in the Neapolitanus of Propertius, a manuscript now at Wolfenbüttel.

It not unfrequently happens that good or instructive readings are found in manuscripts which are in general of small trustworthiness (see below), and whose relations to the general tradition it is not worth while to investigate. These readings may be cited by the name of the MS., or if still greater brevity is required as the readings of inferior MSS. (deteriores), or, as is frequently done, by the symbol S.

Non-extant Manuscripts.—Some of the most valuable of ancient MSS. have disappeared since their discovery in modern times. When this has happened we have to rely upon mere copies, many times of inferior quality, or upon the information which old scholars have given us respecting them. In the latter case what we have are not “collations,” for the art of collation was not understood till the 19th century, but selections or “excerpts” of readings which we have reason to fear are often imperfect and erroneous. Further, it must not be assumed that all readings which are cited as being “ex uetustis codicibus” are necessarily from older or better MSS. than we now possess or indeed from MSS. at all. Scholars since the Renaissance have not always been above inventing codices to obtain currency for their own conjectures. The codices of Bosius (1535-1580) are just as imaginary as the “old plays” which appear as the source of so many of the quotations that head the chapters of the Waverley novels, and suspicion rests on Barth, Lambinus and others.

Some texts and portions of texts of ancient writers are now only known from printed books. The metrical treatise of Terentianus is now preserved in the editio princeps (1497) alone. All known MSS. of Silius Italicus have a considerable gap in the 8th book, first filled up on the authority of Jac. Constantius (1503), and not printed with the rest of the poem till the edition of Aldus (1523). The early printed books are often called by old scholars codices impressi (typis), “printed manuscripts,” a phrase which at first seems curious to us but becomes perfectly intelligible when we examine these codices impressi and observe how closely they follow the codices scripti.

By the methodical employment of these means we shall arrive at a text different from any existing one. It will not be the best one, possible or existing, nor necessarily even a good one. But it will be the most ancient one according to the direct line of transmission, and the purest in the sense of being the freest from traceable errors of copying and unauthorized improvements.

The textual critic has occasionally to deal with the effects of oral transmission. A text so transmitted must in the lapse of time be profoundly though insensibly modified, its forms and expressions modernized, and, if widely disseminated, local variations introduced into it. This is the case with the Homeric poems, the ascertainment of the original form of which is a task beyond the powers of criticism. Even where, as in the Vedas, the sacred books of India, there is proof that the work has been transmitted without change through many centuries, the existence of unintelligible passages and unmetrical verses shows that here too there is work for textual criticism to perform, though in the opinion of most scholars it should be confined to the restoration of such forms as would be unconsciously and inevitably corrupted through changes of pronunciation and the like.

The invention of printing has naturally limited the province of textual criticism, and modified its operations. The writer's autograph, if it is preserved after it has been through the hands of the printer, has seldom more than an antiquarian value. As a source for the text it is superseded by the printed edition, and if there is more than one, then by the latest printed edition, which has been revised in proof by the author, or, in certain cases, by his representative; and the task of the textual critic is restricted to the detection of “misprints,” in other words, of errors which the compositor (the modern analogue to the scribe) has made in “setting up” the manuscript, and which have escaped the notice of the proof-reader and the author or his representative. If, however, this revision has been neglected or incompetently performed, the number of such mistakes may be considerable.

Another question with which the textual critic of modern authors must be prepared to deal is the relative importance of different editions, each of which may have a prima facie claim to be considered authentic. Thus Shakespearean criticism must decide between the evidence of the first folio and the quartos: the critic of Shelley's poems must consider what weight is to be attached to the readings in the posthumous edition by Mrs Shelley, and in unpublished transcripts of various poems. Where there is great or complicated divergence between the editions, as in the case of Marlowe's Faustus, the production of a resultant text which may be relied upon to represent the ultimate intention of the author is well-nigh impossible.

For the bettering of the transmitted text we can call in aids of a partial or subsidiary character which are known in general as testimonia. Such are Anthologies or collections of extracts. The oldest authority for an epithalamium of Catullus (62) is an anthology at Paris written in the 9th century.

Translations from one language into another may help to fix the reading of the original, or this again that of the translation.

In Shelley's Prometheus Unbound, ii. 5, 54,—“Child of Light! thy limbs are burning | Through the vest which seems to hide them”—“limbs” is supported against “lips” (ed. 1) by “membre” in the Italian prose version made by Shelley himself; and similarly in l. 52 “looks” (not “locks”) by the rendering “sguardi.” In his translations of Euripides' Cyclops, 381, “a bowl | Three cubits wide and four in depth, as much | As would contain four amphorae” the Greek original clearly points to “ten amphorae” and four may have come from the previous line.

In direct quotations, either of passages or single words, and either with or without the author's name, we must be sure that the writer is quoting exactly.

