The New Testament in the original Greek - Introduction and Appendix (1882)/Part II

by Fenton John Anthony Hort (primary), Brooke Foss Westcott

 

PART II

THE METHODS OF TEXTUAL CRITICISM

23. Every method of textual criticism corresponds to some one class of textual facts: the best criticism is that which takes account of every class of textual facts, and assigns to each method its proper use and rank. The leading principles of textual criticism are identical for all writings whatever. Differences in application arise only from differences in the amount, variety, and quality of evidence: no method is ever inapplicable except through defectiveness of evidence. The more obvious facts naturally attract attention first; and it is only at a further stage of study that any one is likely spontaneously to grasp those more fundamental facts from which textual criticism must start if it is to reach comparative certainty. We propose to follow here this natural order, according to which the higher methods will come last into view.

SECTION I. INTERNAL EVIDENCE OF READINGS

24—37

24. Criticism arises out of the question what is to be received where a text is extant in two or more varying documents. The most rudimentary form of criticism consists in dealing with each variation independently, and adopting at once in each case out of two or more variants that which looks most probable. The evidence here taken into account is commonly called 'Internal Evidence': as other kinds of Internal Evidence will have to be mentioned, we prefer to call it more precisely 'Internal Evidence of Readings'. Internal Evidence of Readings is of two kinds, which cannot be too sharply distinguished from each other; appealing respectively to Intrinsic Probability, having reference to the author, and what may be called Transcriptional Probability, having reference to the copyists. In appealing to the first, we ask what an author is likely to have written: in appealing to the second, we ask what copyists are likely to have made him seem to write. Both these kinds of evidence are alike in the strictest sense internal, since they are alike derived exclusively from comparison of the testimony delivered, no account being taken of any relative antecedent credibility of the actual witnesses.

A. 25—27. Intrinsic Probability

25. The first impulse in dealing with a variation is usually to lean on Intrinsic Probability, that is, to consider which of two readings makes the best sense, and to decide between them accordingly. The decision may be made either by an immediate and as it were intuitive judgement, or by weighing cautiously various elements which go to make up what is called sense, such as conformity to grammar and congruity to the purport of the rest of the sentence and of the larger context; to which may rightly be added congruity to the usual style of the author and to his matter in other passages. The process may take the form either of simply comparing two or more rival readings under these heads, and giving the preference to that which appears to have the advantage, or of rejecting a reading absolutely, for violation of one or more of the congruities, or of adopting a reading absolutely, for perfection of congruity.

26. These considerations evidently afford reasonable presumptions; presumptions which in some cases may attain such force on the negative side as to demand the rejection or qualify the acceptance of readings most highly commended by other kinds of evidence. But the uncertainty of the decision in ordinary cases is shown by the great diversity of judgement which is actually found to exist. The value of the Intrinsic Evidence of Readings should of course be estimated by its best and most cultivated form, for the extemporaneous surmises of an ordinary untrained reader will differ widely from the range of probabilities present to the mind of a scholar prepared both by general training in the analysis of texts and by special study of the facts bearing on the particular case. But in dealing with this kind of evidence equally competent critics often arrive at contradictory conclusions as to the same variations.

27. Nor indeed are the assumptions involved in Intrinsic Evidence of Readings to be implicitly trusted. There is much literature, ancient no less than modern, in which it is needful to remember that authors are not always grammatical, or clear, or consistent, or felicitous; so that not seldom an ordinary reader finds it easy to replace a feeble or half-appropriate word or phrase by an effective substitute; and thus the best words to express an author's meaning need not in all cases be those which he actually employed. But, without attempting to determine the limits within which such causes have given occasion to any variants in the New Testament, it concerns our own purpose more to urge that in the highest literature, and notably in the Bible, all readers are peculiarly liable to the fallacy of supposing that they understand the author's meaning and purpose because they understand some part or some aspect of it, which they take for the whole; and hence, in judging variations of text, they are led unawares to disparage any word or phrase which owes its selection by the author to those elements of the thought present to his mind which they have failed to perceive or to feel.

B. 28—37. Transcriptional Probability

28. The next step in criticism is the discovery of Transcriptional Probability, and is suggested by the reflexion that what attracts ourselves is not on the average unlikely to have attracted transcribers. If one various reading appears to ourselves to give much better sense or in some other way to excel another, the same apparent superiority may have led to the introduction of the reading in the first instance. Mere blunders apart, no motive can be thought of which could lead a scribe to introduce consciously a worse reading in place of a better. We might thus seem to be landed in the paradoxical result that intrinsic inferiority is evidence of originality.

29. In reality however, although this is the form in which the considerations that make up Transcriptional Probability are likely in the first instance to present themselves to a student feeling his way onwards beyond Intrinsic Probability, the true nature of Transcriptional Probability can hardly be understood till it is approached from another side. Transcriptional Probability is not directly or properly concerned with the relative excellence of rival readings, but merely with the relative fitness of each for explaining the existence of the others. Every rival reading contributes an element to the problem which has to be solved; for every rival reading is a fact which has to be accounted for, and no acceptance of any one reading as original can be satisfactory which leaves any other variant incapable of being traced to some known cause or causes of variation. If a variation is binary, as it may be called, consisting of two variants, a and b, the problem for Transcriptional Probability to decide is whether it is easier to derive b from a, through causes of corruption known to exist elsewhere, on the hypothesis that a is original, or to derive a from b through similar agencies, on the hypothesis that b is original. If the variants are more numerous, making a ternary or yet more composite variation, each in its turn must be assumed as a hypothetical original, and an endeavour made to deduce from it all the others, either independently or consecutively; after which the relative facilities of the several experimental deductions must be compared together.

30. Hence the basis on which Transcriptional Probability rests consists of generalisations as to the causes of corruption incident to the process of transcription. A few of the broadest generalisations of this kind, singling out observed proclivities of average copyists, make up the bulk of what are not very happily called 'canons of criticism'. Many causes of corruption are independent of age and language, and their prevalence may be easily verified by a careful observer every day; while others are largely modified, or even brought into existence, by peculiar circumstances of the writings themselves, or of the conditions of their transmission. There is always an abundance of variations in which no practised scholar can possibly doubt which is the original reading, and which must therefore be derivative; and these clear instances supply ample materials for discovering and classifying the causes of corruption which must have been operative in all variations. The most obvious causes of corruption are clerical or mechanical, arising from mere carelessness of the transcriber, chiefly through deceptions of eye or ear. But, as we have seen (§ 10), the presence of a mental factor can often be traced in corruptions partly mechanical; and under the influence of a lax conception of the proper office of a transcriber distinctly mental causes of change may assume, and often have assumed, very large proportions. Even where the definite responsibilities of transcription were strongly felt, changes not purely clerical would arise from a more or less conscious feeling on a scribe's part that he was correcting what he deemed an obvious error due to some one of his predecessors; while, at times or places in which the offices of transcribing and editing came to be confused, other copyists would not shrink from altering the form of what lay before them for the sake of substituting what they supposed to be a clearer or better representation of the matter.

31. The value of the evidence obtained from Transcriptional Probability is incontestable. Without its aid textual criticism could rarely attain any high degree of security. Moreover, to be rightly estimated, it must be brought under consideration in the higher form to which it can be raised by care and study, when elementary guesses as to which reading scribes are likely in any particular case to have introduced have been replaced by judgements founded on previous investigation of the various general characteristics of those readings which can with moral certainty be assumed to have been introduced by scribes. But even at its best this class of Internal Evidence, like the other, carries us but a little way towards the recovery of an ancient text, when it is employed alone. The number of variations in which it can be trusted to supply by itself a direct and immediate decision is relatively very small, when unquestionable blunders, that is, clerical errors, have been set aside. If we look behind the canons laid down by critics to the observed facts from which their authority proceeds, we find, first, that scribes were moved by a much greater variety of impulse than is usually supposed; next, that different scribes were to a certain limited extent moved by different impulses; and thirdly, that in many variations each of two or more conflicting readings might be reasonably accounted for by some impulse known to have operated elsewhere. In these last cases decision is evidently precarious, even though the evidence may seem to be stronger on the one side than the other. Not only are mental impulses unsatisfactory subjects for estimates of comparative force; but a plurality of impulses recognised by ourselves as possible in any given case by no means implies a plurality of impulses as having been actually in operation. Nor have we a right to assume that what in any particular case we judge after comparison to be the intrinsically strongest of the two or more possible impulses must as a matter of course be the one impulse which acted on a scribe if he was acted on by one only: accidental circumstances beyond our knowledge would determine which impulse would be the first to reach his mind or hand, and there would seldom be room for any element of deliberate choice. But even where there is no conflict of possible impulses, the evidence on the one side is often too slight and questionable to be implicitly trusted by any one who wishes to ascertain his author's true text, and not merely to follow a generally sound rule. Hence it is only in well marked and unambiguous cases that the unsupported verdict of Transcriptional Probability for detached readings can be safely followed.

