Chapter II

The Methods of Criticism



IT has been already pointed out that there are but two kinds of evidence to which we can appeal in prosecuting the work of criticising a text,—external and internal evidence. All methods of criticism are, therefore, but various ways of using these kinds of evidence; and when we undertake to investigate the methods of criticism, we simply inquire how we are to proceed in order to reach firm conclusions as to the text by means of internal and external evidence. We have been busied thus far in merely gathering the external testimony, and the reader is doubtless in a position to appreciate how little the mere collection of the testimony has advanced us in deciding on the text. It is our business now to consider how we may attain a grounded decision as to the true text.

1. Internal Evidence Of Readings.

The most rudimentary method of dealing with the variations that emerge in the collection of the external testimony would be to use the external evidence only to advertise to us the fact of variation and to furnish us with the readings between which choice is to le made, and then to settle the claims of the rival readings on internal grounds. Most crudely performed, this would be to select, out of the readings actually transmitted, that one which seemed to us to make the best sense in the connection, or to account most easily for the origin of the others. It requires no argument to point out the illegitimacy of thus setting aside the external evidence unheard; or the danger of thus staking everything upon our insight into the exact intention of the author or the springs of action that moved men through a millennium and a half of copying, if this insight be exercised extemporaneously, as it were, and without a very severe previous study of the authors and their times and the scribes and their habits. Nevertheless, though all may not be lightly ventured upon its untrained dictum, internal evidence of readings, when carefully investigated, constitutes a most valuable method of criticism, the aid of which we cannot dispense with. It will repay us, therefore, to consider its methods of procedure in some detail.

As has been already intimated, "internal evidence of readings" includes two separate and independent processes. In interrogating any reading as to the evidence that it bears to its own originality, we may make our inquiries with reference to the author, or with reference to the scribes who have transmitted what he wrote j and we may make them in either case absolutely, or relatively to other transmitted readings. We may ask, absolutely, What is the probability that this is the reading that the author would have placed just here? or, relatively, What probability commends this reading, above any of the others that have come down to us, as the reading which the author wrote here? Or we may ask what is the probability that this is the reading which the scribes began with, either absolutely—i.e., in the form, Does this reading suggest an earlier one, out of which it was made by the scribes? or relatively to the other transmitted readings—that is, in the form, What is the probability that the other readings have grown out of this one? When dealing absolutely with each reading, we are seeking directly the autographic text. When dealing relatively with each, we are seeking in the first instance only the earliest transmitted text, and leaving it to a further inquiry to determine whether or not this is the autographic text. In either case we are making use of two separate methods of inquiry; one of which deals with the probability that the author wrote this reading, and the other with the probability that the scribes began with it. The one is appropriately called Intrinsic Evidence, and the other Transcriptional Evidence.

Intrinsic Evidence.

By intrinsic evidence is meant the testimony which each reading delivers, by its very nature, to its fitness to stand in the text. It is elicited by actually trying the reading in question in the passage and testing its appropriateness by the contextual argument, the rhetorical flow of the language, the known style and habits of speech and thought of the author, and the general language and thought-circle of the times and society in which he lived. The danger that attends the use of the method grows out of our tendency to read our own standpoint into our author, instead of reading ourselves back into his. It is easy to become an improver instead of remaining a jimple editor; and it is often very difficult not to make an author speak our thoughts, if not even our language. It cannot, however, be too strongly insisted upon that any attempt to estimate intrinsic probabilities by the rule of what appears to us to be the best reading is simply an attempt to corrupt the text and train it to festoon the trellises of our own desires. All trustworthy appeal to intrinsic evidence is a delicate historical process by which the critic, having steeped himself in the times of the writer and having assimilated himself to his thought and style, thinks his thoughts and estimates the value and fitness of words with his scales. The reading which would be intrinsically certain in Mr. Carlyle might be intrinsically ridiculous in Mr. Ruskin. The reading that we should commend in Lucian might be unthinkable in Epictetus; that which would be appropriate in Lucretius might be impossible in John. The preparation for a just use of this method of criticism consists, therefore, in a serious and sympathetic study of the author in hand; and without this, all appeal to it is but opening the floodgates to the most abounding error.

Above all other processes of criticism this method requires in its user a fine candour and an incorruptible mental honesty which are content to read from the authors with which they deal only what those authors have put into their words, and which can distinguish between what Paul, for instance, says, and what we could wish he had said. Despite what we may have antecedently thought, some writers are ungrammatical, some are obscure, some are illogical, some are inconsequent, some are frightfully infelicitous. And the business of the textual critic is not to correct their grammar, and brighten their obscurities, and perfect their logic, and chasten their style, but to restore their text exactly as they intended to write it, whatever there may be in it to offend our taste or contradict our opinions. Intrinsic evidence in the hands of some critics means nothing else than a ruthless elimination of everything exceptional or even distinctive in an author's style. When Mr. Margoliouth lays it down as a canon for criticising the Attic tragedians that "anything which is difficult or awkward is corrupt," we more than doubt the validity of his methods; and when Mr.McClellan, dealing with the New Testament, states as the "golden canon," that "no reading can possibly be original which contradicts the context of the passage or the tenor of the writing," we recognise the justice of the statement, but desiderate some safeguard that the test shall be applied from the point of sight of the author, and not of the nineteenth-century reader, in whose logical infallibility there may be less reason to believe than in that of the writer who is criticised. Delicate as the process of intrinsic evidence thus becomes, however, it is yet not only a valuable but also an indispensable agent of criticism, and its verdicts sometimes reach a practical certainty. Whenever it is the expression of careful and sympathetic study of an author's thought and style it demands our serious attention, and if, when so used, it distinctly and directly opposes a reading, it may attain a real finality. Cases of this kind, where intrinsic evidence sets itself immovably against a reading, must be very sharply distinguished from those in which it only adjudges one of several readings to be on the whole preferable to the others. In the former case its verdict has an absoluteness which is wholly lacking to the merely relative result reached in the latter. If the other readings, in this case, any or all of them, would have seemed unexceptionable in the absence of the preferred reading, the preference thrown upon this by intrinsic evidence can carry us but a little way towards settling the text, and raises but a faint presumption against any other form of evidence.

The variation in Matt. vi. 1 may perhaps serve as an illustration of the force of intrinsic evidence when thus simply passing on the comparative appropriateness of two readings. The Authorised English Version reads, "Do not your alms before men," which the Revisers change to " Do not your righteousness before men." Which does intrinsic evidence commend? Unquestionably the latter. Throughout this context our Lord is giving instruction concerning righteousness; and having commanded His disciples in the previous chapter (v. 20, sq.) to see to it that their righteousness exceeded that of the scribes and Pharisees, and illustrated the command by instancing the laws against murder, adultery, false swearing, and the like, he proceeds now (vi. 1) to guard against an ostentatious righteousness, and, just as before, illua

trates His command by instancing certain details,— here, almsgiving (2—4), prayer (3—15), and fasting (16—18). To read "righteousness" here is thus far more consonant with the context, and even brings out a connection with the preceding part of the discourse which with the reading " alms" is in danger of being overlooked. "Righteousness," moreover, comes with a Hebraistic flavour straight from the Old Testament, both in the structure of the phrase, "to do righteousness," and in its use as a genus of which "alms" is a species, and thus is especially suitable in the Hebraistic Matthew. We cannot fail to feel that such considerations create a very substantial corroboration of the testimony of those MSS. which contain "righteousness" here. Nevertheless, if " alms" were strongly pressed upon us by external evidence, this intrinsic evidence would not avail to set it aside. For although intrinsic evidence decidedly prefers "righteousness" here, it does not distinctly refuse " alms"; apart from the other reading "alms" would be easily accepted by it, and, hence, if it is otherwise strongly supported, we can receive it as the original reading. Another example of like character is furnished by Luke xv. 21, where the variation concerns the insertion or omission of the repetition from verse 19 of the words "Make me as one of thy hired servants." Intrinsic evidence casts its vote for omission. That the son does not carry out his intention of asking to be made a servant after his father had hasted to claim him as a wellbeloved son, is a fine trait; and we hesitate to believe that such true psychology, and such a beautiful turn

nf composition, have entered the narrative only by a slip from the bungling hand of a sleepy scribe. But after all, may it not have done so? If no copy had omitted the words, we should scarcely have thought of doing so; and hence, even here, intrinsic evidence raises a probability only and does not attain certainty. In a word, intrinsic considerations, in all such cases, give evidence, and oft-times very strong evidence, but scarcely such decisive evidence as can withstand the pressure of a strong probability brought from another quarter.

The evidence is more decisive in such a case as that «f Acts xii. 25, where to read that Paul and Barnabas returned "to Jerusalem," seems flat in the face of the context, although some relief may be got from an unnatural construction. It is to be observed, howiver, that even this result is negative, and in rejecting Eis 'lepovo-aAijjii here, intrinsic evidence does not necessarily commend thereby either of its rivals or O.tto: it contents itself with simply refusing the reading offered to it. This may be illustrated further by the variation at Acts xi. 20. Intrinsic evidence utterly refuses to have anything here except a reading that gives the sense of eAAjjvas; but again this is negative, and does not amount to a demand for just this word. All that we learn from it is that the author of the book placed here some word which contrasted with the "Jews" of v. 19, and which recorded an advance on the previous practice of the Church, and prepared for distinguishing the Christians from the Jews (xi. 26), and for sending missions to the Gentiles (xiii.) It tells us with great positiveness, therefore, that Greek-speaking Jews were not meant here, but veritable Gentiles. It is perhaps a mistake to spring too rashly to the conclusion, however, that this is equivalent to commending eAAijvas and rejecting eAA^no-rds; some other matters need settling first. But if eAAijnoras necessarily means " Greek-speaking Jews," then this evidence does decisively reject it. And if eAAijvas be otherwise well commended, intrinsic evidence accepts it gladly as furnishing just the thought it desires.