A couplet of Propertius is written upon the walls of Pompeii in the following form: “Quisquis amator erit, Scythiae licet ambulet oris, | Nemo adeo ut feriat, barbarus esse uolet.” Here the manuscripts have “Scythicis”—“deo ut noceat,” of which deo is rejected by every one in favour of the Pompeian reading, but Scythicis and noceat are retained on the ground that they are in themselves better than the Pompeian readings, which may be simply due to lapse of memory. In Shelley's Julian and Maddalo, 40,—“(talk) such as once, so poets tell, | The devils held within the dales of Hell | Concerning God, freewill and destiny,”—vales has been suggested to make it harmonize with the passage of Milton to which reference is made: but the argument is not conclusive.

Parodies may prove of service in restoring the form of what is parodied or this in restoring the parody. So also obvious imitations, especially in a highly imitative literature such as Latin poetry. The connexion of the passages must in all these cases be unmistakable.

In Homer, Iliad, i. 4 seq., Aristarchus had the common reading αὐτοὺς δὲ ἑλώρια   τεῦχε κύνεσσι | οἰωνοῖσί τε πᾶσι, but another Homeric critic of note, Zenodotus, read δαῖτα for πᾶσι, and this is supported by the obvious imitation in Aeschylus, Supplices, 800, who has ὄρνισι δεῖπνον.

The support which a reading gains from the evidence of the directly transmitted text and from the auxiliary testimonia may be called its documental probability. To restore a text from the documental evidence available we must know and weigh the causes which tend to vitiate this evidence in its various kinds. We shall speak first of those which affect the direct transmission of texts. These are either external or internal.

External.—A text may become illegible through damp or constant thumbing; portions of it may be torn away; if it is in book form, leaves or whole quires may be detached and either lost or misplaced. When this has taken place on a considerable scale, the critic is helpless; but minor injuries may sometimes be traced and remedied. The weakest parts of a MS. book were the outer margins; and hence the beginnings and the ends of lines, whether of verse or prose, were specially liable to injury. It obviously makes a difference upon which side of a leaf, whether on the verso or the recto, a line was written. Hence the determination of the paging of the archetype (as was done for the archetype of Lucretius by Lachmann) has more than a merely antiquarian value. In ancient classical MSS. the first letters of poems in verse and of paragraphs in prose usually, and the initial letters of lines in verse occasionally, were written separate and by another person than the scribe (who was called the rubricator), and hence were apt to be omitted. Other external circumstances may prejudicially affect a text. The copy from which Shelley's Julian and Maddalo was printed was written on very narrow paper, and the punctuation marks at the ends of the lines were frequently omitted.

Internal.—These errors arise from the default of the scribe or copyist, and, in the case of printed books, the compositor.^[1] They are very numerous. They may be roughly arranged according to the degree in which the volition of the copyist is absent or present, as involuntary or mechanical, semi-voluntary and voluntary; or again as they affect single signs (letters, figures or symbols), words, lines or even larger units such as sentences or paragraphs.

Simple Errors of the Eye.—(a) Confusions of letters. These are very numerous, and different in different scripts or styles of writing (see Palaeography). Thus the Roman letters E and F are liable to be confused in capital script, but not in cursive (e, f), C, G, in capitals, c, e in the cursive writing called Caroline minuscule, c, t, in the angular cursive of the 13th century and later. Texts which have had a long history will often show by the letter-confusions which they exhibit that they have passed through several distinct stages of copying. It is to be observed that two different styles of writing are often found in the same manuscript, the difference being utilized for the purposes of distinction. Thus in Greek cursive MSS. notes were often written in uncials; the use of majuscules or capitals for headings and for the initial letters of lines is well known. (b) Omissions of letters. (c) Shiftings of letters, sometimes by syllables. This is very common in half intelligent or half mechanical copying. In printing we get the disarrangement of type which is known as “pie.” (d) Confusions of symbols and abbreviations.

(a) Examples of confusion of capital letters from Shelley's poems are: Prometheus, i. 553, “Mark that outcry of despair " for “Hark”; Hellas, 472, “Hold each to the other in loud mockery” for “Told.” Of cursive letters: Marenghi, 130, “the dim ocean” for “the dun ocean”; Letter to Maria Gisborne 126, sqq., “,above | One chasm of Heaven smiles like the age of Love | On the unquiet world” for “eye.” (b) Translations from Goethe's Faust; sc. i. 46, “To live more beastily than any beast,” for “beastlily”; ii. 165, “eye” for “eyne” (in spite of the rhyme with 163). (c) Prometh., iv. 575, “Neither to change, nor flatter, nor repent,” for “falter.” In Latin MSS. we often find a mere jumble of letters. (d) Confusion of words through abbreviations is very common in ancient MSS., where they were much employed. At a famous place in the doxology of 1 Timothy iii. 16, the MSS. vary between ὅς (or ὅ) and Θέος. In uncial writing OC () might easily be miswritten or altered to ΘC (Θέος) or vice versa.

Loss of Letters, Syllables, Words or Lines, through Similarity of Writing: Homoeographon.—When similar letters or groups of letters stand next to each other, one of these is liable to be omitted. This is the simplest case and is called haplography.