32. But the insufficiency of Transcriptional Probability as an independent guide is most signally shown by its liability to stand in apparent antagonism to Intrinsic Probability; since the legitimate force of Intrinsic Probability, where its drift is clear and unambiguous, is not touched by the fact that in many other places it bears a divided or ambiguous testimony. The area of final antagonism, it is already evident, is very much smaller than might seem to be implied in the first crude impression that scribes are not likely to desert a better reading for a worse; but it is sufficiently large to create serious difficulty. The true nature of the difficulty will be best explained by a few words on the mutual relations of the two classes of Internal Evidence, by which it will likewise be seen what a valuable ancillary office they discharge in combination.

33. All conflicts between Intrinsic and Transcriptional Probability arise from the imperfection of our knowledge: in both fields criticism consists of inferences from more or less incomplete data. Every change not purely mechanical made by a transcriber is, in some sense, of the nature of a correction. Corrections in such external matters as orthography and the like may be passed over, since they arise merely out of the comparative familiarity of different forms, and here Intrinsic Probability has nothing to do with what can properly be called excellence or easiness. All other corrections, that is, those which bear any relation to sense, would never be made unless in the eyes of the scribe who makes them they were improvements in sense or in the expression of sense: even when made unconsciously, it is the relative satisfaction which they give to his mental state at the time that creates or shapes them. Yet in literature of high quality it is as a rule improbable that a change made by transcribers should improve an author's sense, or express his full and exact sense better than he has done himself. It follows that, with the exception of pure blunders, readings originating with scribes must always at the time have combined the appearance of improvement with the absence of its reality. If they had not been plausible, they would not have existed: yet their excellence must have been either superficial or partial, and the balance of inward and essential excellence must lie against them. In itself therefore Transcriptional Probability not only stands in no antagonism to Intrinsic Probability, but is its sustaining complement. It is seen in its proper and normal shape when both characteristics of a scribe's correction can alike be recognised, the semblance of superiority and the latent inferiority.

34. It is only in reference to mental or semi-mental causes of corruption that the apparent conflict between Transcriptional and Intrinsic Probability has any place: and neither the extent nor the nature of the apparent conflict can be rightly understood if we forget that, in making use of this class of evidence, we have to do with readings only as they are likely to have appeared to transcribers, not as they appear to us, except in so far as our mental conditions can be accepted as truly reflecting theirs. It is especially necessary to bear this limitation in mind with reference to one of the most comprehensive and also most widely prevalent mental impulses of transcribers, the disposition to smooth away difficulties; which is the foundation of the paradoxical precept to 'choose the harder reading', the most famous of all 'canons of criticism'. Readings having no especial attractiveness to ourselves may justly be pronounced suspicious on grounds of Transcriptional Probability, if they were likely to be attractive, or their rivals unacceptable, to ancient transcribers; and conversely, if this condition is absent, we can draw no unfavourable inferences from any intrinsic excellence which they may possess in our own eyes.

35. The rational use of Transcriptional Probability as textual evidence depends on the power of distinguishing the grounds of preference implied in an ancient scribe's substitution of one reading for another from those felt as cogent now after close and deliberate criticism. Alterations made by transcribers, so far as they are due to any movement of thought, are with rare exceptions the product of first thoughts, not second; nor again of those first thoughts, springing from a rapid and penetrating glance over a whole field of evidence, which sometimes are justified by third thoughts. This is indeed a necessary result of the extemporaneous, cursory, and one-sided form which criticism cannot but assume when it exists only as a subordinate accident of transcription. But even the best prepared textual critic has to be on his guard against hasty impressions as to the intrinsic character of readings, for experience teaches him how often the relative attractiveness of conflicting readings becomes inverted by careful study. What we should naturally expect, in accordance with what has been said above (§ 33), is that each reading should shew some excellence of its own, apparent or real, provided that we on our part are qualified to recognise it. If any reading fails to do so, clerical errors being of course excepted, the fault must lie in our knowledge or our perception; for if it be a scribe's correction, it must have some at least apparent excellence, and if it be original, it must have the highest real excellence. Contrast of real and apparent excellence is in any given variation an indispensable criterion as to the adequacy of the evidence for justifying reliance on Transcriptional Probability.

36. Fortunately variations conforming to this normal type are of frequent occurrence; variations, that is, in which a critic is able to arrive at a strong and clear conviction that one reading is intrinsically much the most probable, and yet to see with equal clearness how the rival reading or readings could not but be attractive to average transcribers. In these cases Internal Evidence of Readings attains the highest degree of certainty which its nature admits, this relative trustworthiness being due to the coincidence of the two independent Probabilities, Intrinsic and Transcriptional. Readings thus certified are of the utmost value in the application of other methods of criticism, as we shall see hereafter.

37. But a vast proportion of variations do not fulfil these conditions. Where one reading (a) appears intrinsically preferable, and its excellence is of a kind that we might expect to be recognised by scribes, while its rival (b) shews no characteristic likely to be attractive to them. Intrinsic and Transcriptional Probability are practically in conflict. In such a case either b must be wrong, and therefore must, as compared with a, have had some attractiveness not perceived by us, if the case be one in which the supposition of a mere blunder is improbable; or b must be right, and therefore must have expressed the author's meaning with some special fitness which escapes our notice. The antagonism would disappear if we could discover on which side we have failed to perceive or duly appreciate all the facts; but in the mean time it stands. Occasionally the Intrinsic evidence is so strong that the Transcriptional evidence may without rashness be disregarded: but such cases are too exceptional to count for much when we are estimating the general trustworthiness of a method; and the apparent contradiction which the imperfection of our knowledge often leaves us unable to reconcile remains a valid objection against habitual reliance on the sufficiency of Internal Evidence of Readings.

SECTION II. INTERNAL EVIDENCE OF DOCUMENTS

38–48

38. Thus far we have been considering the method which follows Internal Evidence of Readings alone, as improved to the utmost by the distinction and separate appreciation of Intrinsic and Transcriptional Probability, and as applied with every aid of scholarship and special study. The limitation to Internal Evidence of Readings follows naturally from the impulse to deal conclusively at once with each variation as it comes in its turn before a reader or commentator or editor: yet a moment's consideration of the process of transmission shews how precarious it is to attempt to judge which of two or more readings is the most likely to be right, without considering which of the attesting: documents or combinations of documents are the most likely to convey an unadulterated transcript of the original text; in other words, in dealing with matter purely traditional, to ignore the relative antecedent credibility of witnesses, and trust exclusively to our own inward power of singling out the true readings from among their counterfeits, wherever we see them. Nor is it of much avail to allow supposed or ascertained excellence of particular documents a deciding voice in cases of difficulty, or to mix evidence of this kind at random or at pleasure with Internal Evidence of Readings assumed in practice if not in theory as the primary guide. The comparative trustworthiness of documentary authorities constitutes a fresh class of facts at least as pertinent as any with which we have hitherto been dealing, and much less likely to be misinterpreted by personal surmises. The first step towards obtaining a sure foundation is a consistent application of the principle that knowledge of documents should precede final judgement upon readings.

39. The most prominent fact known about a manuscript is its date, sometimes fixed to a year by a note from the scribe’s hand, oftener determined within certain limits by palæographical or other indirect indications, sometimes learned from external facts or records. Relative date, as has been explained above (§ 8), affords a valuable presumption as to relative freedom from corruption, when appealed to on a large scale; and this and other external facts, insufficient by themselves to solve a question of reading, may often supply essential materials to the process by which it can be solved. But the occasional preservation of comparatively ancient texts in comparatively modern MSS forbids confident reliance on priority of date unsustained by other marks of excellence.

40. The first effectual security against the uncertainties of Internal Evidence of Readings is found in what may be termed Internal Evidence of Documents, that is, the general characteristics of the texts contained in them as learned directly from themselves by continuous study of the whole or considerable parts. This and this alone supplies entirely trustworthy knowledge as to the relative value of different documents. If we compare successively the readings of two documents in all their variations, we have ample materials for ascertaining the leading merits and defects of each. Readings authenticated by the coincidence of strong Intrinsic and strong Transcriptional Probability, or it may be by one alone of these Probabilities in exceptional strength and clearness and uncontradicted by the other, are almost always to be found sufficiently numerous to supply a solid basis for inference. Moreover they can safely be supplemented by provisional judgements on similar evidence in the more numerous variations where a critic cannot but form a strong impression as to the probabilities of reading, though he dare not trust it absolutely. Where then one of the documents is found habitually to contain these morally certain or at least strongly preferred readings, and the other habitually to contain their rejected rivals, we can have no doubt, first, that the text of the first has been transmitted in comparative purity, and that the text of the second has suffered comparatively large corruption; and next, that the superiority of the first must be as great in the variations in which Internal Evidence of Readings has furnished no decisive criterion as in those which have enabled us to form a comparative appreciation of the two texts. By this cautious advance from the known to the unknown we are enabled to deal confidently with a great mass of those remaining variations, open variations, so to speak, the confidence being materially increased when, as usually happens, the document thus found to have the better text is also the older. Inference from the ascertained character of other readings within the identical text, transmitted, it is to be assumed, throughout under identical conditions, must have a higher order of certainty than the inferences dependent on general probabilities which in most cases make up Internal Evidence of Readings.