These examples illustrate the nature and the limitations of this method of criticism. It cannot be used idly, and it is very easy to abuse. But when exercised with care, and guided by a sympathetic insight into the literary character of the author under treatment, it is capable of much, and indispensable to the critic. It is chary of giving a positive verdict with too great decision; but it may be safely asserted that no conclusion to which it does not give at least its consent can be accepted as final in any case of textual criticism.

Transcriptional Evidence.

By transcriptional evidence is meant the testimony which each reading bears to its own origination. It is elicited by comparing together the whole series of claimants to a place in the text, in any given passage, with a view to discovering in what order they must have arisen—that is, which one of them, on the assumption of its originality, will best account for the origin of all the rest, or to what reading the whole body of extant readings points, as their source and fountain. The danger to which this method is exposed resides in our liability to come to conclusions on the ground of tendencies to error which we may observe in ourselves, rather than on the ground of the actual tendencies that led astray the scribes who have transmitted ancient books to us. Our only safeguard against this danger is to make preparation for using this method by a thorough study of the character of scribes' work, and of the errors to which they were liable as exhibited in the actual errors which they have made. A few hours of careful scrutiny of a series of acknowledged errors actually occurring in our codices will do more towards fitting us for the exercise of this nice process than any length of time spent in d, priori reasoning. Above all, it must be remembered that in criticising—say, for instance, the text of the New Testament—we are dealing with a writing which has had not one but many scribes successively engaged upon it, and that, therefore, we are to deal with a complex of tendencies which may have been engaged in progressively corrupting a text, and that in even exactly opposite directions. The greatest difficulty of the process is found in experience to reside less, however, in inability to arrange any given series of readings in an order which may well have been, on known tendencies of scribes, the order of their origination, than in inability to decide which of several orders, in which they seem equally capable of being arranged, is the actual order of their origination. Just because the tendencies to error ran through a very wide range and pulled in divergent directions, it often seems equally easy to account for each rival reading as a corruption of some other; and the acute editor is seldom at a loss to defend the reading which he prefers, by pointing out some way in which the rival readings may have grown out of it. The only remedy against this ever-present danger is a more careful study of the MSS. themselves, and a more rigid exclusion of all undue subjectivity from our judgments. What is difficult is not impossible; and, as experience grows, it is usually discovered that we can with ever-increasing confidence select from a body of readings the one which actually did stand at the root of all the others. Wherever this can be done, transcriptional evidence may be able to deliver a very decided verdict.

A circumstance which appears, at first sight, sufficiently odd, operates to give us especial confidence in the union of transcriptional and intrinsic evidence in the same finding. Just because intrinsic evidence asks after the best reading and transcriptional evidence after the reading that has been altered by the scribes, they are frequently found, at first examination, in apparent conflict. An obviously satisfactory reading is not especially apt to be changed by a scribe; it is often the play of his mind about a reading that puzzles him in one way or another, that distracts his attention from or intrudes his conjecture into his writing. When we ask which is the best reading, therefore, we often select the one which appeared also to the scribe to be the best, and which, when we ask after the original reading, just on this account appears to be a scribe's correction of a less obviously good or easy reading. Rarely, this contradiction between the two forms of internal evidence is ineradicable. Commonly, however, it is only the signal to us that we have carelessly performed our work in the one process or the other, and thus directs us to a further study, and finally to a complete reconciliation of the divergent findings. The reading that seemed to us intrinsically unlikely comes often on deeper study to seem intrinsically certain; or else the reading which seemed at first certainly derivative, comes to be seen to be without doubt original. Whenever these two so easily opposing forms of evidence can be shown to unite heartily and certainly in favour of one reading, they raise a presumption for it that will not yield to any other kind of evidence whatever. But, for precisely the same reason, whenever they seem hopelessly set in opposition to one another, we may with the greatest justice suspect the conclusions at which we have arrived by the one or the other,—perhaps by both.

The very essence of a preparation to engage in criticism by the aid of transcriptioual evidence is experience of actual scribes' work. Nothing can quite take the place of familiarity with MSS. themselves. Where this is impossible, facsimiles may form a partial substitute; and even the information given in the digests may be turned to excellent account by the diligent student. Some primary hints of how various readings have arisen in the text, which may serve as a basis for further and more direct studies, are all that it is possible to set down here.

Considered from the point of view of their effect on the text, various readings are either additions, omissions, or substitutions. But such a classification is of small use to the student of transcriptional evidence. What he desires to know is how various readings originate, that he may have some means of investigating the origin of the readings that come .before him. From this point of view, all readings may be broadly classified as intentional and unintentional corruptions. Every change brought into the text is the result either of a conscious and intentional alteration made by the scribe, or of an unintentional and unconscious slip into which he has fallen. Taking the mass of various readings together, a very inconsiderable proportion of them can be attributed to intentional changes, and any detailed classification of them is so far arbitrary that many readings may be equally easily accounted for on two or more hypotheses, and hence may be assigned indifferently to either of two or more classes. With this explanation a rough classification of the sources of error may be ventured, as follows :—

I. Intentional corruptions:

1. Linguistic and rhetorical correctional

2. Historical corrections.

3. Harmonistic corrections.

4. Doctrinal corruptions.

5. Liturgical corruptions.

II. Unintentional corruptions:

1. Errors of the eye.

2. Errors of the memory.

3. Errors of the judgment.

4. Errors of the pen.

5. Errors of the-speech.

Most of the corruptions which may be fairly classed as intentional fall under the head of linguistic and rhetorical corrections, and were introduced, we may believe, almost always in good faith and under the impression that an error had previously crept into the text and needed correcting. Sometimes they were the work of the scribe himself, sometimes of the official corrector (somewhat analogous to the modern proof-reader) under whose eye the completed MS. passed before it left the "publishing house." Examples may be found in the correction of dialectic forms, such as the rejection of the second aorist termination in a, and the substitution of the more common forms—e.g., rjKOofj.evt rjA.#€Te, ^\Oov for ijA0a/«v, rJA^are, rj\Oav; the euphonic changes which transform A.7j/xi/fo/, \-^}OeK into \.rj\[rofW.i, A^^ets or ocKaK«v into eyKaKeiv; the smoothing out of the grammar, as, e.g., when in Matt. xv. 32 fjftepai i-pets is changed into jj/xepas Tp«s, or in Matt. xxi. 23 IXBovros avrov into eXOovri avrif, or in Mark vii. 2 ep.ejjul/avro is inserted and thereby a difficult sentence rendered easy. Here, too, may be ranged such corrections as the change of the participles K/aofas and tm-apafas in Mark ix. 26 into Kpa£av and£av in order to make them agree grammatically with their neuter noun irvevjjui.. Examples of corrections for clearing up historical difficulties may be found in the change of "Isaiah the prophet" into "the prophets" in Mark i. 2; of "sixth " into "third " in John xix. 14, and the like. Harmonistic corruptions, though not confined to the Gospels (compare, for example, Acts ix, 5, 6 with xxvi. 14, 15), are, of course, most frequent there, and form, whether consciously introduced or unconsciously, one of the most fertile sources of corruption. Familiar examples may be found in the assimilation of the Lord's Prayer as recorded by Luke to the fuller form as recorded by Matthew, and the insertion of "unto repentance" in Matt. ix. 13 from Luke v. 32. Something very similar has often happened to the quotations from the Old Testament, which are enlarged from the Old Testament context or more closely conformed to the LXX. wording. Examples may be found in the addition of lyyt£« /*oi .... Ti3 Oto/witi avrZv Kcu out of Isa. xxix. 13 into Matt. xv. 8, and of ou i^euSo/xap-nyjijo-as in Rom. xiii. 9. On the other hand, it is doubtful if any doctrinal corruptions can be pointed to with complete confidence. Even the Trinitarian passage in 1 John v. 7 and part of 8 may have innocently got into the text. The most likely instances are the several passages in which fasting is coupled with prayer in some texts—as, e.g., in [Matt. xvii. 21], Mark ix. 29, Acts x. 30, 1 Cor. vii. 5; but even these are doubtful. Liturgical corruptions, on the other hand, are common enough, but can seldom be assigned to intention except in the service-books, where they deceive nobody, or in certain MSS. redacted for use as service-books, which have been fitted for public reading by such changes as inserting "And turning to His disciples He said," at Luke x. 22 (the beginning of a lesson), or of "But the Lord said," at Luke viii. 31, or the change of "His parents" into "Joseph and Mary," at Luke ii. 41, and the like.