Similarity operates differently if the similar groups stand in different lines of the exemplar. Then the copyist's eye is apt to slip from the first of two similarly written groups to the second; and he will thus omit all that is between. The term homoeoteleuton (“similarity of ending”) is often used of these omissions, but it is not adequate, as similarity anywhere may produce the same result.

Examples of homoeographon and haplography. Shelley's Cenci, v. 4, 136, “whose love was [as] a bond to all our loves”: a similar omission in Witch of Atlas, 599. In Stanzas written in Dejection near Naples the two lines 4, 5, “The purple noon's transparent might, | The breath of the moist earth is light," were printed in the 1st edition, “The purple noon's transparent light," owing to the homoeographon “might” “light.”

Omissions through Simple Negligence.—Groups of letters, words, syllables and lines are often omitted without any contributory cause. Short words or such as are not necessary to the sense are especially prone thus to disappear.

Examples of omission. Shelley's Prometheus, iii. i, 70, “No refuge! No appeal! Sink with me [then];” Cenci, i. i, 26, “Respited [me] from Hell! So may the Devil | Respite their souls from Heaven!”; Hellas, 657, “Bask in the [deep] blue noon divine”; Julian and Maddalo, 218, where “Moans, shrieks, and curses, and blaspheming prayers” is absent in the earlier editions though required for the rhyme; so lines 299-301 of the Letter to Maria Gisborne.

Repetitions: Dittography.—Letters, groups of letters, words and lines may be written twice (or even oftener) instead of once.

Other repetitions of words already written and anticipations of words yet to be written are also found, through the scribe's eye wandering into the preceding or the following context. Wherever the word or group of words repeated is not the one that he has just copied loss is liable to occur.

Dittography is common enough in manuscripts but is usually detected in reading proofs. In the unique MS. of Cicero's treatise De Republica, 2, 33, 57, secutus appears as “secututus secutus.” Other kinds of repetition are Shelley's Witch of Atlas, 611 seq., “Like one asleep in a green hermitage, | With gentle sleep about its eyelids playing” (sleep for smiles has come from the previous line); Revolt of Islam, 4749, “Where” for “When” appears to have come from “Where” in 4750 or 4751. Often the word thus extruded is irrecoverable; Ginevra, 125 sqq., “The matin winds from the expanded flowers | Scatter their hoarded incense and awaken | The earth, until the dewy sleep is shaken | From every living heart which it possesses | Through seas and winds, cities and wildernesses”; the second “winds” is a repetition of the first, but what should stand in its place,—“lands” or “strands” or “waves” or something else—no one can say.

Confusions of Words.—Words are not only changed through confusion of single letters or abbreviations, but also through general resemblance or (a semi-voluntary change) through similarity of meaning.

Shelley, Prometheus, ii. 2, 53: “There streams a plume-uplifting wind” for “steams.” In Shelley's lines, When the lamp is shattered, vv. 5-6, “When the lute is broken, | Sweet tones are remembered not,” the printed edition had “notes” for “tones.” In Mrs Gaskell's Cranford, ch. xiv. (near the end), “The lunch—a hot savoury mutton-chop, and a little of the cold loin sliced and fried—was now brought in” is the reading of most if not all the editions; but “loin” should be “lion,” the reference being to the pudding, “a lion with currant eyes,” described earlier in the chapter. In Shelley's “Evening: Ponte al Mare, Pisa,” 20, “By darkest barriers of enormous cloud” for “cinereous”; “Hymn to Mercury” (trans.), 57, “And through the tortoise's hard strong skin” for “stony.” Shelley's “The Boat on the Serchio,” 117, “woods of stunted fir” for “pine” which the rhyme requires; Prince Athanase, 250, “And sea buds burst beneath the waves serene” for “under.”

The same character frequently attaches to transpositions of words and parts of words. The copyist does not as a general rule consciously intend a change, but he falls into one through the influence of dominant associations. He substitutes an order of words which, in respect of syntax, metre or rhythm is more familiar to him.

Transpositions of words, if not purely accidental, as in Chaucer, “Parson's Tale,” p. 689 (ed. Skeat), “God yaf (gave) his benison to Laban by the service of Jacob and to Pharao by the service of Joseph,” where the MSS. transpose Laban and Pharao, are generally to a more usual order, as in Shelley's Witch of Atlas, 65, “She first was changed” to “she was first changed.” An instance of transposition of words in part is in Shelley's “Invocation to Misery,” l. 27, “And mine arm shall be thy pillow,” where the 1st ed. had “thine arm” and “my pillow.”

Faulty Divisions of Words.—These will generally imply an exemplar in which the words were without any division or without a sufficient one. Under this head we may class errors which arise from the omission or the insertion of such marks as the apostrophe and the hyphen.

Examples of wrong division of words. Chaucer's House of Fame, iii., 1975, “Of good or misgovernement” which should be “mis (i.e., bad) governement”; Shelley's Prometheus, iii. 2, 22, “Round many peopled continents” for “many-peopled," ib. 26, “the light laden moon” for “light-laden”; Revolt of Islam, 4805, “Our bark hung there, as one line suspended | Between two heavens,” for “on a line.”

With this we may class faulty division of sentences. Wrong punctuation is a common error and usually easy to correct.