41. The method here followed differs, it will be observed, from that described above in involving not a single but a threefold process. In the one case we endeavour to deal with each variation separately, and to decide between its variants immediately, on the evidence presented by the variation itself in its context, aided only by general considerations. In the other case we begin with virtually performing the same operation, but only tentatively, with a view to collect materials, not final results: on some variations we can without rashness predict at this stage our ultimate conclusions; on many more we can estimate various degrees of probability; on many more again, if we are prudent, we shall be content to remain for the present in entire suspense. Next, we pass from investigating the readings to investigating the documents by means of what we have learned respecting the readings. Thirdly, we return to the readings, and go once more over the same ground as at first, but this time making a tentative choice of readings simply in accordance with documentary authority. Where the results coincide with those obtained at the first stage, a very high degree of probability is reached, resting on the coincidence of two and often three independent kinds of evidence. Where they differ at first sight, a fresh study of the whole evidence affecting the variation in question is secured. Often the fresh facts which it brings to light will shew the discordance between the new and the old evidence to have been too hastily assumed. Sometimes on the other hand they will confirm it, and then the doubt must remain.

42. To what extent documentary authority alone may be trusted, where the Internal Evidence of Readings is altogether uncertain, must vary in different instances. The predominantly purer text of one document may undoubtedly contain some wrong readings from which the predominantly less pure text of another is free. But the instances of this kind which are ultimately found to stand scrutiny are always much fewer than a critic’s first impression leads him to suppose; and in a text of any length we believe that only a plurality of strong instances confirming each other after close examination ought to disturb the presumption in favour of the document found to be habitually the better. Sometimes of course the superiority may be so slight or obscure that the documentary authority loses its normal weight. In such cases Internal Evidence of Readings becomes of greater relative importance: but as its inherent precariousness remains undiminished, the total result is comparative uncertainty of text.

43. Both the single and the triple processes which we have described depend ultimately on judgements upon Internal Evidence of Readings; but the difference between isolated judgements and combined judgements is vital. In the one case any misapprehension of the immediate evidence, that is, of a single group of individual phenomena, tells in full force upon the solitary process by which one reading is selected from the rest for adoption, and there is no room for rectification. In the other case the selection is suggested by the result of a large generalisation about the documents, verified and checked by the immediate evidence belonging to the variation; and the generalisation itself rests on too broad a foundation of provisional judgements, at once confirming and correcting each other, to be materially weakened by the chance or probability that some few of them are individually unsound.

44. Nevertheless the use of Internal Evidence of Documents has uncertainties of its own, some of which can be removed or materially diminished by special care and patience in the second and third stages of the process, while others are inherent and cannot be touched without the aid of a fresh kind of evidence. They all arise from the fact that texts are, in one sense or another, not absolutely homogeneous. Internal knowledge of documents that are compared with each other should include all their chief characteristics, and these can only imperfectly be summed up under a broad statement of comparative excellence. At first sight the sole problem that presents itself is whether this document is ‘better’ or ‘worse’ than that; and this knowledge may sometimes suffice to produce a fair text, where the evidence itself is very simple. Yet it can never be satisfactory either to follow implicitly a document pronounced to be ‘best’, or to forsake it on the strength of internal evidence for this or that rival reading. Every document, it may be safely said, contains errors; and second only to the need of distinguishing good documents from bad is the need of leaving as little room as possible for caprice in distinguishing the occasional errors of ‘good’ documents from the sound parts of their text.

45. General estimates of comparative excellence are at once shown to be insufficient by the fact that excellence itself is of various kinds: a document may be ‘good’ in one respect and ‘bad’ in another. The distinction between soundness and correctness, for instance, lies on the surface. One MS will transmit a substantially pure text disfigured by the blunders of a careless scribe, another will reproduce a deeply adulterated text with smooth faultlessness. It therefore becomes necessary in the case of important MSS to observe and discriminate the classes of clerical errors by which their proper texts are severally disguised; for an authority representing a sound tradition can be used with increased confidence when its own obvious slips have been classed under definite heads, so that those of its readings which cannot be referred to any of these heads must be reasonably supposed to have belonged to the text of its exemplar. The complexity of excellence is further increased by the unequal distribution of the mental or semi-mental causes of corruption; while they too can be observed, classified, and taken into account, though with less precision than defects of mechanical accuracy. Where the documentary witnesses are not exclusively MSS having continuous texts in the original language, but also, for instance, translations into other languages or quotations by later authors, similar deductions are required in order to avoid being misled as to the substantive text of their exemplars. Thus allowance has to be made for the changes of phraseology, real or apparent, which translators generally are prone to introduce, and again for those which may be due to the defects or other peculiarities of a given language, or the purpose of a given translation. In quotations account must in like manner be taken of the modifications, intentional or unconscious, which writers are apt to make in passages which they rapidly quote, and again of the individual habits of quotation found in this or that particular writer. In all these cases on the one hand comparative excellence is various and divided; and on the other an exact study of documents will go a great way towards changing vague guesses about possible errors into positive knowledge of the limits within which undoubted errors have been actually found to exist. The corrective process is strictly analogous to that by which evidence from Transcriptional Probability is acquired and reduced to order: but in the present case there is less liability to error in application, because we are drawing inferences not so much from the average ways of scribes as a class as from the definite characteristics of this or that documentary witness.

46. The true range of individuality of text cannot moreover be exactly measured by the range of contents of an existing document. We have no right to assume without verification the use of the same exemplar or exemplars from the first page to the last. A document containing more books than one may have been transcribed either from an exemplar having identical contents, or from two or more exemplars each of which contained a smaller number of books; and these successive exemplars may have been of very various or unequal excellence. As regards alterations made by the transcriber himself, a generalisation obtained from one book would be fairly valid for all the rest. But as regards what is usually much more important, the antecedent text or texts received by him, the prima facie presumption that a generalisation obtained in one book will be applicable in another cannot safely be trusted until the recurrence of the same textual characteristics has been empirically ascertained.

47. A third and specially important loss of homogeneousness occurs wherever the transmission of a writing has been much affected by what (§§ 5, 6) we have called mixture, the irregular combination into a single text of two or more texts belonging to different lines of transmission. Where books scattered in two or more copies are transcribed continuously into a single document (§ 46), the use of different exemplars is successive: here it is simultaneous. In this case the individuality, so to speak, of each mixed document is divided, and each element has its own characteristics; so that we need to know to which element of the document any given reading belongs, before we can tell what authority the reading derives from its attestation by the document. Such knowledge evidently cannot be furnished by the document itself; but, as we shall see presently, it may often be obtained through combinations of documents.

48. Lastly; the practical value of the simple application of Internal Evidence of Documents diminishes as they increase in number. It is of course in some sort available wherever a text is preserved in more than a single document, provided only that it is known in each variation which readings are supported by the several documents. Wherever it can be used at all, its use is indispensable at every turn; and where the documents are very few and not perceptibly connected, it is the best resource that criticism possesses. On the other hand, its direct utility varies with the simplicity of the documentary evidence; and it is only through the disturbing medium of arbitrary and untrustworthy rules that it can be made systematically available for writings preserved in a plurality of documents. For such writings in fact it can be employed as the primary guide only where the better documents are in tolerably complete agreement against the worse; and the insufficiency must increase with their number and diversity. Wherever the better documents are ranged on different sides, the decision becomes virtually dependent on the uncertainties of isolated personal judgements. here is evidently no way through the chaos of complex attestation which thus confronts us except by going back to its causes, that is, by enquiring what antecedent circumstances of transmission will account for such combinations of agreements and differences between the several documents as we find actually existing. In other words, we are led to the necessity of investigating not only individual documents and their characteristics, but yet more the mutual relations of documents,

SECTION III. GENEALOGICAL EVIDENCE

49—76

A. 49—53. Simple or divergent genealogy

49. The first great step in rising above the uncertainties of Internal Evidence of Readings was taken by ceasing to treat Readings independently of each other, and examining them connectedly in series, each series being furnished by one of the several Documents in which they are found. The second great step, at which we have now arrived, consists in ceasing to treat Documents independently of each other, and examining them connectedly as parts of a single whole in virtue of their historical relationships. In their prima facie character documents present themselves as so many independent and rival texts of greater or less purity. But as a matter of fact they are not independent: by the nature of the case they are all fragments, usually casual and scattered fragments, of a genealogical tree of transmission, sometimes of vast extent and intricacy. The more exactly we are able to trace the chief ramifications of the tree, and to determine the places of the several documents among the branches, the more secure will be the foundations laid for a criticism capable of distinguishing the original text from its successive corruptions. It may be laid down then emphatically, as a second principle, that all trustworthy restoration of corrupted texts is founded on the study of their history, that is, of the relations of descent or affinity which connect the several documents. The principle here laid down has long been acted upon in all the more important restorations of classical texts: but it is still too imperfectly understood to need no explanation. A simple instance will show at once its practical bearing.