So long, however, as we are dealing with corruptions which may with some plausibility be classed as intentional, we are on the confines of the subject. The fecund causes of the abounding error that has crept into the text lie rather in the natural weakness of flesh, limiting the powers of exact attention. From each of the sources of error which have been tabulated above as unintentional have sprung many kinds of corruption. Under errors of the eye, for instance, are to be classed all those mistakes, of whatever kind, which have arisen through a simple misreading of the MS. that lay before the copyist to be copied. The ancient mode of writing in continuous lines, and the similarity that existed between some of the letters, facilitated such errors. A considerable body of omissions have arisen from what is called "homoeoteleuton" or "like-ending." When two succeeding clauses or words end alike, the last is apt to be omitted in copying; the copyist, having written out the first, glances back at the MS. for the next clause, and, his eye catching the like-ending of the second clause, he mistakes this for what he has just written, and so passes on to the following words, thus omitting the second clause altogether. The same result often happens when the same sequence of letters occurs twice near together, and when two consecutive clauses begin alike instead of ending alike—a case which differs in name rather than in fact from the one just described. An example of "homoeoteleuton" may be found at 1 John ii. 23, where the whole clause, "He that confesseth the Son, hath the Father also," is omitted in some codices because both it and the preceding clause end with the words Tov irarepa l^ei. An instance in which only a few letters are involved is the omission of 6 'Irjo-ov; in Matt. ix. 28, which is apparently due to the custom of writing "Ii^rous in abbreviation, thus: AereiAYToicoic,—in which oic was easily mistaken for the preceding oic. Other examples are the omission of the whole verse, Luke xviii. 39, in a few codices, and of a clause in John vi. 39 by C.

Another error of the eye arises from mistaking similar letters for one another, such as, e.g., the confusion of (one way or the other) ei and H (Luke xvi. 20, eiAKoytevos—rj\Kiap.evos; 2 Cor. xii. 1, 8r;—Set); Tt and Ti (John vii. 31, /mj ir\eiovaftrjri TrAetova); H and N (Matt. xvii. 12, oo-a rjSeAijo-av—ocrav O€jcrav); 6 and o (Luke vii. 13, ea-ir\ay^vuj-Orjeinr\ayxyurov); Y and B (Aa/3tS—AawS), and the like. Possibly the famous reading ®«>s in 1 Tim. iii. 16 may have arisen as an error of the eye whereby oc was mistaken for the abbreviation 0c, which differs from it only by two light lines; although it may have equally well arisen as a strengthening correction or a mere blunder of a scribe, who mechanically added the lines which he had so frequently attached to this pair of symbols. The misreading of abbreviations was also a fertile source of error, and may be classed with errors of the eye. One of the most frequent instances results in the insertion of 6 'irjo-ovs after airois, by first doubling the oic, and then mistaking it for the abbreviated oic. In like manner we have Kcupi3 in Rom. xii. 11, probably through a misreading

of the abbreviated Kro> (Kvpuo) for K,pw (kcu/ju>). So too, the Kara iravra. of Acts xvii. 25 may have arisen from misreading Kvtattanta (kiu Ta iravra). A still more striking instance is found at Acts xiii. 23, where the abbreviation cp&m (or Ccothrain) has been misread as if it were cpi&N (or Ccothrian), and thus oxorJJpa 'j(rovv transmuted into rrtorrjpiav. Still another class of errors of the eye arises from the wandering eye taking up and inserting into the text a word or part of a word from a neighbouring line or a neighbouring column. Perhaps the form 'Acra.(f> in Matt. i. 7 has so come into the text from the influence of the 'Iuxraif>a.T, which stands immediately beneath it. Even whole lines may be omitted or exchanged by a similar slip, and this may be the true account to give of the varied relative position of the clauses in 1 Cor. i. 2. Another error of the eye of somewhat similar kind produces an assimilation of neighbouring terminations — as, for example, in Rev. i. 1, where rau oyyeAou avrov Tov 8ov\ov avrov stands for Tov ayyeAou avrov Tu> oov\ia avrov.

As errors of memory we should class all that brood which seem to have arisen from the copyist holding a clause or sequence of letters in his somewhat treacherous memory between the glance at the MS. to be copied and his writing down what he saw there. Hence the numerous petty changes in the order of words; the substitution of synonyms, as etWv for fyrj fa Matt. xxii. 37, « for airo, and the reverse (cf. Acts xii. 25), o/x/xaTW for 6(f>Oa\jj.wv in Matt. ix. 29, and the like; permutation of tenses, as, e.g., jSaTrri'cravres for /3aim'£ovTes in Matt, xxviii. 19, and the like. Here, too, belong many of the harmonistic corruptions, and the conformation of quotations from the Old Testament to the LXX. text, the scribe allowing his memory unconsciously to affect his writing.

As errors of the judgment may be classed many misreadings of abbreviations, as also the adoption of marginal glosses into the text, by which much of the most striking corruption which has ever entered the text has been produced. As the margin was used for both corrections and glosses, it must have been often next to impossible for the scribe to decide what to do with a marginal note. Apparently he solved his doubt generally by putting the note into the text. Doubtless this is the account to give of the abundant interpolation that deforms the text of such codices as those cited by the symbol D. More interesting examples are afforded by such explanatory notes as "who walk not according to the flesh but according to the spirit," inserted at Rom. viii. 1, to define "those in Christ Jesus" of the text; or as the account of how it happened that the waters of Bethsaida were healing, inserted at John v. 3, 4. Even more important instances are the pericope of the adulteress inserted at John vii. 53, sq., and the last twelve verses of Mark, both of which appear to be scraps of early writings inserted from the margin, where they had been first written with an illustrative or supplementary purpose. What a sleepy or stupid scribe could do in this direction is illustrated by such a reading as Se'£ao-$at i^/uas lv TroAAots T£>v a.vTi-ypdif>iav Outojs evprjrai Ko.l ov xa^ws ^Airicra/xev, which stands in a minuscule copy at 2 Cor. viii. 4, 5.

Under errors of the pen we class all that great body of variations which seem to be due to a simple careless miswriting of what lay rightly enough in the mind of the scribe at the time, such as, e.g., transpositions, repetitions, petty omissions of letters, and the like. It is impossible to draw any sharp line of demarcation between this class and errors of the eye or memory, and many readings combine more than one slip in their origin. For instance, when in Matt. ix. 15 we read Otanapsh in Codex D instead of oTAN&TTApeH, we recognise that there has been confusion of N and n, and then homoeoteleuton at work in omitting Att after An; but the result ia simply the omission of two letters. So, in 1 Cor. vii. 34, when D, E, omit the second Ko! in the sequence of letters MeMepiCT&iK&iH, we scarcely know whether to call it simple incuria, or to explain it by homoeoteleuton of the Tai and Km. On the other hand, when S writes eis To. dyia twice in Heb. ix. 12, or B repeats i(f>vyov ol Se Kpanjo-aires in Matt; xxvi. 56, 57, we have before us a simple blunder; and the like is found in every codex. Matters of this kind call for remark only when the slip of the scribe creates a difference in sense which may mislead the reader—as, e.g., when E, M, etc., transform e\a/3ov in Mark xiv. 65 by a simple transposition of letters into 6/SaAov, and H corrects this into HfiaX\ov; or when H, by a careless repetition, inserts an article into the phrase ex/SaA/Wra [tu] Saiftofio in Luke ix. 49. A more difficult case occurs at Matt. xxvi. 39, where N A, C, D, etc., read rrpoceAeooN, but B, M, II, etc., npoeA6a>N; either the former is a careless insertion, or the latter a careless omission of c, helped by the neighbourhood of the other round letters o and e

Finally, by errors of speech we mean all those which have grown out of the habitual forms of speech (in grammar, lexicography, or pronunciation) to which the scribe was accustomed, and which therefore he tended to write. His purism obtruded itself in correcting dialectic forms or Hebraistic turns of speech into accordance with his classical standard. Examples of this have been given under another caption. Sometimes, on the other hand, the idiom would be too elegant for his appreciation, and he would unconsciously conform it to his habitual speech. An instance may be seen in Acts xvi. 3, where D, E, H, L, P, substitute rjS«o-av yap awaiTes rov Trarepa Olvtov ori *E\jv vTrrjp\ev for the correct r/8eurav yap TrdvTfi on "EXXrjv o TraTTjp O.vtov vinjp^ev—to the ruin of the proper emphasis. The most considerable bodj of corruptions of this sort, however, grows out ol what is technically called "Itacism," that is, out of that confusion of vowels and diphthongs which was prevalent in pronunciation and could not fail to affect here and there the spelling. It consequently happens that i is continually getting written for « and vice versd, and ai and e; ij, t, and «; ij, 01 and v; o and ia; rj and e are confused in the spelling. For determining the age of these confusions of sounds in the speech of the people, we are dependent on epigraphical material, and on its testimony they must be carried back to a very remote antiquity. The confusion of « and i, for instance, occurs even in an Attic inscription earlier than 300 B.c., and was already prevalent in other regions' before that. From the end of the third century it was prevalent everywhere, while in the second century A.d. the distinction between the two was a crux orthographica. At the same time it must be remembered that a standard spelling was current, and carefully written MSS. tried to conform to it; so that we are not surprised to learn that the MSS. differ much among themselves in the amount and in the classes of itacism that have found their way into their pages. For instance, among the papyrus fragments of Homer, those usually cited as N and S are very free from itacism, while fi (of the first century B.c.) is full of it. Among New Testament MSS. N shows a marked preference for the spelling in i, and B for the spelling in «. Allowance for such particular characteristics must be made in passing judgment on readings ; but it must also be borne in mind that all the codices of the New Testament were copied at a time when itacistic spelling was current, and hence are more or less untrustworthy when the point is to distinguish between the vowels thus confused. The most common confusions are those between a and i, a> and o, at and e; and after these those between ij and the two pairs t and «, and 01 and v. The effect of the first may be illustrated by the readings e'SeTe and Bere in Phil. i. 30, or the readings larai, «arat in Mark v. 29. The most common effect of the confusion between o and ia is to confound the indicative and subjunctive moods; the following are examples: Matt. xiii. 15, "ao-oyiai K, TJ, X, A, tao-o/xai H, B, C, D, L, etc.; 1 Cor. xv. 49, , A,C,D, etc., c^opeo-o/xEv B, 46; 2 Cor.vii. 1, the way in which, even by the most honest copying, the text of any document may become corrupt; and to serve as examples of the kind of facts with which the student must have a personal familiarity in order to be prepared to trace back a reading to its source in a scribe's error, or to classify a body of readings according to their origination. It is important for him next to obtain an intimate knowledge of the habits, so to speak, of the important individual MSS. in order to check by familiarity with the habits of the one scribe the conclusions that are reached from a study of the general habits of all scribes. A fact in point has been already mentioned: N tends to write i everywhere for' «, and B to write « everywhere for i, and a knowledge of this fact is a help in determining readings involving « and i, for which these codices are sponsors. That A loves synonyms, or in other words the scribe that wrote this codex had an active mind that worked as he copied, and so felt the sense of what he wrote more than most scribes, is an important fact to know when we are deciding on the probability of a synonymous reading that A supports. That the scribe of N was a rapid penman, proud apparently of his handwriting; and that B's scribe was on the contrary a careful, plodding fellow, who copied the text before him with only such petty slips as such a writer would fall into,—brief omissions, doubling of short words, repetitions of letters and such stupidities,—these and such facts enable us to pass ready judgment on variations which might otherwise somewhat puzzle us.