As an example of mispunctuation we may take Shelley's Triumph of Life, 188 sqq., “ ‘If thou can'st, forbear | To join the dance, which I had well forborne’ | Said the grim Feature of my thought ‘Aware | I will unfold,’ ” &c., for “said the grim Feature (of my thought aware) ‘I will unfold.’ ”

Grammatical Assimilations.—These are often purely mechanical errors: but they may be semi-voluntary or even voluntary, the copyist desiring to set the syntax right.

Examples: Shelley's Rosalind and Helen, 63, “A sound from thee, Rosalind dear” instead of there; Mask of Anarchy, 280 seq., “the daily strife | With common wants and common cares | Which sow the human heart with tares,” for “sows.”

Insertions (or Omissions) of Seemingly Unimportant Words.—These, inasmuch as they must often import some judgment on the sense of the passage copied, will be frequently semi-voluntary if not voluntary.

Examples: Shelley, Prometheus, iii. i, 5, “The soul of man like [an] unextinguished fire.” So in Triumph of Life, 265, “Whom from the flock of conquerors | Fame singled out for her thunder-bearing minion,” out seems to be due to the compositor.

False Recollections.—The passage which a copyist is reproducing may suggest to him something else and he will write down what is thus in his mind instead of what is before his eyes.

There is a noteworthy instance in Horace, Odes, iii. 18, 11 seq., “festus in pratis uacat otioso | cum boue pagus” where some MSS. give pardus, a reminiscence of Isaiah xi. 6, “The leopard (pardus) shall lie down with the kid.” In iv. i. 20, for “trabe citrea” many MSS. have “trabe Cypria,” which occurs in i. 1, 13.

Incorporation of Marginalia.—The copyist may erroneously suppose that something written in the margin, between the lines or at the top or the foot of the page which he is copying, is intended to be placed in the text. The words so incorporated may appear side by side with the genuine reading or they may expel it.

In Horace, Odes, iii. 27, 47, “amati | cornua monstri” (of the bull which carried off Europa), more than one MS. has “cornua tauri,” an explanation of monstri. The celebrated passage about the three heavenly witnesses inserted in the Epistle of St John (v. 2) seems to have been originally a comment explanatory of the text.

Transpositions of Lines and Passages.—This kind of transposition is really arrested loss. An accidental omission is discovered, and the person responsible, or another, places what is omitted in the margin at the foot of the page or in some other part of the text, usually adding a mark to show where it ought to have been. The next copyist may easily overlook this sign and thus the passage may be permanently displaced.

In Chaucer's Canterbury Tales, most MSS. place the couplet, “And eek of many another maner cryme | Which nedeth nat rehercen at this tyme,” which should stand after v. 8 of the “Friar's Tale,” in the Prologue to the Tale before the fourth line from the end. In the “Monk's Tale” a block of 88 lines (3565-3652) is transposed in most MSS. to follow 3956.

Interpolation.—This is the deliberate alteration of an exemplar by way of substitution, addition or omission, but when it takes the particular form of omission it is naturally very hard to detect. Interpolation then always has a motive. The most frequent motive is the removal of some difficulty in the sense, expression or metre of the text, and especially obvious gaps or corruptions which the interpolator endeavours to fill or to heal. Fraudulent interpolation, whether the fraud be pious or otherwise, does occur, but is comparatively rare. The removal or the mitigation of objectionable matter is also occasionally found. Interpolation is then a voluntary alteration, but in practice it is often hard to distinguish from other changes in which its motive is absent.

The usual character of scribes' alterations is well illustrated by a passage in Bacon's Advancement of Learning, II. xix., “For these critics have often presumed that that which they understand not is false set down: as the Priest that where he found it written of St Paul Demissus est per sportam” [Acts ix. 25] “mended his book, and made it Demissus est per portam, because sporta was an hard word, and out of his reading.” Shelley in Triumph of Life, 201 seq., wrote, “And if the spark with which Heaven lit my spirit | Had been with proper nutriment supplied,” but the printed editions made it “sentiment.” The transcript used for the printed edition of Marenghi apparently often corrupted what was rare and strange to what was commonplace; e.g., l. 119, “dewglobes” to “dewdrops.” Interpolation is sometimes due to an inopportune use of knowledge, as when a quotation or a narrative is made to agree with what the interpolator has read elsewhere. The text of the Septuagint, a translation of the Old Testament made from MSS. older than those accessible to Origen, was much altered by him in order to make it conform more closely to the Hebrew text with which he was familiar, and in the Synoptic Gospels changes are found, the aim of which is to “harmonize” the accounts given by the different evangelists. Deliberate alteration is occasionally due to disapproval of what stands in the text or even to less creditable reasons. There is an old and seemingly trustworthy tradition that some lines in Homer's “Catalogue of the Ships,” Iliad, ii. 553-555 and 558, were introduced there to gratify the vanity or ambition of the Athenians. Insertions of this or of a similar character may be of almost any length, from a few words to a whole chapter or a complete poem. Literary forgery has never set any bounds to itself, and the history of every literature will supply examples of entire works being foisted upon authors and personages of repute. A notable one was the Epistles of Phalaris, a late Greek forgery, demonstrated to be such by Bentley in a treatise which is a model of what such a demonstration should be.