50. Let it be supposed that a treatise exists in ten MSS. If they are used without reference to genealogy by an editor having a general preference for documentary evidence, a reading found in nine of them will in most eases be taken before a rival reading found only in the tenth, which will naturally be regarded as a casual aberration. If the editor decides otherwise, he does so in reliance on his own judgement either as to the high probability of the reading or as to the high excellence of the MS. He may be right in either case, and in the latter case he is more likely to be right than not: but where an overwhelming preponderance of the only kind of documentary evidence recognised is so boldly disregarded, a wide door is opened for dangerous uncertainty.

51. Another editor begins by studying the relations of the MSS, and finds sufficient evidence, external or internal, for believing that the first nine MSS were all copied directly or indirectly from the tenth MS, and derived nothing from any document independent of the tenth. He will then know that all their variations from the tenth can be only corruptions (successful cursory emendations of scribes being left out of account), and that for documentary evidence he has only to follow the tenth. Apart therefore from corruptions in the tenth, for the detection of which he can obviously have no documentary evidence, his text will at once be safe and true.

52. If however the result of the second supposed editor's study is to find that all the nine MSS were derived not from the tenth but from another lost MS, his ten documents resolve themselves virtually into two witnesses; the tenth MS, which he can know directly and completely, and the lost MS, which he must restore through the readings of its nine descendants, exactly and by simple transcription where they agree, approximately and by critical processes where they disagree. After these processes some few variations among the nine may doubtless be left in uncertainty, but the greater part will have been cleared away, leaving the text of the lost MS (with these definite exceptions) as certain as if it were accessible to the eyes. Where the two ultimate witnesses agree, the text will be as certain as the extant documents can make it; more certain than if the nine MSS had been derived from the tenth, because going back to an earlier link of transmission, the common source of the two witnesses. This common source may indeed be of any date not later than the earliest of the MSS, and accordingly separated from the autograph by any number of transcriptions, so that its text may vary from absolute purity to any amount of corruption: but as conjecture is the sole possible instrument for detecting or correcting whatever errors it may contain, this common source is the only original with which any of the methods of criticism now under discussion have any concern. Where the two ultimate witnesses differ, the genealogical method ceases to be applicable, and a comparison of the intrinsic general character of the two texts becomes the only resource.

53. The relations of descent between existing documents are rarely so simple as in the case supposed. To carry the supposition only one step further, the nine MSS might have been found to fall into two sets, five descended from one lost ancestor and four from another: and then the question would have arisen whether any two of the three authorities had a common origin not shared by the third. If it were ascertained that they had, the readings in which they agreed against the third would have no greater probability than the rival readings of the third, except so far as their common ancestor was found to have higher claims to authority as a single document than the third as a single document. If on the other hand the nine could not be traced to less than two originals, a certain much diminished numerical authority would still remain to them. Since however all presumptions from numerical superiority, even among documents known to be all absolutely independent, that is, derived from the autograph each by a separate line of descent, are liable to be falsified by different lengths and different conditions of transmission, the practical value of the numerical authority of the two supposed witnesses against the third could not be estimated till it had been brought into comparison with the results yielded by the Internal Evidence of all three witnesses.

B. 54—37. Genealogy and Number

54. It is hardly necessary to point out the total change in the bearing of the evidence here made by the introduction of the factor of genealogy. Apart from genealogy, the one MS becomes easily overborne by the nine; and it would be trusted against their united testimony only when upheld by strong internal evidence, and then manifestly at great risk. But if it is found that the nine had a common original, they sink jointly to a numerical authority not greater than that of the one; nay rather less, for that one is known absolutely, while the lost copy is known only approximately. Where for want of sufficiently clear evidence, or for any other reason, the simplification of pedigree cannot be carried thus far, still every approximation to an exhibition of their actual historical relations presents them in a truer light for the purposes of textual criticism than their enumeration in their existing form as so many separate units. It enables us on the one hand to detect the late origin and therefore irrelevance of some part of the prima facie documentary evidence, and on the other to find the rest of it already classified for us by the discovered relations of the attesting documents themselves, and thus fitted to supply trustworthy presumptions, and under favourable circumstances much more than presumptions, as a basis for the consideration of other classes of evidence.

55. It would be difficult to insist too strongly on the transformation of the superficial aspects of numerical authority thus effected by recognition of Genealogy. In the crude shape in which numerical authority is often presented, it rests on no better foundation than a vague transference of associations connected with majorities of voices, this natural confusion being aided perhaps by the application of the convenient and in itself harmless term ‘authorities’ to documents. No one doubts that some documents are better than others, and that therefore a numerical preponderance may have rightly to yield to a qualitative preponderance. But it is often assumed that numerical superiority, as such, among existing documents ought always to carry a certain considerable though perhaps subordinate weight, and that this weight ought always to be to a certain extent proportionate to the excess of numbers. This assumption is completely negatived by the facts adduced in the preceding pages, which shew that, since the same numerical relations among existing documents are compatible with the utmost dissimilarity in the numerical relations among their ancestors, no available presumptions whatever as to text can be obtained from number alone, that is, from number not as yet interpreted by descent.

56. The single exception to the truth of this statement leaves the principle itself untouched. Where a minority consists of one document or hardly more, there is a valid presumption against the reading thus attested, because any one scribe is liable to err, whereas the fortuitous concurrence of a plurality of scribes in the same error is in most cases improbable; and thus in these cases the reading attested by the majority is exempt from the suspicion of one mode of error which has to be taken into account with respect to the other reading. But this limited prima facie presumption, itself liable to be eventually set aside on evidence of various classes, is distinct in kind, not in degree only, from the imaginary presumption against a mere minority; and the essential difference is not altered by the proportion of the majority to the minority.

57. Except where some one particular corruption was so obvious and tempting that an unusual number of scribes might fall into it independently, a few documents are not, by reason of their mere paucity, appreciably less likely to be right than a multitude opposed to them. As soon as the numbers of a minority exceed what can be explained by accidental coincidence, so that their agreement in error, if it be error, can only be explained on genealogical grounds, we have thereby passed beyond purely numerical relations, and the necessity of examining the genealogy of both minority and majority has become apparent. A theoretical presumption indeed remains that a majority of extant documents is more likely to represent a majority of ancestral documents at each stage of transmission than vice versa. But the presumption is too minute to weigh against the smallest tangible evidence of other kinds. Experience verifies what might have been anticipated from the incalculable and fortuitous complexity of the causes here at work. At each stage of transmission the number of copies made from each MS depends on extraneous conditions, and varies irregularly from zero upwards: and when further the infinite variability of chances of preservation to a future age is taken into account, every ground for expecting a priori any sort of correspondence of numerical proportion between existing documents and their less numerous ancestors in any one age falls to the ground. This is true even in the absence of mixture; and mixture, as will be shown presently (§§ 61, 76), does but multiply the uncertainty. For all practical purposes the rival probabilities represented by relative number of attesting documents must be treated as incommensurable,

C. 58, 59. Manner of discovering genealogy

58. Knowledge of the Genealogy of Documents, as of other facts respecting them, can sometimes be obtained to a certain extent from external sources, under which may be included various external indications furnished by themselves; but it is chiefly gained by study of their texts in comparison with each other. The process depends on the principle that identity of reading implies identity of origin. Strictly speaking it implies either identity of origin or accidental coincidence, no third alternative being possible. Accidental coincidences do occur, and have to be reckoned for: but except where an alteration is very plausible and tempting, the chance that two transcribers have made the same alteration independently is relatively small, in the case of three it is much smaller, and so on with rapidly increasing improbability. Hence, while a certain number of identities of reading have to be neglected as capable of either interpretation, the great bulk may at once be taken as certain evidence of a common origin. Such community of origin for a reading may of course as regards the two or more attesting documents be either complete, that is, due to a common ancestry for their whole texts, or partial, that is, due to ‘mixture’, which is virtually the engrafting of occasional or partial community of ancestry upon predominantly independent descent.

59. Here, as in the investigation of the comparative excellences of continuous texts, we are able to arrive at general conclusions about texts by putting together the data furnished by a succession of variations of reading. What we have to do is to note what combinations of documents, large or small, are of frequent recurrence. Wherever we find a considerable number of variations, in which the two or more arrays of documents attesting the two or more variants are identical, we know that at least a considerable amount of the texts of the documents constituting each array must be descended from a common ancestor subsequent to the single universal original, the limitation of ancestry being fixed by the dissent of the other array or arrays. Each larger array may often in like manner be broken up into subordinate arrays, each of which separately is found repeatedly supporting a number of readings rejected by the other documents; and each such separate smaller array must have its own special ancestry. If the text is free from mixture, the larger arrays disclose the earlier divergences of transmission, the smaller arrays the later divergences: in other words, wherever transmission has been independent, the immediate relations of existing documents are exhibited by those variations which isolate the most subordinate combinations of documents, the relationships of the ultimate ancestors of existing documents by those variations in which the combinations of documents are the most comprehensive; not necessarily the most numerous individually, but the most composite.