Above all, however, it ig necessary to remember that every attempt to account for the errors that occur in our MSS. is an attempt to bring the accidental under rule, and every effort to classify them according to their sources is only an effort to group the effects of human carelessness; so that much must remain over of which we can only speak as instances of incuria. It may be useful to the student to look at a brief list of slips of the scribe of N, gleaned from the digest of the Epistle to the Hebrews, and to consider how many of them can be assigned to the several classes mentioned above :—

Incuria of K in Hebrews.

Heb. i. 5. Omit avrco from "I shall be [to him]

for a father."

„ i. 8. Omit rrfi evftmjros pa/?Sos.

„ i. 12. Add xat with a-u 8e.

„ ii. 18. Omit ?rapao-#«s.

., iv. 9. Omit the whole verse.

„ iv. 11. Omit Tk.

„ viii. 3. Omit Km.

viii. 10. Mov for fioi.

,, ix. 5. Evea-Ttv for ecrriv.

ix. 12. Eis T(i ayia written twice.

,, x. 7. Omit rjKta.

x. 11. Order changed to A«t. KaO.

„ x. 26. Trjs eiriyiwcriav for rrjv

,, x. 32. A/xopTias for ij/tepas.

,, x. 36. Change of order to xpetav

x. 39. Eis OTnuAias for ets aTroAeiav.

„ xi. 5. On for Sioti.

„ xi. 8. Change of order to Kjpovofuav XajL/3a.

There are in this list instances of errors of the eye (homoeoteleuton, the wandering eye catching a neighbouring word, confusion of similar letters), of the memory, of the judgment, of the pen, and of the speech, — and others also. It looks as if the scribe were taking a sly nap when he was writing the tenth chapter, and as if he either nodded again or was interrupted by an unthinking chatterer at xiii. 18, where, at least, we find a very odd case of repetition.

Efforts have been made to generalise upon the phenomena of the various readings, and so to furnish "canons of criticism" for the guidance of the student. Transcriptional evidence cannot, however, be reduced to stiff rules of procedure. All "canons of criticism" are only general averages, and operate like a probability based on a calculation of chances. A "chance" is always open that this particular instance is one of the exceptions. But, although to use them as strict rules to square our conclusions by were but to invite error, general rules are very useful, as succinctly embodying the results of broad observation. If we use them only as general guides, and expect to find exceptions to them continually turning up, the following three rules are valuable:—

1. The more difficult reading is to be preferred: founded on the observed tendency of scribes to render the sense smooth by correction or unconscious tinkering.

2. The shorter reading is to be preferred: founded on the observed habit of scribes to enlarge rather than shorten the text.

3. The more characteristic reading is to be preferred: founded on the observed tendency of scribes to reduce all they touch to their own level, and so gradually eliminate everything especially characteristic of an author.

Not co-ordinate with these, but above them and inclusive of them, stands the one great rule that embodies the soul of transcriptional evidence: that reading is to be preferred from which the origin of all the others can most safely be derived. Knowledge of the habits of scribes and of the phenomena of MSS. is needed to interpret this rule. Commonsense is here even more than usually needed. But given the knowledge and common-sense, this one rule adequately furnishes the worker in this department of evidence.

That much could be done towards settling the text of any work by the use of intrinsic and transcriptional evidence alone, which would be generally recognised as sound, is certain. But it is equally clear that a special danger attends processes that are so nice and delicate, of the intrusion of those wishes that are fathers to thoughts; and in criticising the text of a book that stands in such close relation to our dearest beliefs as the New Testament, this danger reaches its maximum. This does not render the method of internal evidence of readings invalid; nor does it exonerate critics from the duty of using it,—with strict honesty and a severe exclusion of improper subjectivity. But it throws sufficient doubt on individual judgment in attaining some of its results, to render it desirable to test its conclusions by some less easily warped method of investigation. We gladly remember, then, that besides "internal evidence of readings" we have "external evidence of readings " 'to depend on, and proceed to inquire after the methods of using it.

2. External Evidence Of Readings.

(a) Comparative Criticism and Internal Evidence of Documents.

The crudest method (hit could be adopted to decide between readings on the ground of external evidence would be simply to count the witnesses for each reading and follow the greatest number. It requires little consideration to perceive the illegitimacy of such a method. The great practical difficulty stands in the way of adopting the principle that the majority shall rule, that we cannot certify ourselves that we have the majority. For this, we must first collate every known copy, and even then the doubt would hang over us that mayhap the majority of copies are yet unknown: have not, indeed, the majority actually perished? If we should adopt a simple majority principle, therefore, we could never reach certainty; we could never be sure that the copies as yet unknown, or hopelessly lost, might not alter the balance; and we should be betraying the text into the hands of the chance that has preserved one MS. and lost another. A greater theoretical difficulty lies behind. Who can assure us that the many are the good? The majority of MSS. are late MSS.; and if it be the original text that we are seeking, is it likely that the many MSS. of the eleventh century will better help us to it than the few of the fourth? Dare we overmatch the multitude of years by the multitude of copies,—our two codices of the fourth century by the mixed hordes that throng on us from the fourteenth? If corruption be largely due to the fortunes of hand-copying, it will of necessity be progressive, and the MSS. of the earlier centuries may be rightfully presumed to be purer and better than those of the later. We may even expect to find in them the parents of the very later codices which now would crowd them out of the witness-stand. If so, to follow mere numbers is to betray the text into the hands of the later corruption.

Shall we, then, say that not the most MSS. but the oldest shall rule? This certainly would be a far better canon. But it is met again, on the threshold of practical use, by a double difficulty,—theoretical and practical. After all, it is not the mere number of years that is behind any MS. that measures its distance from the autograph, but the number of copyings. A MS. of the fourth century may have been copied from another but little older than itself, and this again from another but a little older than it, and so on through a very long genealogy; whereas a MS. of the eleventh century may have been copied from one of the third, and it from the autograph. It is not, then, the age of the document, but the age of the text in it, that is the true measure of antiquity; and who shall certify us that many of our later documents may not preserve earlier texts than our earliest MSS. themselves ?—or, indeed, that all our later documents may not be of purer descent than our few old codices 1 With the frankest acceptance of the principle that the age of a document is presumptive evidence of the age of the text, it is clear that we can reach little certainty in criticism by simply agreeing to allow weight to documents in proportion to their age. And here the practical difficulty enters the problem: how much greater weight shall we allow to greater age? Certainly two fourth-century documents cannot reduce all tenth-century documents to no value at all, simply by reason of their greater age: but how nice the question as to the exact increment of weight that must be added for each century of additional life! Professor Birks set himself once to investigate this question; and his conclusion was "that on the hypothesis most favourable to the early MSS., and specially to the Vatican [B], its weight is exactly that of two MSS. of the fifteenth century, while the Sinaitic [n"| weighs only one-third more than an average MS. of the eleventh century." Mr. Monro was at pains to point out certain errors in Professor Birks' calculations which appear to vitiate his conclusions. But for the purposes of actual criticism were they not valueless even if correct? How is it possible to calculate the value of each document relatively to all the others on the ground of age alone t Let us confess it: to admit that the older a MS. is the more valuable it is likely to be, carries us but an infinitesimal way towards the actual work of criticism, and it is entirely impossible to apportion their values to codices by their ages. Though we may feel that a MS. of the fourth century ought to be a better and safer witness than one or two, or a hundred, or a thousand for that matter, of the fifteenth, we cannot certify ourselves of this with regard to any given MS.; and we certainly cannot arrange all our MSS. in a table of relative weights as resulting from their relative ages, and then use this table as a touchstone for our critical problems. It is a plain fact that MSS. need not and do not always vary in weight directly according to age.