Special Conditions conducing to Corruption.—The chief of these is strangeness or difficulty in the matter to be copied. Proper names, technical expressions, quotations from foreign languages, and frequent change of subject, are all likely to cause difficulty to a scribe and error in his work.

Careful and continuous regard to the various kinds of errors and defaults that are found in transcription will enable us to judge whether a reading which it is suggested stood in the archetype of our text is likely to have been corrupted to the reading, or readings, which stand in the extant manuscripts or editions. If it is, we say of this reading that it is transcriptionally probable.

Some precautions must be observed. First we must rule out any proposal which assumes confusions of letters and abbreviations which are not attested for the particular tradition. Secondly, since different scribes are prone to different kinds of error, we must ever bear in mind the particular failings of the scribes responsible for the transmission of our text as these failings are revealed in the apparatus criticus.

Maxims of criticism to which we may here refer are that “harder readings are better than easier” and that “the shorter reading is generally the truer.” The first maxim is indisputable, provided we understand by “harder” harder to the scribe, and by “easier” easier to the scribe. The characteristic of scribes' emendations or interpolations is that they are superficial. Their mark is that at the time of their making they “combine the appearance of improvement with the absence of its reality” (Westcott and Hort, New Testament, i. p. 27). The second maxim refers to the well-known fact that accretions from marginalia, &c., lengthen and at the same time weaken a text.

The virtues of a scribe are honesty and care (or in a single word fidelity) and intelligence. But it is rare to find these combined in a very high degree, and out of them we can least easily dispense with fidelity. Paradoxical as it may seem, the mechanical corruptions of a stupid but faithful copyist may tell us more than the intelligent copyings of a less faithful one.

A nice question is how far any alteration of the text of the exemplar is compatible with fidelity. Is a scribe, who recognizes under a corruption the word certainly intended, to perpetuate the error of the exemplar? Considering the liability of corruption to breed corruption we can hardly blame him if he does not, and we may say that it is no derogation to his fides if he makes self-evident corrections. But with these he must stop.

At certain epochs in the transmission of literature systematic efforts have been made to improve the transmitted texts, and these efforts have naturally been accompanied by a good deal of emendation both successful and unsuccessful. Such an epoch was the revival of Latin and Greek learning in the 15th century, and a modern scholar would for that reason naturally prefer to have a manuscript to work on, which was written immediately before this epoch to one which was written immediately after it.

The fidelity of a scribe has to be judged chiefly by internal tests, and these are best applied to his work in passages where there is no reasonable doubt of the correctness of the transmitted text. But there are two tests of a more objective character that may be used—orthography, and indication of lacunae or other faults in his exemplar. A scribe who preserves in his spelling the traces of a bygone age is probably trustworthy. If faithful in small things, he is likely to be faithful in great. A scribe again who scrupulously records the presence of a lacuna or illegibility in what he is copying, inspires us with confidence in the rest of his work.

As regards the use of testimonia, it may be observed to begin with that their value must depend on the trustworthiness of the texts of the writers from whom they are taken, and further upon that of the text used by the translator, the excerptor or the quoter, about which we can know nothing for certain, though we may sometimes make probable inferences. In the case of quotations we must allow for failures of memory.

Many times in the course of his investigations the critic will be confronted with problems which cannot be resolved by considerations of transcriptional or documental probability. To take an instance already referred to, it is not clear at first sight whether in the couplet from Propertius Scythiae is more likely to be a misrecollection of some text of the 1st century A.D., or Scythicis some scribe's assimilation which made its way into the transmitted text in the course of the next thousand years.

This leads us to consider Intrinsic Probability. By this is meant the likelihood that the writer of our text would at the time of writing have written, or not have written, a particular thing. Two questions which may be separated, though they are not entirely distinct, are here involved. What was the meaning of the writer? And how did he express it? The sense may be clear though the words may no longer be determinable.

A reading may be impugned on a number of grounds: that it gives no sense or an inappropriate sense, that it involves a usage or an idiom not current at the assumed time of writing, or foreign to the reputed author, or to the style in which he then was writing, that it involves some metrical or rhythmical anomaly, or that the connexion of thought which it produces is incoherent or disorderly. These charges cannot be played off against each other. It is no answer to the objection that a reading in some Roman poet makes nonsense to say that its Latinity is perfect or its metre excellent. But they may reinforce each other, and to such corroboration great weight must be assigned.