D. 60—65. Complications of genealogy by mixture

60. In the texts just mentioned, in which transmission has followed exclusively the simple type of divergent ramification, cross divisions among documents are impossible, except to the limited extent within which accidental coincidence can operate. If L M are two transcripts of the original, L¹ L² of L, and M¹ M² of M, the five distributions (i) L¹ L² against M¹ M², (ii) L¹ against L² M¹ M², (iii) L² against L¹ M¹ M², (iv) M¹ against L¹ L² M², and (v) M² against L¹ L² M¹ are all possible and all likely to occur; but the two distributions (vi) L¹ M¹ against L² M² and (vii) L¹ M² against L² M¹ are impossible as results of divergent genealogy. In the second distribution L² appears to desert its own primary array and join the array of M; but the truth is that in a text transmitted under these conditions L¹ must have introduced a corruption, while L² has merely remained faithful to a reading of the original which had been faithfully preserved by L and M alike. On the other hand in the sixth distribution either L¹ M¹ must have the wrong reading and L² M² the right, or vice versa: if L¹ M¹ are wrong, either L and M must have both concurred in the error, which would have rendered it impossible for either L² or M² to be right, or L¹ and M¹, transcribed from different exemplars, must have each made the same change from the true reading of L and M preserved by L² and M², which is impossible except by accidental coincidence; and mutatis mutandis the case is the same if L¹ M¹ be right and L² M² wrong, and again for the two corresponding alternatives of the seventh distribution. In this fact that the sixth and seventh combinations, that is, cross combinations, cannot exist without mixture we have at once a sufficient criterion for the presence of mixture. Where we find cross combinations associated with variations so numerous and of such a character that accidental coincidence is manifestly incompetent to explain them, we know that they must be due to mixture, and it then becomes necessary to observe within what limits the effects of mixture are discernible.

61. In so far as mixture operates, it exactly inverts the results of the simpler form of transmission, its effect being to produce convergence instead of divergence. Corruptions originating in a MS belonging to one primary array may be adopted and incorporated in transcripts from other MSS of the same or of other primary arrays. An error introduced by the scribe of L¹ in one century, and unknown to L² M¹ M² may in a later century be attested by all the then extant representatives of L¹ L² M¹, those of M² alone being free from it, the reason being that, perhaps through the instrumentality of some popular text which has adopted it, it has found its way into intermediate descendants of L² and of M¹. It follows that, whenever mixture has intervened, we have no security that the more complex arrays of existing documents point to the more ancient ramifications: they may just as easily be results of a wide extension given comparatively late by favourable circumstances to readings which previously had only a narrow distribution. Conversely a present narrowness of distribution need not be a mark of relatively recent divergence: it may as easily (see § 76) be the only surviving relic of an ancient supremacy of distribution now almost obliterated by the invasion of mixture. This is of course a somewhat extreme case, but it is common enough: as a matter of fact, mixture is found to operate on every scale, from the smallest to the largest.

62. Mixture being thus liable to confuse and even invert the inferences which would indubitably follow from the conditions of transmission were transmission exclusively divergent, we have next to enquire what expedients can be employed when mixture has been ascertained to exist. Evidently no resource can be so helpful, where it can he attained, as the extrication of earlier unmixed texts or portions of texts from the general mass of texts now extant. The clearest evidence for tracing the antecedent factors of mixture in texts is afforded by readings which are themselves mixed or, as they are sometimes called, ‘conflate’, that is, not simple substitutions of the reading of one document for that of another, but combinations of the readings of both documents into a composite whole, sometimes by mere addition with or without a conjunction, sometimes with more or less of fusion. Where we find a variation with three variants, two of them simple alternatives to each other, and the third a combination of the other two, there is usually a strong presumption that the third is the latest and due to mixture, not the third the earliest and the other two due to two independent impulses of simplification. Peculiar contexts may no doubt sometimes give rise to this paradoxical double simplification: but as a rule internal evidence is decisive to the contrary. If now we note the groups of documents which support each of the three variants; and then, repeating the process with other conflate readings, find substantially the same groups of documents occupying analogous places in all cases, we gain first a verification of the presumption of mixture by the mutual corroboration of instances, and next a determination of one set of documents in which mixture certainly exists, and of two other sets of documents which still preserve some portion at least of two more ancient texts which were eventually mixed together. Sometimes the three groups are found nearly constant throughout, sometimes they have only a nucleus, so to speak, approximately constant, with a somewhat variable margin of other documents. This relative variability however, due to irregularity of mixture, does not weaken the force of the inferences to be drawn from each single instance. If a reading is conflate, every document supporting it is thereby shown to have a more or less mixed text among its ancestry; so that, in considering any other doubtful variation, we have empirical evidence that the contingency of mixture in each such document is not a priori unlikely. About those documents which habitually support the conflate readings we learn more, namely that mixture must have had a large share in producing their text. Similarly we learn to set an especial value on those documents which rarely or never support the conflate readings; not necessarily as witnesses to a true text, for in all these eases each true reading is paired with a simple wrong reading, but as witnesses to texts antecedent to mixture.

63. The results thus obtained supply the foundation for a further process. It is incredible that mixed texts should be mixed only where there are conflate readings. In an overwhelming proportion of cases the composition of two earlier readings would either be impossible or produce an intolerable result; and in all such cases, supposing the causes leading to mixture to be at work, the change due to mixture would consist in a simple replacement of one reading by another, such change being indifferently a substitution or an addition or an omission. Here then we should find not three variants, but two only: that is, the reading of the mixed text would be identical with one of the prior readings; and as a matter of course the documents attesting it would comprise both those that were descended from the mixed text and those that were descended from that earlier text which the mixed text has here followed. When accordingly we find variations exhibiting these phenomena, that is, having one variant supported by that set of documents which habitually attests one recurring factor of mixture in conflate readings, and another supported by all the remaining documents, there is a strong presumption that a large portion of the adverse array of documents is descended from no line of transmission independent of the remaining portion, (that is, independent of the set of documents which habitually attests the other factor of mixture in conflate readings,) but merely echoes at second hand the attestation of that remaining portion of the array: the lines of descent of the two groups which together make up the array are in short not parallel but successive. It follows that the documentary authority for the two variants respectively is virtually reduced to that of the two groups habitually preserving the separate factors of mixture.

64. It is true that variability in the margin of attestation, if we may for brevity repeat a phrase employed above (§ 62), may render it uncertain with which portion of the composite array certain documents should be classed, thus weakening but not destroying the force, whatever it may be, of their opposition to the reading of the single array. It is true also that the authority of the portion of documents which belongs to the mixed text does not become actually nothing: it is strictly the authority of a single lost document, one of the sources of the mixture, belonging to the same line of transmission as the earlier group of documents supporting the same reading independently of mixture, and thus adding another approximately similar member to their company. These qualifications do not however affect the substantial certainty and efficacy of the process here described, as enabling us in a large number of variations to disentangle the confusion wrought by mixture. It is independent of any external evidence as to dates, being founded solely on the analysis and comparison of the extant texts: but of course its value for purposes of criticism is much enhanced by any chronological evidence which may exist.

65. On the other hand there is much mixture of texts for which the extant documentary evidence antecedent to mixture is too small or uncertain to be detached from the rest, and therefore to yield materials for the application of this process. In such cases we have to fall back on the principle of Internal Evidence of Groups, to be explained presently, which is applicable to mixed and unmixed texts alike.

E. 66–72. Applications of genealogy

66. After this brief sketch of the modes of discovering genealogical facts by means of the extant texts, which will, we hope, be made clearer by the concrete examples to be given further on, we come to the uses of the facts so obtained for the discrimination of true from false readings. One case of the examples given in § 51 shews at once that any number of documents ascertained to be all exclusively descended from another extant document may be safely put out of sight, and with them of course all readings which have no other authority. The evidence for the fact of descent may be of various kinds. Sometimes, though rarely, it is external. Sometimes it consists in the repetition of physical defects manifestly not antecedent to the supposed original, as when the loss of one or more of its leaves has caused the absence of the corresponding portions of text in all the other documents. Sometimes the evidence is strictly internal, being furnished by analysis of the texts themselves, when it is found that a fair number of mere blunders or other evidently individual peculiarities of the supposed original have been either reproduced or patched up in all the supposed derivative documents, and secondly that these documents contain few or no variations from the text of the supposed original which cannot be accounted for by natural and known causes of corruption.

67. This summary reduction of documentary evidence by the discovery of extant ancestors of other existing documents is however of rare occurrence. On the other hand, wherever a text is found in a plurality of documents, there is a strong probability that some of them are descended from a single lost original. The proof of common descent is always essentially the same, consisting in numerous readings in which they agree among themselves and differ from all other documents, together with the easy deducibility, direct or indirect, of all their readings from a single text. In the absence of the second condition the result would differ only in being less simple: we should have to infer the mixture of two or more lost originals, independent of each other as well as of the remaining extant documents.