A great step forward is taken when we propose to allow MSS. weight, not according to their age, but according to the age of the text which they contain. To Tregelles must be ascribed the honour of introducing this method of procedure, which he appropriately called "Comparative Criticism." It is a truly scientific method, and leads us for the first time to safe results. Briefly stated, it proceeds as follows. The earlier versions and citations are carefully ransacked, and a list of readings is drawn from these dated sources which can be confidently declared to be ancient. Each MS. is then tested, in turn, by this list. If a MS. contains a considerable proportion of these readings, or of readings which on grounds of transcriptional probability are older than even these, it is demonstrated to contain an old text. If, on the other hand, a MS. fails to contain these readings, and presents instead variants which according to transcriptional probability appear to have grown out of them, or which can be proved from dated citations to have been current at a later time, its text may be assumed to be late. From an examination of the MSS. thus proved to exhibit an early text, we may next obtain a very clear general notion of what the earlier text is, and this will serve us as a more extended test of the age of texts contained in MSS., and we may confidently divide them into two great classes—the early and the late.

Here, it is plain, our feet rest on firm ground. What may be done towards settling the text by this method may be observed in the text which Dr. Tregelles actually framed, and which stands to-day as his suitable and honourable monument. But a little consideration will satisfy us that, as an engine of criticism, this method is far from perfect. It will furnish us with a text that is demonstrably ancient, and this, as a step towards the true text, is a very important gain. It is something to reach a text that is certainly older than the fourth century,—that was current in the third or second century. But this can be assumed to be the autographic text only if we can demonstrate that the text current in the second or third century was an absolutely pure text. So far from this, however, there is reason to believe that the very grossest errors that have ever deformed the text had entered it already in the second century. By this method, therefore, we may deal successfully with all cases of variation in which the older and later texts stand opposed as bodies, and thus may sift out a vast rabble of late corruptions; but we stand, with it only to help us, helpless before all cases in which the oldest witnesses themselves differ. This result might have been anticipated. If our touchstone only reveals to us texts that are ancient, we cannot hope to obtain for our result anything but an ancient text. What we wish, however, is not merely an ancient but the true text.

Yet another process has been developed for our aid in this perplexity. It has been pointed out that the way is open to the estimation of MSS., not by the age of the parchment on which they are written, nor yet by the age of the text which they contain, but by the actual excellence of the text which they contain. This is another great advance. For we are now invited to assign weight to MSS. according to their real value. The process by which this method undertakes to ascertain the relative value of the different MSS. is appropriately called "Internal Evidence of Documents," and proceeds by interrogating each MS. as to its own value, by testing it by the only kinds of evidence available—namely, intrinsic and transcriptional evidence. A rude example of what is intended by this will, perhaps, be its best explanation. Let us suppose two copies of a will or deed to be laid before us, and it to be our task to determine which is the better—i.e., the more correct.

What would be the common-sense procedure? Beyond doubt, we should begin by noting every point in which they differed; and then, taking this list of various readings, we should ask, in the case of each reading, which appeared to be the original. We should have two ways of determining this: in each case we should ask, Which reading is it probable, considering the context, style, and the like, the author wrote] and, Which reading, considering the known habits of the scribes, the accidents to which they are liable, and the like, is it probable that the scribe had before him in order to produce the other? When these two modes of inquiry resulted in the same answer, the reading .would be determined by a high degree of probability. Now, after having thus passed through the whole list of various reading?, we could count up what proportion of them had been determined in favour of one MS. and what proportion in favour of the other. This would furnish us with a fair general estimate of the comparative value of the two copies. If, for instance, the two differed in a hundred places, and the two varieties of internal evidence of readings united in commending the readings of one in ninety of these, and those of the other in only ten, we should have no difficulty in greatly preferring the former to the latter copy. Nay, it would not be strange if we now revised our decision in some of the other ten cases, and allowed our demonstrably better copy to determine their readings on documentary grounds. No doubt such a method offers us only probable results; but it is scarcely open to doubt but that, so far as they go, they are sounil results, and in favourable cases the probability may reach moral certainty. It is equally plain that the method is not essentially affected if the documents we have to compare are a dozen instead of two, or even a hundred or a thousand; nor yet if our two varieties of evidence fail to give us clear or united testimony in a number of the readings. It would still remain true that the relative value of the MSS. could be ascertained by determining the proportionate number of their special readings which internal evidence will commend. After its own relative value has been assigned to each MS. of a work by this method, we may proceed to its textual criticism on documentary grounds, allowing each MS. the weight thus indicated. This is not reasoning in a circle. By one process, tentatively applied, we attain a general notion of the value of each MS. When a considerable number of readings have been used in this work, errors in their estimation check one another, and our general result is sound. It is quite consistent next to treat all these readings as still undecided: this is but to recognise that tentative results as to the details are provisional. We may, therefore, justly call in the MSS. according to the relative values which have been assigned them by our tentative results en masse to decide now on each reading in detail.

Precisely this process has been applied to the MSS. of the New Testament. And we are asked to determine the relative weight of the witnesses for each disputed reading by allowing to them the weights assigned them by this method of testing. It would be idle to dispute the validity of the process. It is transparently just and scientific. It is equally impossible to doubt that it will enable us to come to conclusions on which we can depend. Especially when taken in connection with the former method, which marshals MSS. according to the age of the texts they exhibit, this method, which marshals them according to the tested value of their texts, will lead us to very important conclusions, both in the way of testing the results obtained by the former method, and in carrying them some steps farther. The mere fact that the results of this method accord with those obtained by the former, so far as they were legitimate, gives us confidence in using it. It may be in one sense an accident that our oldest MSS. should be shown by comparative criticism to contain the most ancient text, although an accident in the line of the pre-existing presumption. But it cannot be by mere accident that the text obtained as the most ancient should in the main accord with that obtained as the best. And it is reasonable to be led by this accordant result of two independent methods to put confidence in the further results obtained by one of them which in the nature of the case cannot be tested by the other. We are justified, therefore, in using internal evidence of documents to decide for us the readings in which the older text is itself divided.

As already intimated, Dr. Tregelles' text may be taken as the type of the results attainable by comparative criticism. He was accustomed to divide the MSS. into classes, thus: (a) Uncial MSS. of the most ancient class,—i.e., those earlier than the seventh century; (6) Later uncial MSS. of special importance;

(c) Certain important MSS. in minuscule letters;

(d) The later uncials. He aimed at citing the testimony of all the uncial MSS., those of the minuscules the text of which was ancient, all versions down to the seventh century, and the fathers down to and including Eusebius. In class (6) he included L, X, Y, A, ®, H, of the Gospels, P of Acts and the Catholic Epistles, and F, G, of Paul. In class (c) he included 1, 33, 69 of the Gospels, 13, 31, 61 of Acts and the Catholic Epistles, 17, 37, 47 of Paul, and 38 of the Apocalypse. To these might well be added, now, the minuscules cited in the lists of minuscules given in the proper place above. The other classes (a), (d), may be gathered from the lists of uncial MSS. given above. When tested by internal evidence of documents, the MSS. arrange themselves in a not dissimilar classification. As is practically universally confessed, B is by this means shown to be the best single MS., and N stands next to it. Naturally enough the documents most like B are given the next place. But the general character of such codices as D, D2, G3, F2, is not very high, when tested by internal evidence of documents, although their text is certainly very old, as comparative criticism satisfactorily proves. Among the versions, the palm falls to the Memphitic and Thebaic.

A various reading that occurs in Matt. vi. 4 may serve us as an example of the working of these processes. Shall we read in this verse simply, "And thy Father that seeth in secret shall reward thee "? or shall we add the word "openly" at the clcwe 1 Tischendorf states the evidence thus :—omit iv ria ^avepco: NBDZ 1. 22. 108. 209. al5 cdd81" ap

Aug (multa exx. Latino, sic reddet tibi palam

in Greeds quce priora sunt non invenimus

palam) ff1 k, vg fr sax cop syrcu (Or «.*».«m. non liquet quo spectet), Gyp Aug Hier Chrom al; insert cv Tio ^avepw: E K L M S U XTid (e spatio) a b c f g1 h q syrsch etp go arm seth al Const Chr Op al. In order to interpret the evidence by comparative criticism, we may arrange the matter as follows:—

We observe that the addition "openly" does not occur in any known Greek MS. before the eighth century, or in any version or patristic citation before the fourth century. Some good later uncials, L of the eighth century, and apparently also X of the ninth, witness for it, but the better minuscules, again, omit it. No second-century version contains it, but all later ones do, with the sole exception of the Latin Vulgate. Its absence from this and from Jerome's quotations is probably to be explained by Augustine's precise statement that many Latin copies of his day contained it, but none of the earlier Greek copies,— which in itself is a very strong testimony to the superior antiquity of the omission. On this evidence the conclusion is probable that lv ria (f>avepia, balancing the previous ev ria Kpinrria, was first introduced into the Greek text late in the third or early in the fourth century. When we now withdraw our attention from the question of antiquity, and consider the witnesses according to their values, as determined by "internal evidence of documents," we discover that the best witnesses array themselves for omission. On this ground, too, therefore, we decide to omit the words.