To set the meaning of a passage in a foreign language before us we must frequently have recourse to translation. But this method of representation is a very imperfect one; we may easily impose on ourselves and others by strained and ambiguous renderings. A more subtle danger to which we are especially liable in the case of a dead language is that of our acquiescing in a sense which satisfies us but which would not have satisfied the ancient writer. Above all we must avoid applying our own standards of taste, style and morality to the judgment of the text before us. The textual critic has no concern with what the writer ought to have thought or said; his business is solely with what he did say or think or might have said or thought. Amongst the legitimate reasons for suspecting the correctness of a text are patent contradictions in a passage or its immediate neighbourhood, proved and inexplicable deviations from the standards for forms, constructions and usages (mere rarity or singularity is not enough), weak and purposeless repetitions of a word (if there is no reason for attributing these to the writer), violations of the laws of metre and rhythm as observed by the author, obvious breaks in the thought (incoherence) or disorderly sequence in the same (double or multiple incoherence).

Where the critic has ascertained the earliest form of a reading in his text, he will apply to it the tests of intrinsic probability. No part of a text can be considered exempt from this scrutiny, though for a very large part of it it may be dispensed with. It should, however, be here observed, that whoever takes a reading without investigation, on the authority either of a manuscript or of a great scholar, or of a number of scholars, ceases for the time being to be a textual critic.

After every such critical examination four conclusions are possible—acceptance, doubt, rejection and alteration. In other words, a critic may deliberately pronounce that what stands in the text represents what the author wrote or might well have written, that it is doubtful whether it does, that it certainly does not, or, in the last event, that it may be replaced with certainty by something that does. In the three first cases his judgment will be governed by considerations of intrinsic probability alone: but in the last it must regard transcriptional probability as well. No alteration of a text, or emendation, is entitled to approval, unless in addition to providing the sense and diction required, it also presents a reading which the evidence furnished by the tradition shows might not improbably have been corrupted to what stands in the text. These tests, and these alone, are emendations bound to satisfy; but others are often tacitly imposed upon them. Of this the transposition of lines is the most notable example. This kind of change is troublesome to estimate and inconvenient to adopt, as it involves placing passages where we are not accustomed to look for them; but to the question, did the author write the passage here or there? the matter of our trouble or inconvenience is wholly irrelevant. There is, however, one class of cases in which no conclusion may be drawn, documental and intrinsic probability both failing us. This is where two alternative readings, neither of which can have come from the other, have equal external support and equal intrinsic merit. Isolated discrepancies of this kind may be due to some accident to our text at a period now beyond our power to trace. Numerous and striking discrepancies may be due to the fact that there was more than one edition or recension of it in early times, or to the author leaving his work in such a condition that such discrepancies must inevitably gain currency. In the case of dramas, different acting editions will give rise to them.

Up to this point all schools of textual criticism are theoretically at least in accord. But here begins a divergence which has done more than anything else to discredit the study with the outside world. It emerges because in all judgments on textual matters it is presupposed that they will be acted on, that a reading accepted will remain in the text, a rejected one obelized, enclosed between brackets or removed, and, in this last case, something else substituted in its place.

The “conservative” critic's chief concern is for the safety of the traditional and by preference the transmitted text. He urges very rightly that if alteration is carried beyond a certain point it cuts away its own foundation, and so all certainty is destroyed. His objective is the minimum of change. And as the need of making a text compels some sort of decision in every case, the “doubtful” readings of the tradition, some of which on the evidence would be doubtfully accepted and others doubtfully rejected, will all appear with the accepteds in the text. As to alterations (emendations) that are less than certain, his attitude is clearly if somewhat crudely expressed in the dictum that it is better to leave in the text “what if not the original reading is at least the remains of it.” The corresponding thesis of the opposite school would be that it is better to present to the reader something which the author might have written than something which he could not: or, in other words, that “stopgaps” should be preferred to débris.

An editor of a corrupt and disputed text may reasonably adopt either of two methods of procedure. He may present the text in the purest form which the external evidence warrants, and place all plausible suggestions for its improvement in notes or appendices. The text will be faithful but unreadable, and his work will be that of an honest man but of a textual antiquarian, not a textual critic, since he declines the duty of “the restoration of the text, as far as possible, to its original form.” On the other method the editor will provide all necessary information about the evidence for the text in the notes of his critical apparatus; but in the text itself he will give whatever in each case is supported by the balance of the probabilities. Each and every case he will decide on its own merits and without reference to decisions upon the other cases not now before him. Special consideration will be paid to “doubtful” readings, which will be distinguished in his work as “doubtfully accepted” or “doubtfully rejected.” Legitimate doubt arises when the evidence pro et contra of documental and intrinsic probability is equal, or nearly equal, or when documental probability points strongly to one side and intrinsic probability to another. Illegitimate doubt is the uncertainty of the doubter as to whether he has examined the whole of the evidence. Such doubt is much more frequently felt than acknowledged, and its effect upon critical work is highly injurious. On the one hand, it is apt to take refuge in an uncritical acceptance of the traditional readings, and, on the other hand, to produce a crop of hesitant and mutually destructive conjectures which a reader naturally resents as a needless waste of his time.