68. The manner of recovering the text of a single lost original, assuming the fact of exclusive descent from it to have been sufficiently established, will be best explained by a free use of symbols. Let us suppose that the extant descendants are fourteen, denoted as abcdefghiklmno; that, when their mutual relationships are examined, they are found to fall into two sets, abcdefghi and klmno, each having a single lost ancestor (X and Υ respectively) descended from the common original; and again that each of these sets falls similarly into smaller sets, the first into three, ab, cdef, and ghi, the second into two, kl and mno, each of the five lesser sets having a single lost ancestor (αβγδε respectively) descended from the common subordinate original, αβγ from X, δε from Y. Let us suppose also that no cross distributions implying mutual or internal mixture can be detected. We have then this pedigree:

O

X

Y

α

β

γ

δ

ε

a

b

c

d

e

f

g

h

i

k

l

m

n

o

69. Readings in which all fourteen documents agree belonged indubitably to the common original O. On the other hand the genealogical evidence now before us furnishes no indication as to the readings of Ο in variations in which all the descendants of X are opposed to all the descendants of Y: for reasons already given (§ 57) the proportion nine to five tells us nothing; and the greater compositeness of abcdefghi, as made up of three sets against two, is equally irrelevant, since we know that each larger set has but a single ancestor, and we have no reason for preferring X singly to Υ singly. These variations therefore we reserve for the present. Where however the descendants of either X or Υ are divided, so that the representatives of (say) γ join those of δ and ε against those of α and β, and the question arises whether the reading of X is truly represented by αβ or by γ, the decision must be given for that of γ, because, mixture and accidental coincidence apart, in no other way can γ have become at once separated from αβ and joined to δε; in other words, the change must have been not on the part of γ but of αβ, or rather an intermediate common ancestor of theirs. The reading thus ascertained to have been that of both X and Υ must also, as in the first case, have been the reading of O. Accordingly, so far as the whole evidence now before us is concerned, that is, assuming absence of mixture with documents independent of O, all readings of αβ against γδε may be at once discarded, first as departures from the text of O, and next as departures from the text of the autograph, since the direct transmission of all the documents passes through O, and thus it is not possible, on the present conditions, for αβ to agree with the autograph against Ο except by conjecture or accidental coincidence. The same results follow in all the analogous cases, namely for readings of y against αβδε, α against βγδε, δ against αβγε, and ε against αβγδ. The combinations αγ against βδε and βγ against αδε are possible only by mutual mixture among descendants of X antecedent to αβγ, since they form cross distributions with the assumed combination aβ against γδε: but this particular mixture would not interfere with the present operation of fixing the reading of X by coincidence with the reading of Y, because there would be no more mixture with Υ than in the other cases, and the force of the consent of Υ with part of the descendants of X remains the same whatever that part may be.

70. It will be seen at once what a wide and helpful suppression of readings that cannot be right is thus brought about by the mere application of Genealogical method, without need of appeal to the Internal Evidence of either Texts or Readings except so far as they contribute in the first instance to the establishment of the genealogical facts. Precisely analogous processes are required where any of the five lesser sets are divided, say by opposition of cd to ef, so that we have to decide whether the true reading of β is found in cd or in ef. The final clear result is that, when we have gone as far as the discoverable relations among our documents admit, we have on the one hand banished a considerable number of the extant variants as absolutely excluded, and on the other ascertained a considerable number of readings of O, in addition to those parts of the text of Ο in which all its descendants agree.

71. Two elements of uncertainty as to the text of Ο alone remain. First, the condition presupposed above, absence of mixture from without, does not always hold good. Where mixture from without exists, the inference given above from the concurrence of γ with δε against αβ becomes but one of three alternatives. It is possible that mixture with a text independent of Ο has affected γ and Υ alike, but not αβ; and if so, αβ will be the true representatives of X and of O. This possibility is however too slight to be weighed seriously, unless the reading of γ and Υ is found actually among existing documents independent of O, provided that they are fairly numerous and various in their texts, or unless the hypothesis of mixture is confirmed by a sufficiency of similarly attested readings which cannot be naturally derived from readings found among the descendants of O. Again, it is possible that the reading of αβ is itself due to mixture with a text independent of Ο: and if so, though rightly rejected from the determination of the reading of O, it may possibly be of use in determining the reading of an ancestor of O, or even of the autograph itself. But both these contingencies need be taken into account only when there is already ground for supposing mixture from without to exist.

72. The second element of uncertainty is that which always accompanies the earliest known divergence from a single original. Given only the readings of X and Y, Genealogy is by its very nature powerless to shew which were the readings of O. It regains its power only when we go on to take into account fresh documentary evidence independent of O, and work towards an older common original from which both it and Ο are descended. Ο then comes to occupy the place of X or Y, and the same process is repeated; and so on as often as the evidence will allow. It must however be reiterated (see § 52) that, when Ο has come to mean the autograph, we have, in reaching the earliest known divergence, arrived at the point where Genealogical method finally ceases to be applicable, since no independent documentary evidence remains to be taken up. Whatever variations survive at this ultimate divergence must still stand as undecided variations. Here therefore we are finally restricted to the Internal Evidence of single or grouped Documents and Readings, aided by any available external knowledge not dependent on Genealogy.

F. 73–76. Variable use of genealogy according to unequal preservation of documents

73. The proper method of Genealogy consists, it will be seen, in the more or less complete recovery of the texts of successive ancestors by analysis and comparison of the varying texts of their respective descendants, each ancestral text so recovered being in its turn used, in conjunction with other similar texts, for the recovery of the text of a yet earlier common ancestor. The preservation of a comparatively small number of documents would probably suffice for the complete restoration of an autograph text (the determination of the earliest variations of course excepted) by genealogy alone, without the need of other kinds of evidence, provided that the documents preserved were adequately representative of different ages and different lines of transmission. This condition however is never fulfilled. Texts are not uncommonly preserved in a considerable assemblage of documents the genealogy of which can be fully worked out, but is found to conduct to one or two originals which, for all that appears to the contrary, may be separated from the autograph by many ages of transmission, involving proportionate possibilities of corruption. Here Genealogical method retains its relative value, for it reduces within narrow limits the amount of variation which need occupy an editor when he comes to the construction of his text: but it leaves him in the dark, as all criticism dealing only with transmitted variations must do, as to the amount of correspondence between the best transmitted text and the text of his author. These cases correspond to such limited parts of the documentary evidence of more adequately attested texts as represent single stages of textual history.

74. In those rare cases, on the other hand, in which extant documentary evidence reaches up into quite ancient times the process may be carried back to a stage comparatively near the autograph: but here the evidence is as a matter of fact never abundant enough for more than rough and partial approximations to the typical process described above. Here too, as always, we have to ascertain whether the confusing influence of mixture exists, and if so, within what limits. Under such circumstances any chronological and geographical information to be obtained from without has great value in interpreting obscure genealogical phenomena, especially as marking the relative date and relative independence of the several early documents or early lost ancestors of late documents or sets of documents.

75. In proportion as we approach the time of the autograph, the weight of composite attestation as against homogeneous attestation increases; partly because the plurality of proximate originals usually implied in composite attestation carries with it the favourable presumption afforded by the improbability of a plurality of scribes arriving independently at the same alteration; partly because the more truly composite the attestation, that is, the more independent its component elements, the more divergences and stages of transmission must have preceded, and thus the earlier is likely to have been the date for the common original of these various generations of descendants, the later of which are themselves early. Nothing of course can exclude the possibility that one line of transmission may have ramified more rapidly and widely than another in the same time: yet still the shorter the interval between the time of the autograph and the end of the period of transmission in question, the stronger will be the presumption that earlier date implies greater purity of text. But the surest ground of trusting composite attestation is attained when it combines the best documentary representatives of those lines of transmission which, as far as our knowledge goes, were the earliest to diverge. Such are essentially instances of ascertained concordance of X and Υ (§ 69), in spite of the dissent of some descendants of one or both.

76. The limitation to "the best documentary representatives" is necessary, because the intrusion of mixture in documents, or in lost originals of documents or of documentary groups, may disguise the actual historical relations (see § 61), and give the appearance of greater compositeness of attestation to readings which have merely invaded lines of transmission that for a while were free from them. It thus becomes specially necessary to observe which documents, or lost originals of documents or documentary groups, are found to shew frequent or occasional mixture with texts alien from their own primary ancestry, and to allow for the contingency accordingly. Many cases however of ambiguous interpretation of evidence are sure to remain, which the existing knowledge of the history of mixture is incompetent to clear up; and for these recourse must be had to evidence of other kinds.