Practically much the same division of evidence is met with in the more important matter of the insertion or omission of the doxology to the Lord's Prayer (Matt. vi. 13). There is, however, this important difference: the doxology appears in witnesses as early as the second century. For its omission are quoted: N, B, D, Z, 1, 17, 118, 130, 209; scholia in the margin of many copies that contain it; a, b, c, ff1, g2, 1, vg., cop.; Or., Nyss., Cses.,Cyrhr., Max., Cyp.,Tert., etc. For its insertion: E, G, K, L, M, S, U, V, A, II, 2, $, very many others, f, g1, [k] q, syrutr, etOT, etbr, scth., arm., go. [sah], [Teaching of the Twelve Apostles,] Constt., Chrys., and later fathers generally. The MS. evidence does not differ markedly from the distribution observed in Matt. vi. 4. But among the versions a doxology is f onnd in the second century Curetonian Syriac and the Sahidic (=Thebaic); and in the fathers, in the early second century " Teaching of the Apostles." There is no question, therefore, but that a doxology is found attached to the Lord's Prayer as early as the very opening of that century. Nevertheless, the oldest MS. in which it is found dates no higher than the sixth century (2). Even with comparative criticism alone beneath our feet, we are not helpless here; for when we observe that the doxology appears in the Becond century in as many differing forms as there we documents that contain it, that it occurs in no MS. before the sixth century, and in no commentator on the Lord's Prayer before Chrysostom at the end of the fourth century, conclusions as to its late origin present themselves with some force, and we can suspect that it entered the Greek Testament about the end of the third or opening of the fourth century. When we call in "internal evidence of documents," we see that the best old documents are ranged for omission, and our conclusion is strengthened accordingly

The reading in John vii. 8, the evidence in the case of which was analysed a few pages back, is distinctly more difficult to deal with. The two oldest and best MSS. are here set in opposition to one another; the second-century versions are divided as throe to one, but the best and the worst agree against the second best, and the most stand with the second MS. against the best. This is typical of the division of the evidence throughout. How, then, can we decide the matter on grounds either of the antiquity of the witnesses or of their excellence? Cases of just this complexity meet us on nearly every page of the New Testament. What are we to do with them t

These examples have been designed to illustrate both the strength and the limitations of the method of criticism which we are expounding. That much can be accomplished by it is clear. That it is scientific and sound, so far as it will carry us, is equally certain. But it is also true that it is helpless whenever the old or the good documents are pretty evenly divided; and that when, as in the New Testament, we have many documents to deal with, it does not always carry with it that practical certainty which we desiderate. The reason of both shortcomings is that its decisions rest everywhere, at bottom, on an arithmetical balance. Let us try to explain.

By this method of criticism, when all the old MSS. stand opposite the later, and when all the good MSS. stand opposite the bad, we have no difficulty in deciding the reading. But they will not always so arrange themselves; perpetually some of the older are on the side of the later, some of the better on the side of the worse. What are we to do in such cases? Even if we are confident that N, B, A, 0, D,

when combined, may stand against the world, how do we judge the group to be weakened by the defection of A? or of C i or of B? or of N, B? or of A, C, D? or of any two or any three or any four of them? These are puzzling questions. But until they are answered this method of criticism is helpless before the immense variety of divided testimony which meets the critic in every part of his work. Clearly, in such cases everything depends at bottom on our knowing not only that N, B, C, D, present an old, and E, S, TJ, V, a late text; or that N, B, C, present a good and most minuscules a bad text; but also, very accurately indeed, the exact proportional excellence and consequent weight of each MS.: how much better precisely B is than N, and N is than C, and C is than V or 10 or 19. How else can we estimate the effect of each defection? Often decision on the bearing of documentary evidence will absolutely depend on an exact knowledge of the precise value of each MS., and a consequent ability to estimate the weight each brings to a group with its presence, or takes from it by its absence. Obviously this means (at our present stage) nothing less than ability to speak of MSS. in terms of numerical formulse, and the whole matter of documentary evidence becomes an arithmetical balance. If, assuming an ordinary minuscule of the fourteenth century to rank as 1 in weight, we know that B ranks as 2000, and N as 1800, and G as 1600, and so on, we can accurately estimate the value of each group and by a simple sum in arithmetic settle the text. But unless we know this or something equivalent to it, the bearing of the documentary evidence is constantly escaping us. We cannot tell what effect on the weight of N B A C D, for example, the defection of B will have; we cannot tell whether N B D Z may not be enough to carry our suffrages, and N B D not enough; whether E K L M S U X may not be too weak to follow, but EGKLMSUVAHS^ too strong not to follow. Manage it by whatsoever method we please, and conceal the fact from others or ourselves by any way of speaking of it that we may, the whole process of criticism which deals with MSS. as separate units amounts to nothing less, at bottom, than an attempt to settle readings by an open or veiled arithmetical balance. We are not now arguing whether such a method be not fundamentally wrong; but only that it cannot be carried successfully through any case where the testimony is well divided unless the arithmetical balance be accurately estimated. And it is clearly apparent that such a balance is not accurately estimated, and, indeed, cannot be. But by as much as it is not, by so much is our criticism but little removed in all nice problems from guesswork.

Let us try to realise in thought still further, what is implied in the very attempt to decide readings by such a balance. No less than this: the possibility of overwhelming all early and good testimony by the sheer numbers of late and bad testimony. Does not the very principle of an arithmetical balance yield the point that the early and good may be overborne by the late and bad, if only the latter be numerous enough? So, in pretending to estimate and weigh witnesses, we fall into the trap of merely counting them. What we want is a method which will allow later testimony to overrule earlier, only if it be good enough to do so. But this method and all methods of a mere balance of individual documents inevitably puts itself in the position that the best and oldest may be overborne, if only we can produce a sufficient number of later documents. Say that B is made equal to two thousand thirteenth-century copies, and ten or a hundred thousand nineteenth-century copies, it would be in the power of an enterprising printer to produce enough very debased copies to overbear its testimony. The procedure would be transparently ridiculous, no doubt; but this only proves that we need some method of criticism which is not capable of such a reductio ad absurdum,—which does not proceed on an assumption which can only arbitrarily protect us from such a conclusion. Something else is needed beyond knowledge of the general relative age of the texts that documents contain, or the general relative goodness of them, or anything that concerns single documents, before we can reach very secure results.

That those who have made use of "comparative criticism" have avoided the weakness of an arithmetical balance in dealing with all that class of readings in which the older text differs from the later is no doubt true. But they have done it by confessedly or practically ignoring all later testimony. In this they have built better than their theory gave them ground for, and they have given us a text, consequently, better than their theory would legitimately defend. It has not unjustly been made their reproach that because they had discovered that the better testimony was to be found in a certain body of witnesses, they arbitrarily treated all the rest as if they had no testimony to offer at all. And in all that class of variations in which the older documents differ among themselves, these great critics have continually fallen a prey to the imperfection of their method, and their results have depended less on a scientific procedure than on a certain personal quality which we may call "critical tact," and which is but another name for a keen appreciation of the bearing of internal evidence of readings. The discovery of a single MS. (N) revolutionised Tischendorf's text. Tregelles, always more cautious and consistent, was yet repeatedly led into the most patent errors. Every one who has attempted to decide on the weight of documentary groups on any large scale has necessarily been made to feel very keenly that very much of criticism which depends on such methods, wherever internal evidence of readings is not really decisive, is little removed from arbitrary decision or guesswork. From all which it is clear that some method which will enable us to deal with MSS. in groups and classes rather than as individuals is absolutely necessary before we can determine more than the outlines of the text with confidence.

(6) Internal Evidence of Groups.

A method of procedure which will relieve us from these difficulties has been pointed out under the .appropriate name of "internal evidence of groups." Internal evidence of readings is the evidence of its own value which each reading supplies when subjected to the tests of intrinsic and transcriptional probability. Internal evidence of documents, as we have just seen, is the evidence of its own value which each document furnishes; and is obtained by noting what proportion of the characteristic readings of a document approve themselves as probably genuine under the twofold test of intrinsic and transcriptional evidence. This process can be carried, with equal ease, a step higher, and be applied to any given group of documents, and thus become internal evidence of groups. Nothing prevents our collecting all the readings supported by any group of documents in which we may be for the time interested, and then trying the list in each of its items in turn by transcriptional and intrinsic evidence. If the majority of its characteristic readings, when thus tested, approve themselves, the group is a good group; if the majority are condemned, it is a bad group; and the proportion between those approved and those condemned will furnish an accurate criterion of the actual value of the group. When two or more groups are successively subjected to this testing, the proportional result obtained in each case supplies data for determining their relative values.