The so-called “conservative text” is neither an antiquarian's text nor a critic's text, but a compromise between the two. When it is conscientiously obtained, it is arrived at by handicapping, more or less heavily, intrinsic probability as compared with documental probability, or by raising the minimum of probability which shall qualify a reading for admission into the text until it is in agreement with the notions of the editor. Both of these procedures are arbitrary in their principle, and liable to be erratic in their application. The text will suffer whichever course is adopted, and it will suffer the more the more conservative is the editor, as may easily be shown. Thus, to take the latter one, if we suppose that of two editors of equal competence A requires a probability of four-fifths to admit a reading into his text and B a probability of three-fifths only, then in all the cases in which the probability lies between these two fractions B will be right seven times to A's three, while outside these limits there will be no difference between them.

Many persons appear to suppose that decisions upon doubtful points can be avoided by the expedient of leaving the traditional reading in possession of the text. The rule is a simple one and easy to apply. But owing to the constitution of the human mind it has consequences which possibly they have not contemplated. The great works of classical literature are not studied as pathological specimens, and they will be studied the less the more they contain to repel and disquiet the reader. If a corruption is left in a text when something might be substituted which would at least, as a “stopgap,” give the sort of sense required, then one of two things must happen. Either the sense of the passage is blotted out for the reader and the conservation of the corruption is tantamount to the expunging of the rest of the sentence, or else he will obtain the required sense by wresting the meaning of the other constituents of the context until they furnish it. So far so good: the requisite sense has been obtained, but the price has now to be paid. And the price is that the reader's perception of the signification of the word or words so wrested is dimmed and impaired, and his power of discriminating and understanding them when he meets them again is shot with doubt and error. In dealing with writings in dead languages this is particularly mischievous.

There are two reasons in particular why the part which emendation plays in the shaping of Greek and Latin texts is apt to be overlooked. Most people take their notions of a classical book not from its traditional form but from a “received” or vulgate text. This in the case of most writings is fairly readable, because it has been purged by the continuous emendation of scholars during several centuries. But the received conjectures which make this text acceptable have no more authority in themselves than equally good conjectures which have not yet won their way into the text, and it is clearly illogical to treat a text largely built upon conjectures as if it were now beyond the reach of conjecture. Again, it has often happened that readings which have been discovered by conjecture, and as such received into a text, have afterwards been found to have the support of MSS. Thus in one speech of Cicero, pro Caelio, some thirty conjectures of critics were found to be attested by a single recently discovered MS. Such readings it is now commonly the practice to transfer to the credit of the MS. and to suppress the fact that they were originally discovered by emendation. These confirmations, as they are called, should be carefully recorded in all critical texts, inasmuch as they constitute the most striking justification of the critical method.

Some examples from Shelley's poems are Prometheus, ii. 3, 50, “See'st thou shapes within the mist” (Zupitza for “I see thin shapes”); ib. iv. 4, 242, “Purple and azure, white and green, and golden” (and inserted by Rossetti); Prince Athanase, 150 sqq. “the rugged path | Where she once saw that horseman toil, with brief | And blighting hope, who with the news of death | Struck body and soul as with a mortal blight” (blighting, condemned by Rossetti, is cancelled in the Bodleian MS.).

It is a weakness of conservative critics to extol interpretation (or exegesis) at the expense of emendation. Some have even ventured to say that the successful defence of a passage in a text is a greater service than its successful correction. This is not true. The service to the text is the same, what was previously dark being now made clear. But the emendation deserves the higher praise as being in most instances the more difficult achievement. The fault of the opposite school, on the other hand, is to disparage interpretation and to regard correction as the proper field of a scholar and gentleman. This bias is reflected in the maxim that “correction should precede interpretation,” which is no more than a half-truth. For emendation must inevitably fail unless it express the meaning which the proper interpretation of the passage has shown to be required. Further, a corrector may propose the right word with the wrong meaning. Yet the custom is to give the credit of the emendation to him, and not to a successor who has seen what the right sense was and that this was the only word to express it, whereas the first scholar blundered once if not twice, first assigning the wrong sense to the passage and then selecting what (in most cases) would be the wrong word to express it. The proper course would be not to mention the first conjecturer or to mention him only for his error.

One of the most vexed questions of textual criticism, and one which divides scholars more perhaps than any other, is the question to what extent admitted imperfections and inconsistencies may properly be left in a text as due to the default of an author rather than of a scribe or compositor. No universal rule is here attainable. Each case must be considered on its merits; and the critic's procedure must of necessity be “eclectic”—an epithet often used with a tinge of reproach, the ground for which it is not easy to discover. Two general considerations may be indicated. If the autograph of a work is not accessible, there is no means of distinguishing between the involuntary errors of a scribe and the involuntary errors—“slips of pen”—of an author. For the latter are in fact only scribe's mistakes, the author being his own amanuensis. To take the example given under Confusions of Words above, loin for lion in Cranford is probably a printer's error, but it is conceivable that it is due to a deflexion of the authoress's mind or pen through the accidental proximity of the “mutton chop.”