 

SECTION IV. INTERNAL EVIDENCE OF GROUPS

77, 78

77. We have reserved for this place the notice of another critical resource which is in some sense intermediate between Internal Evidence of Documents and Genealogical Evidence, but which in order of discovery would naturally come last, and the value of which will have been made more apparent through the inherent and the incidental defects of Genealogical Evidence described in the preceding paragraphs. This supplementary resource is Internal Evidence of Groups. In discussing Internal Evidence of Documents, we spoke only of single documents: but the method itself is equally applicable to groups of documents. Just as we can generalise the characteristics of any given MS by noting successively what readings it supports and rejects, (each reading having previously been the subject of a tentative estimate of Internal Evidence of Readings, Intrinsic and Transcriptional,) and by classifying the results, so we can generalise the characteristics of any given group of documents by similar observations on the readings which it supports and rejects, giving special attention to those readings in which it stands absolutely or virtually alone. In texts where mixture has been various, the number of variations affording trustworthy materials for generalisations as to any one group can be only a part of the sum total of variations; but that part will often be amply sufficient. The evidence obtained in this manner is Internal Evidence, not Genealogical. But the validity of the inferences depends on the genealogical principle that community of reading implies community of origin. If we find, for instance, in any group of documents a succession of readings exhibiting an exceptional purity of text, that is, readings which the fullest consideration of Internal Evidence pronounces to be right in opposition to formidable arrays of Documentary Evidence, the cause must be that, as far at least as these readings are concerned, some one exceptionally pure MS was the common ancestor of all the members of the group; and that accordingly a recurrence of this consent marks a recurrence of joint derivation from that particular origin, and accordingly a strong presumption that exceptional purity is to be looked for here again. The inference holds equally good whether the transmission has been wholly divergent, or partly divergent and partly mixed; and any characteristic, favourable or unfavourable, may be the subject of it.

78. The value of Internal Evidence of Groups in cases of mixture depends, it will be seen, on the fact that by its very nature it enables us to deal separately with the different elements of a document of mixed ancestry. In drawing general conclusions from the characteristics of the text of a document for the appreciation of its individual readings successively, we assume the general homogeneousness of its text; but this assumption is legitimate only if unity of line of ancestry is presupposed. The addition of a second line of ancestry by mixture introduces a second homogeneousness, which is as likely as not to conflict with that of the first, and thus to falsify inferences drawn from the first, unless there be means of discriminating from the rest of the text the portions taken from the second original. But each well marked group of which the mixed document is a member implies at least the contingency of a distinct origin; and thus, in readings in which the document is associated with the rest of the group, its authority need not be that which it derives in the bulk of its text from its fundamental or primary original, but is strictly that belonging to the common ancestor of its secondary original and of the other members of the group. Such readings might be truly described as forming a series of minute fragments of a copy of the lost document which was the secondary original, leaving corresponding gaps in the more or less faithfully preserved text of the primary original, except where conflate readings have wholly or partly preserved both texts. In the next Part we shall have ample opportunity of illustrating what has here been said.

SECTION V. RECAPITULATION OF METHODS IN RELATION TO EACH OTHER

79—84

79. To recapitulate. The method of Genealogy is an application of one part of the knowledge of Documents; and like the method founded on the Internal Evidence of Documents it involves three processes; first the analysis and comparison of the documentary evidence for a succession of individual variations; next the investigation of the genealogical relations between the documents, and therefore between their ancestors, by means of the materials first obtained; and thirdly the application of these genealogical relations to the interpretation of the documentary evidence for each individual variation. The results of the interpretation of documentary evidence thus and thus alone made possible are various. In the first place it winnows away a multitude of readings which genealogical relations prove to be of late origin, and which therefore cannot have been derived by transmission from the autograph. Where the extant evidence suggests but is insufficient to prove thus much, and in the case of all other variants, this method so presents and limits the possible genealogical antecedents of the existing combinations of documentary evidence as to supply presumptions in favour of one variant against another varying from what amounts under favourable circumstances to practically absolute certainty down to complete equipoise.

80. So far as genealogical relations are discovered with perfect certainty, the textual results which follow from them are perfectly certain too, being directly involved in historical facts; and any apparent presumptions against them suggested by other methods are mere guesses against knowledge. But the inequalities and occasional ambiguities in the evidence for the genealogical relations frequently admit of more than one interpretation, and this greater or less substitution of probability for certainty respecting the documentary history reduces the textual verdict to a presumption, stronger or weaker as the case may be. Genealogical presumptions ought however to take precedence of other presumptions, partly because their immediate basis is in itself historical not speculative, and the subject-matter of all textual criticism is historical, partly because the generalisations by which that historical basis is ascertained involve less chance of error than the analogous generalisations required for any kind of Internal Evidence.

81. The only safe order of procedure therefore is to start with the reading suggested by a strong genealogical presumption, if such there be; and then enquire whether the considerations suggested by other kinds of evidence agree with it, and if not, whether they are clear and strong enough to affect the prima facie claim of higher attestation. If they appear so to be, a full re-examination becomes necessary; and the result, especially if similar instances recur, may be the discovery of some genealogical complication overlooked before. No definite rule can be given as to what should be done where the apparent conflict remains, more especially where the documentary evidence is scanty or obscure. For our own part, in any writing having fairly good and various documentary attestation we should think it dangerous to reject any reading clearly supported by genealogical relations, though we might sometimes feel it equally necessary to abstain from rejecting its rival.

82. Next in value to Genealogical Evidence is Internal Evidence of Documents, single or in groups. But where the documents exceed a very small number, the Internal Evidence of single Documents, as has already been explained (§ 48), is rendered for the most part practically inapplicable by the unresolved complexity. The Internal Evidence however of Groups of Documents is always applicable if there are documents enough to form groups. It is the best substitute for Genealogical Evidence proper in texts, or in any parts of texts, in which genealogical relations are too obscure for use; and it affords the most trustworthy presumptions for comparison with purely genealogical presumptions, having similar merits derived from the form of the processes by which it is obtained, while relating to a different class of phenomena. The highest certainty is that which arises from concordance of the presumptions suggested by all methods, and it is always prudent to try every variation by both kinds of Internal Evidence of Readings. The uncertainty however inherent in both, as dependent on isolated acts of individual judgement, renders them on the whole untrustworthy against a concurrence of Genealogy and Internal Evidence of Documents; though a concurrence of clear Intrinsic with clear Transcriptional Probability ought certainly to raise at least a provisional doubt.

83. Textual criticism fulfils its task best, that is, is most likely to succeed ultimately in distinguishing true readings from false, when it is guided by a full and clear perception of all the classes of phenomena which directly or indirectly supply any kind of evidence, and when it regulates itself by such definite methods as the several classes of phenomena suggest when patiently and circumspectly studied. This conformity to rationally framed or rather discovered rules implies no disparagement of scholarship and insight, for the employment of which there is indeed full scope in various parts of the necessary processes. It does but impose salutary restraints on the arbitrary and impulsive caprice which has marred the criticism of some of those whose scholarship and insight have deservedly been held in the highest honour.

84. Nevertheless in almost all texts variations occur where personal judgement inevitably takes a large part in the final decision. In these cases there is no failure of method, which strictly speaking is an impossibility, but an imperfection or confusion of the evidence needed for the application of method. Here different minds will be impressed by different parts of the evidence as clearer than the rest, and so virtually ruling the rest: here therefore personal discernment would seem the surest ground for confidence. Yet here too, once more, the true supremacy of method is vindicated; for it is from the past exercise of method that personal discernment receives the education which tends to extinguish its illusions and mature its power. All instinctive processes of criticism which deserve confidence are rooted in experience, and that an experience which has undergone perpetual correction and recorrection.

SECTION VI. CRITICISM AS DEALING WITH ERRORS ANTECEDENT TO EXISTING TEXTS

85—95

A. 85—92. Primitive errors

85. The preceding pages have dealt exclusively with the task of discriminating between existing various readings, one variant in each case being adopted and the rest discarded. The utmost result that can be obtained under this condition is the discovery of what is relatively original: whether the readings thus relatively original were also the readings of the autograph is another question, which can never be answered in the affirmative with absolute decision except where the autograph itself is extant, but which admits of approximative answers varying enormously in certainty according to the nature of the documentary evidence for the text generally. Even in a case in which it were possible to shew that the extant documents can be traced back to two originals which diverged from the autograph itself without any intermediate common ancestor, we could never be quite sure that where they differed one or other must have the true reading, since they might independently introduce different changes in the same place, say owing to some obscurity in the writing of a particular word. In almost all actual cases an interval, short or long, must have divided the autograph from the earliest point or points to which genealogy conducts us back; and any interval implies the possibility of corruption, while every addition to the length of the interval increases the probability of corruption. On the other hand documentary evidence including a fair variety of very ancient attestation may bring the meeting-point of the extant lines of transmission so near the autograph that freedom from antecedent corruption ceases to be improbable, without however thereby becoming a priori probable. In such cases therefore any investigation of the ultimate integrity of the text is governed by no theoretical presumptions: its final conclusions must rest on the intrinsic verisimilitude or suspiciousness of the text itself.

86. These considerations have an important bearing on certain paradoxical conflicts of evidence respecting transmitted variations, which present themselves occasionally in most texts and frequently in many; and which are peculiarly apt to mislead editors to whom textual criticism is only a subordinate province of interpretation. The reading clearly indicated by Genealogical or other evidence obtained from whole texts, or by Transcriptional Evidence of Readings, or by both together, may be as clearly condemned by Intrinsic Evidence. We are not speaking of the numerous cases in which readings that have seemed to a critic in the first instance too strange to be true approve themselves on better knowledge, perhaps as no more than tolerable, but oftener still as having a peculiar impress of truth which once apprehended cannot easily be questioned; or in which competent critics receive opposite impressions from the same reading, one holding it to be impossible, the other to have the stamp of originality. These differences of judgement throw no light upon readings which all competent critics feel on consideration to be impossible, and yet which are strongly attested by, it may be, every kind of evidence except Intrinsic Evidence.