Thus we may at will obtain, by this process, grounded decision as to the weight of any given group, and so determine the actual composite value of any combination of documents. If, for instance, we are studying the reading in John vii. 8, which we Tiave already had before us, we may take the group N D K M II 17** 389 psor, and trace it throughout the Gospels, collecting all the readings which it supports into a list. Next we may test this list of readings by transcriptional and intrinsic evidence, and thus attain a very good, and certainly a well-grounded notion of the value of this group. It only remains, now, to return to the reading in hand, and allow the group there the weight which we are thus led to assign to it. We no longer try to estimate the weight of the group by the sum of the weights of its component parts; we no longer need to raise question as to the relative values of the separate MSS., and the effect of the defection of this one or that; we treat the group as a unit, and estimate its value as a whole. Instead of speculating as to the difference between N D K M II 17** 389 pscr and B N D K M II 17** 389 p**, or trying to calculate it by adding the weight of B to the weight of the former group, we simply go with this process to the places where these groups occur, collect the readings actually supported by each, and try each separately by the only kinds of evidence applicable, and so find for each in turn what its actual value is. The result is oddly portentous for all attempts to estimate readings by arithmetical balances. As a mere matter of fact, wherever N D K M H 17** 389 pBCr, or its essential elements, occur, it is usually in support of an obviously wrong reading; and wherever B is added, this greater group usually supports an obviously right reading. In other words, the former is a bad and the latter a good group.

Two practical limitations, in the use of internal evidence of groups, need statement at the outset. In estimating the value of any group, we must confine ourselves within the limits of the section of the New Testament in which the reading we are to study occurs, and, in the first instance at least, within the strict limits of the group we are investigating. There is every reason to believe that our great MSS. which contain, or once contained, the whole New Testament, were made up directly or remotely of copies of different codices in the several parts of the New Testament; and, indeed, that in the early days of the Church each section was usually written in a volume apart. The result would naturally be that the Epistles of Paul, say, for instance, in Codex B, would have a very different history, could it be discovered, from that of the Gospels in the same codex. As a matter of fact, also, the result of the actual test gives a different value to the same apparent group in the several sections. Very divergent weights are assigned by it to A in the Gospels and in the rest of the New Testament. In the Gospel of Mark £ A is excellent, but B G in Paul is very suspicious. Experience thus teaches us that the value of the separate groups must be studied apart for each great section of the New Testament. The same experience teaches that it is not safe to confound two groups which look alike. No man knows whether B N D L has the same value as, or more or less value than, B N D, until he has actually tested the matter empirically. We may afterwards learn from actual trial the limits within which each group may vary without essentially altering its weight, but we must be chary of assumption in this matter. Take the group D E F G in Paul. If we add N to it its value is unaffected. Or if we add B to it, it is essentially the same. If, however, we add both N and B, the group immediately changes from bad to good.

The immense advance that is made, by the introduction of this method, on all criticism that depends on estimating the values of groups from the values of the members that compose them, is apparent at a glance. All the difficulties and dangers of an arithmetical balance are escaped at a single step. We now estimate the weight of any group which supports a given reading, not by the age of the MSS. which compose it, nor by the age of the texts which these MSS. contain, nor by the value of the separate MSS., but by the tested value of the group itself. Each group stands before us as a unit; each is first tested as a unit, and then used as a unit. The full importance of thus escaping the arithmetical balance will not be appreciated, however, until we realise that the union of two codices will not necessarily, and indeed is sure not to, be the same in weight as the sum of their values. For example, N B is not the same as N + B; and any system which proceeds openly or practically by an arithmetical balance is sure, therefore, to lead to error, which cannot be legitimately escaped until we learn to deal with groups in some way or other as units of testimony. Internal evidence of groups assigns to N B no weight as a composition of N and B, but recognises it as a third thing (just as blue plus yellow make the third thing, green), and seeks to discover its own value as it betrays it from the readings it supports; it thus accords it only the weight which it makes good its claim to.

The soundness of this method of work is bound up inseparably with that of internal evidence of documents, from which it differs rather in name than in fact. It does for groups of documents just what the former process does for single documents. It makes no assumptions as to how documents come to be grouped; it accepts as a fact that here is a circumiscribed group supporting a series of readings, and then asks what kind of readings, good or bad, does this group support 1 It thus estimates the value of a witness by the character of what he witnesses to, '—by his habits of truthfulness or the contrary elsewhere,—and gives him credit accordingly. No less obvious than that the application of this method will' give us secure results is it, however, that it will entail a great deal of labour. It is far easier to guess at the weight of a group, or to leave it unguessed and fall back on internal evidence of readings as our sole dependence, than laboriously to test the weight of a group. The beginner may well be somewhat appalled at the prospect of painfully tracing every chance combination of documents through the crowded digests of a Tischendorf or a Tregelles, and even after this labour is completed, of feeling that the most trying task is still before him, —the careful testing of each one of the readings thus

obtained by internal evidence, with a view to deter mining the value of the witnessing group. Yet, the result is worth the labour: royal roads have not a good reputation for safety, and the very thorns in this path have their useful lessons to teach. And it is right to point out that the number of groups needing testing is found in practice far fewer than would & priori be thought likely. The New Testament MSS. do not arrange themselves in every conceivable grouping, and the student will not proceed far in this work without discovering that the number of varying groups that actually occur is comparatively small, and further, that these may be reduced to yet fewer by attending only to the essential core of each,—a core that can only be empirically discovered, but which yet, after a while, can be with certainty abstracted.

In a matter of this kind no one can afford to accept implicitly the results of other investigators and simply apply them to special cases. It is strongly recommended that every student actually study for himself the value of some few selected groups at the very outset, and that he be prepared to test all results of others in the same line of work, and to make trial of any group that puzzles him in any special reading. At the same time, the beginner may be allowed to stand on the shoulders of the masters of the science, and perceive the bearing of evidence through their eyes. Dr. Hort, in particular, has worked out the values of the chief groups throughout the New Testament, and his results may be safely accepted as sound. The most interesting of these results is the very high character given to the compound B N, which approves itself as nearly always right, whether it stands alone, or with whatever further body of documents, and that throughout the New Testament. Next to B N, B conjoined with some other primary document, such as B L, B C, B T, and the like, whether alone or with other support, forms the most weighty series of groups, and this, again, throughout the New Testament. The only outstanding exception to this last generalisation is formed by B G in Paul's Epistles, whether alone or with other documents short of the whole body of primary uncials, which is usually condemned by internal evidence. B D in Paul is a good group, although B D Gr is bad, and although it hardly attains the very high excellence of the like group B D in the Gospels and Acts, whether alone or in combination with other documents. On the other hand, N D is everywhere, and in every combination (if B be absent), very suspicious. Even with secondary witnesses only adjoined to it, B stands the test excellently; and if clear slips of its scribe be excluded, even when wholly alone, B attains great excellence and stands forth as plainly the best single codex known. On the other hand, compounds of N with other documents (B being absent) are usually not strongly commended, and compounds of documents excluding both N and B are commonly condemned by internal evidence. In the Apocalypse N falls to a low level, and A rises to the height of the best single MS., while A C is the best binary group, and is usually to be trusted, whether it stands alone or in combination with other documents. A very special discredit is thrown on D G in Piiul's Epistles, whether it stands alone or in any combination, provided only that both B and N are not adjoined to it.

These generalisations, all of which the student would do well to test by actual trial, already put us in a position to deal with most readings. For instance, in John vii. 8 internal evidence of groups clearly commends ovn-ta; for the good group B L T etc. supports it, while the bad group N D etc. supports its opponent. So too in Matt. vi. 4 the group that omits lv Tia if>o.v€pia—viz., B N D Z—is seen, at a glance, to be one of the strongest possible. The same is true of the group that omits the doxology in the Lord's Prayer. In a word, internal evidence of groups puts an engine of criticism into our hands which cuts the knots that seemed incapable of being unloosed by the older methods, and enables us to reach assured convictions as to the bearing of the external evidence, where before we stood helpless.

If in any case Dr. Hort's generalisations do not seem easily or safely applicable, or the results of their application bring us to a conclusion which seems difficult to square with internal evidence of readings, it is the duty of the inquirer to subject the special group before him to a renewed and independent testing. But even with the most easily studied and safely interpreted groups, it must be remembered always that we reach general and probable results only, and not invariable and unmistakable ones. The character assigned thus to groups of MSS., like the character assigned to individual MSS. by internal evidence of documents, is general character, and is quite consistent with the best groups being sometimes in error. The rules of procedure derived from internal evidence of groups are, therefore, not without exceptions. This may be illustrated by such a reading as that found in Matt, xxvii. 49. Here N, B, C, L, U, r, five minuscules, some mixed Latin MSS., a copy of the Jerusalem Syriac, the ^thiopic version, and Chrysostom, with perhaps some other fathers, insert the sentence, "But another, taking a spear, pierced His side, and there came forth water and blood," to the confusion of the narrative. The intrinsic evidence seems immovable against the insertion; the transcriptional evidence seems to judge it an assimilation to John xix. 34, clumsily done. Bvit if the internal evidence is thus united against the insertion, we can scarcely insist on inserting it on account of the testimony of internal evidence of groups. Though this group is about as strong a one as can occur, yet internal evidence of groups gives us only the comparative weights of groups when considered throughout all their readings; it does not give us an exceptionless rule to apply mechanically. We learn from it what amount of correctness N B C L U r is apt to exhibit, not what amount it must have in every reading. The way is open for us to find some exceptions to the general excellence of the group, and hence to find »n exception here.