Passing over this class we come to one about which there may frequently be serious doubt. What is clearly erroneous or faulty may as clearly be intended, and therefore not to be removed by the critic. In Chaucer's “Miller's Tale” (3451, 3457) astromie is used for astronomie, and Noë and Noël (Christmas) confused, “Nowélis flood” (3451, 3457), because the speaker is an illiterate carpenter. In the Prologue to the “Parson's Tale” (10) there is, on the other hand, a mistake of Chaucer's own, which no judicious critic would think of removing, the constellation Libra being said to be “the moon's exaltation” when it should be Saturn's. But this error in an astrological detail would not warrant us in assigning to the poet the blunder about Jacob and Laban in the same tale (see above). Much depends on the precision with which an error can be corrected: wherever there are more plausible ways than one of doing this, the faulty reading must be allowed to remain. Collateral as well as direct evidence must be obtained. If there are a number of instances where there is faultiness which is hard to remove, it is probable that the evil lies too deep for emendation. The author's own carelessness may be to blame, or, as in the case of Virgil and Lucan, he may not have been allowed to put the finishing touches to his work.

Certain lapses from grammatical correctness and metrical regularity that we find in the poems of Shelley are undoubtedly due to the author, though the number of these has been reduced (as Mr Buxton Forman has pointed out) with our improved knowledge of the sources of the text. Amongst such lapses we may instance Prince Athanase (287), “The shadow of thy moving wings imbue | Its deserts and its mountains”; “To a Skylark” (80), “Thou lovest—but ne'er knew love's sad satiety.” The solecism in the Preface to the Adonais, “My known repugnance to the narrow principles of taste on which several of his earlier compositions were modelled prove at least that I am an impartial judge,” would probably have been corrected by the poet if his attention had been called to it; but the two first ones, with others, cannot be thus regarded. We may detect occasional laxity also in his handling of his verse. Lines are left unrhymed: e.g., Julian and Maddalo (211); Rosalind and Helen (366). Or the same word is used in place of another rhyming word: Revolt of Islam (3573 and 3576, 3829 and 3831). In the Daemon of the World (341-2), Shelley himself cancelled a metrical reading for one that makes the verse a syllable too short. It is in this department of criticism that the personal equation has the freest play, and hence the natural adherents of either school of critics should be specially on their guard against their school's peculiar bias.

The part which conjectural emendation should play must obviously be very different in different texts. In the New Testament, for example, this part is very small indeed, though it cannot be altogether excluded. Colossians ii. 58 is corrupt as it appears; but the adoption of a correction recommended by Bishop Lightfoot and Dr C. Taylor will restore it to sense.

It has been maintained that emendation (being guessing) is no part of textual criticism at all, though judgment upon emendation is. The position approaches to paradox and is not likely to be generally accepted. But it does contain an element of truth and indicates a well-founded reproach against the majority of those who practise conjecture. Nothing has discredited emendation as a means of improving texts more than the want of method, common care and research, which those addicted to it show. Some of the most distinguished scholars have offended worst. The Milton of Bentley, England's greatest critic, is a by-word. To examine all the causes which may vitiate emendations would mean writing a treatise upon human frailty. But the reason why the vast majority of them fail is that the vast majority of them should never have been made at all. Their proposers do not take even elementary precautions to be right. An inquirer who examines the stars with a shilling telescope is not likely to make observations of value, and even a trained astronomer has to allow for his “personal equation”—a point to which even a finished critic rarely attends. Successful emendation requires a rare union of qualifications—insight, prudence, patience and familiarity with the author emended and the conditions of his text. If any of these is absent, the work is apt to be wasted.

Authority, as already hinted, has properly no place in textual criticism. For his facts a textual critic may, and often must, be beholden to others: but never for his opinions. It adds nothing to the evidence for a reading that it has been approved by a Lachmann or a Madvig or rejected by a Stoeber or a Carutti: and an appeal to names on any such question confuses issues and deters inquiry. But inasmuch as there are many persons, including most makers of school editions, who prudently and modestly desire a better road to truth than their own investigations can discover and think thus to find it, it will not be amiss to observe on the one hand that the concurrence of a succession of editors in a reading is no proof and often no presumption either that their agreement is independent or that their reading is right; and on the other that, though independence may generally be granted to coinciding emendations of different scholars, yet from the general constitution of the human mind it is likely that not a few of these will be coincidences in error rather than in truth.

One of the marks of a great textual critic is his attention to details. He will not consider his work upon the text complete until he has made it, as far as he can, such as the author would approve in every particular. Accordingly he will restore the spelling of the author if that can be ascertained: he will not accept the corruptions which have been introduced into it by copyists or printers, even though these may not affect its sense, nor will he modernize it so as to bring it into harmony with that of a later and to him a more familiar age. Thus, to take an example, he will not print a critical text of Plautus with two letters (Y and Z) which were no part of the Latin alphabet in the age of that comedian; still less will he introduce into Latin texts distinctions, such as i, j and u, v, which were not used till long after the middle ages.

As time goes on, textual criticism will have less and less to do. In the old texts its work will have been performed so far as it is performable. What is left will be an obstinate remainder of difficulties, for which there is no solution or only too many. In the newer texts, on the other hand, as experience has already shown, it will have from the outset but a very contracted field.

(J . P. P.)

1. For the convenience of the general reader these errors have been illustrated as far as possible from English authors and especially from the poems of Shelley (ed. Hutchinson).