87. The true solution lies in the fact that the subject matter of the different kinds of evidence is not identical. Intrinsic Evidence is concerned only with absolute originality; it pronounces which of two or more words or phrases a given author in a given place was more likely to use, or, in extreme cases in either direction, whether either of them was what he must have used or could not possibly have used. All other kinds of evidence are concerned only or predominantly with relative originality: they pronounce, speaking roughly, which of two or more readings is more likely to have given rise to the others, or is found in the best company, or has the best pedigree. The apparent conflict therefore is dependent on the assumption, usually well founded, that the two originalities coincide. Where they do not, that is, where corruption has preceded the earliest extant documentary evidence, the most nearly original extant reading may nevertheless be wrong, simply because the reading of the autograph has perished. What an editor ought to print in such a case, supposing he has satisfied himself that the best attested reading is really impossible, may vary according to circumstances. But it is clearly his duty in some way to notify the presumed fact of corruption, whether he can offer any suggestion for its removal or not.

88. In the cases just mentioned, while the best attested reading is found to be impossible, the other reading or readings shown by evidence not Intrinsic to be corruptions of it are or may be found quite possible, but not more: they derive their prima facie probability only from an assumed necessity of rejecting their better attested rival. In other cases the reading (or one of the readings) shown to be of later origin has very strong Intrinsic Evidence in its own favour; that is, we have a combination of positive clear Intrinsic Evidence for the worse attested reading with negative clear Intrinsic Evidence against the better attested reading. So complete an inversion of the ordinary and natural distributions of evidence always demands, it need hardly be said, a thorough verification before it can be accepted as certain. It does however without doubt occasionally occur, and it arises from a state of things fundamentally the same as in the former cases, with the difference that here a transcriber has happened to make that alteration which was needed to bring back the reading of the autograph, that is, has in the course of transcription made a successful Conjectural Emendation. No sharp line can in fact be drawn between the deliberate conjectural emendations of a modern scholar and many of the half or wholly unconscious changes more or less due to mental action which have arisen in the ordinary course of transcription, more especially at times when minute textual accuracy has not been specially cultivated. An overwhelming proportion of the cursory emendations thus made and silently embodied in transcribed texts are of course wrong: but it is no wonder that under favourable circumstances they should sometimes be right. It may, once more, be a matter of doubt what form of printed text it will here be most expedient under given circumstances to adopt. The essential fact remains under all circumstances, that the conjectural origin of these readings is not altered by the necessity of formally including them in the sum of attested readings; and that an editor is bound to indicate in some manner the conjectural character of any attested reading which he accepts as the reading intended by the author, and yet which he does not believe to have been received by continuous transmission from the autograph.

89. We have dwelt at some length on these two classes of variations because at first sight they appear to furnish grounds for distrusting the supremacy of what we have ventured to call the higher kinds of evidence. They not unnaturally suggest the thought that, whatever may be said in theory respecting the trustworthiness of evidence not Intrinsic, it breaks down in extreme cases, and must therefore contain some latent flaw which weakens its force in all. But the suspicion loses all plausibility when it is seen that it springs from a confusion as to the subject matter of attestation (see § 87), and that the attestation itself remains as secure in extreme cases as in all others. The actual uncertainties arise not from any want of cogency of method, but from inadequate quantity or quality of the concrete evidence available in this or that particular text or variation.

90. Both the classes of variations just considered imply corruption in the earliest transmitted text. The same fact of corruption antecedent to extant documentary evidence has to be recognised in other cases, some of which form a third class of variations. Besides the variations already noticed in which the evidence shews one variant to have been the parent of the rest, while yet on Intrinsic grounds it cannot be right, there are others in which the variants have every appearance of being independent of each other, while yet on Intrinsic grounds none having sufficiently good documentary attestation, or even none at all, can be regarded as right: that is to say, a convergence of phenomena points to some lost reading as the common origin of the existing readings. Fourthly, there may be sufficient grounds for inability to accept the transmitted text even in places where the documents agree.

91. In all four cases the ground of belief that the transmitted text is wrong is Internal Evidence of Readings. In the third it is or may be a combination of Intrinsic and Transcriptional Evidence: in the first, second, and fourth it is exclusively Intrinsic Evidence, except where recognition of corruption is partly founded on perception of the lost original reading, which, as we shall see shortly, involves the use of Transcriptional Evidence. The use of Internal Evidence of Readings in detecting corruption is precisely identical with its use, or rather one of its uses, in the discrimination of attested readings. In coming to a decision on the strength of Intrinsic Evidence, a critic makes one of three affirmations respecting two variants α and β; (1) α is more probable than β; (2) α is not only more probable than β, and is not only suitable to the place, but is so exactly and perfectly suitable that it must be right; and (3) β is not only less probable than α, but so improbable absolutely that it cannot be right, so that α as the only remaining variant must be right: (2) and (3) of course include (1), and also are compatible with each other. Now in pronouncing a text corrupt, he affirms neither more nor less than in the fundamental proposition of the third instance, in which he equally finds his whole evidence exclusively in the reading condemned, and in its own relations to the context, without reference to any other variant. In both procedures the affirmation has against it all the uncertainties which we have pointed out as inherent in the exclusive use of Intrinsic Evidence: nevertheless there are places in nearly all texts where its force is so convincing that the most cautious critic cannot refuse to make the affirmation, and in every ill preserved text they abound.

92. The first, second, and fourth cases are essentially the same. The presence of more than one variant in the first and second case does not place them on a different footing from the fourth, because all but the one are by supposition subsequent to the one, and are therefore virtually out of sight when the question of accepting the most original of attested readings as the true reading arises. A critic may doubtless feel less reluctant to pronounce a reading corrupt when he sees that it gave trouble to ancient scribes; but the encouragement is due to corroboration of personal judgement, not to any kind of evidence; it comes from the ancient scribes in the character of critics, not as witnesses to a transmitted text. On the other hand the third case has an advantage over the others by combining a certain measure of Transcriptional with Intrinsic Probability. The supposition of corruption has the strength of a double foundation when it not only accounts for our finding an impossible text but supplies a common cause for two readings, the apparent independence of which would otherwise be perplexing; and this it does even in the absence of any perception as to what conjectural reading would fulfil the various conditions of the case.

B. 93—95. Removal of primitive errors by conjecture

93. In discussing the corruption of texts antecedent to extant documents, the forms in which it presents itself, and the nature of the critical process by which it is affirmed, we have reserved till last a brief notice of the critical process which endeavours to remedy it, that is, Conjectural Emendation. Although in practice the two processes are often united, and a felicitous conjecture sometimes contributes strong accessory evidence of corruption, it is not the less desirable that they should be considered separately. The evidence for corruption is often irresistible, imposing on an editor the duty of indicating the presumed unsoundness of the text, although he may be wholly unable to propose any endurable way of correcting it, or have to offer only suggestions in which he cannot place full confidence.

94. The art of Conjectural Emendation depends for its success so much on personal endowments, fertility of resource in the first instance, and even more an appreciation of language too delicate to acquiesce in merely plausible corrections, that it is easy to forget its true character as a critical operation founded on knowledge and method. Like the process of detecting corruption, it can make no use of any evidence except Internal Evidence of Readings, but it depends on Intrinsic and Transcriptional Evidence alike. Where either there is no variation or one variant is the original of the rest, that is, in the fourth, first, and second of the cases mentioned above, two conditions have to be fulfilled by a successful emendation. As regards Intrinsic Evidence, it must, to attain complete certainty, be worthy of the second form of affirmation noticed above, that is, be so exactly and perfectly suitable to the place that it cannot but be right; or, to attain reasonable probability, it must be quite suitable to the place positively, and free from all incongruity negatively. As regards Transcriptional Evidence, it must be capable of explaining how the transmitted text could naturally arise out of it in accordance with the ordinary probabilities of transcription. Where there are more independent variants than one, that is, in the third case, the only difference is that the suggested correction must in like manner be capable of giving rise naturally to every such transmitted Reading. Thus in all cases the problem involved in forming a judgement on a suggested Conjectural Emendation differs in one respect only from the ordinary problems involved in deciding between transmitted readings on the strength of Intrinsic and Transcriptional Evidence combined, and of these alone; it consists in asking whether a given reading out of two or three fulfils certain conditions well absolutely, whereas in other cases we ask which of two or three readings fulfils the same conditions best.

95. The place of Conjectural Emendation in the textual criticism of the New Testament is however so inconsiderable that we should have hesitated to say even thus much about it, did it not throw considerable light on the true nature of all textual criticism, and illustrate the vast increase of certainty which is gained when we are able to make full use of Documentary Evidence, and thus confine Internal Evidence to the subordinate functions which alone it is normally fitted to discharge.