If, however, the estimation of the value of the various groups which is attained by internal evidence of groups allows for exceptions, and attains only a probable force, it becomes immediately important to check its results by some other independent method of criticism, which will enable us to determine which are the readings in which the exceptions are found. That an independent method lies within our reach may be hinted by our use of internal evidence of groups itself. We shall not proceed far in using this method before we realise what has been already remarked: that the number of groups that actually occur in the digests is far short of the calculable number of possible combinations of the documents. We shall observe a certain persistency in some MSS. in getting together, and a certain persistency in keeping apart manifested by others. Nor will accident account for this. It is, no doubt, possible that two or more MSS. may occasionally unite in a reading by accident. But how rarely and in what a narrowly limited class of readings this can occur, a very little reflection will assure us. Only in such obvious corrections or in such unavoidable corruptions as two scribes might independently stumble upon, can codices agree accidentally. The improbability of many MSS. falling independently into an identical corruption of even this kind, and the still greater improbability of a plurality of MSS. falling independently into a considerable series of identical corruptions, is too immense to be apprehended. MSS. which fall frequently together can owe their frequent conjunction to nothing else than common inheritance. This is, indeed, the principle on which all textual criticism proceeds. We seek the original text of the New Testament in the extant MSS., because we judge that where these MSS. agree, this agreement can be accounted for in no other way than by common inheritance from the ancestor of all. The same principle is, of course, valid for any given group of MSS. short of all: their union in a body of readings common to them, and more or less confined to them, is proof that they are preserving in these readings parts of a MS. which, for these parts, lay at the root of all the MSS. in the group. When we gather together the readings of any given group of codices, we are gathering, therefore, a body of readings from a lost MS., the common parent in these readings of all the codices of this group. And when we test this list of readings by internal evidence of groups, we are only in appearance performing a process different from internal evidence of documents; we are testing a lost document, a body of the readings of which we have recovered, instead of an extant document all of the readings of which are before us. Internal evidence of groups is, therefore, simply internal evidence of documents applied to lost documents, a list of the readings of which has come down to us, and nothing more. This is why we have said that its validity is bound up with the validity of internal evidence of documents, and must stand or fall with it.

From this point of view we may understand why we find it in practice of the utmost importance to confine the examples of the use of any given group which we are testing, strictly within the bounds of the group that stands before us. Every MS. added to the group may carry us another step back for the common parent of the (now enlarged) group. If B C D in Paul, for instance, is being tested, we must exclude all readings supported by M B C D, because we do not know whether the common ancestor of N B C D may not bo another MS. from the common ancestor of B C D, and thus we may be confusing two MSS. in our investigation and therefore obtaining results inapplicable to either. No doubt everything in N B C D must have been in the MS. which stood at the head of the subgroup BCD; otherwise it could not have been inherited by B and C and D. And if our purpose were to recover as much as possible of the common ancestor of B C D, we should have to collect all readings found in these three MSS., no matter what others were added to them. But since our purpose is to test the value of this reconstructed MS., our first duty is to select from the whole mass of its readings those in which it differs from the opposing group, just as, in internal evidence of documents, we confined our attention to the list of various readings. To pay attention to all the readings of any MS. or group of MSS. gives us no basis of comparative judgment, since the readings common to both documents or groups cannot discriminate between them. Consequently, for internal evidence of groups the labour is lost which is spent on collecting readings which we cannot use, for the sake of sifting them out again. And it is worse than lost. Suppose we are testing the value of B. Is it valid to take account of the readings for which B N witness? Certainly not, in order to obtain a value to assign to B when it stands alone. And simply for this reason; B N is not B, but the common ancestor of B and N; and the value of this common ancestor of the two cannot be assigned to either separately without leading to extensive error. No doubt B has preserved in all cases where B and N stand together the reading of the common ancestor of them both. But this does not prove that it has preserved it also where B and N differ: N may have, then, preserved it and B lost it; and this is the case that we are now investigating. To confuse passages in which B N stand together with those in which B stands alone, is to lend to B everywhere the weight that belongs to it only when preserving the reading of the common ancestor of it and N,—is practically to deny that any corruption has entered B in all the course of descent from the common ancestor of it and N down to the writing of the MS. itself. Conversely, to attempt to estimate B N from the known value of B (as is done by all methods of criticism that treat the MSS. separately only) is to attribute to the common ancestor of B N all the change that has entered through the many possible copyings which have taken place in the descent from it to B.

How empirical the foundations of this method of investigation are may be estimated from the fact that although, as just explained, the addition of a MS. to a group may make every difference in its value, on the other hand experience shows that it may make no difference at all. This, too, is due to the fact that MSS. agree together not by accident but by inheritance. Suppose the new MS. added is a near kinsman of those already tested, the descendant of the same common immediate ancestor or of one of the codices already in the group. Evidently, in such a case, its presence or absence will make no difference in the results of our testing process. For instance, we know that F of Paul is a copy of G3. Now, if we are investigating the value of D G of Paul, it is obvious that it is all one whether we allow F to join them or not. With or without F it is the same common exemplar that lies at the base of the group. It follows as a rule of procedure that we must take nothing for granted in using this process, but try all things, and learn the effect of each addition only by actual testing.

The practice of internal evidence of groups is thus wholly independent of any genealogical considerations. It proceeds, and must proceed, in utter ignorance of all genealogies. It tests the composite value of every combination of documents that faces it; and it is all one to it whether this combination is one which chance has thrown together or which inheritance has compacted, whether it unites in a common ancestor at once or only in the autograph itself. All it knows is, Here are documents united. All it asks is, Do they form a good or a bad combination? Yet behind internal evidence of groups the student will see genealogies clamouring for recognition. He notes the peculiarities of the groupings,—some groups frequently occurring, others, apparently equally possible, never occurring at all. He notes the verdicts of internal evidence of groups,—some groups uniformly condemned, others, apparently just like them, almost as uniformly commended. Why is it that D, the African Latin, and the Curetonian Syriac, stand so often together? Why is it that B D is so generally good, and N D so generally bad? The student would be something other than human if he did not wish to know the cause of all this. And the hope lies close that all may be explained and a new and powerful engine of criticism be put into our hands by the investigation of the genealogical affiliations of the MSS. which are suggested by these facts. The results of internal evidence of groups suggest not only the study of genealogies, but also certain genealogical facts on which that study may be begun. Every one must suspect that MSS. that are frequently in company are close of kin. Every one must suspect that the groups which support little else but corruptions are composed of the remaining representatives of a corrupt stock. Everybody must perceive that if such hints are capable of being followed out, and the New Testament documents arranged in accordance with their affiliations, we shall have a means of reaching the true text which will promise more than all other methods combined.

(c) Genealogical Evidence.

These hints have been followed out with the result of developing another method of criticism, which may be appropriately called "The Genealogical Method." This method proceeds by examining minutely all the documents representing a text, with a view to tracing out the resemblances between them and so classifying them in smaller and larger groups according to likeness. It assumes only the self-evident principle that community in readings argues community of origin, and that, therefore, a classification of documents according to their resemblances is a classification of them according to origin. If this be true of all MSS. taken together, so that we can group all New Testament MSS., for instance, together as MSS. of the New Testament by virtue of their community in the general text of the New Testament, it is, of course, true of the minor resemblances also, and we can equally safely group the MSS. into numerous subgroups, each characterised by their special readings, and each, therefore, forming a family sprung from a common more proximate origin. Community in erroneous readings is as sure a test of relationship as community in correct ones: the point is not the kinds of readings that are involved, but the communion in them. Each MS. on becoming parent of others impresses its actual characteristics on its progeny, whether these characteristics be excellences or depravities; and we may, therefore, select from the mass of MSS. the progeny of each parent, by selecting those MSS. possessing the same characterising peculiarities. The labour involved in this method of criticism, again, is no doubt very great. Every document has to be examined minutely, and compared with every other one. Those most alike are to be put together into small groups of close kinsmen; these small groups are then to be compared, and those closest to one another put together as constituting a higher and more inclusive group; these higher groups are then in like manner to be compared and grouped into yet higher groups; and so on, until we reach a point at which they all unite in one great group, inclusive of all the extant MSS. of the work, with the oldest transmitted text as their common source. The result of the labour is, however, here too, worth the expenditure. Its effect is to arrange all the witnesses in the form of a genealogical tree, and so to enable us to see at a glance the relative originality of the witness of each, — to sift out those combinations of documents which must represent only a lately originated corruption, and to trace out the combinations which will take us back to the original of all.

All this will most easily be made clear by a concrete example. Mr. Robinson Ellis finds that the MSS. of Catullus so class themselves as to admit of a genealogical arrangement which, with a little compression, we may represent thus :—

In this special instance, B, a, and b, are lost; but let us suppose for the-moment that all the MSS. marked on the plan are still in our hands. We should, then, have thirteen MSS.:—A, B, a, b, 1, 2, 3, 4, 5, 6, 7, 8, 9. Should each of these be allowed the same weight? Clearly B and 9, say for instance, stand in very different relations to the autograph, and, when the two differ, it would be manifestly unfair to allow to 9 equal weight with B. We can even go further: there is nothing legitimately in 9 which was not already in B, and if 9 differs from B, it does so only by error, and is, worthless. There is absolutely nothing legitimately in any of the codices 1—6 which is not already in a, or in the codices 7—9 which is not already in b, or in the whole array a, b, 1—9, which is not already in B. If, then, B is extant, all its descendants are useless to us; when they agree with B they are mere repeaters of testimony already in hand, and when they differ from B they are introducers of new error, and in both casee they must be absolutely neglected as useless and confusing. That B has two children (a, b) and nine grandchildren (1—9) standing by its side, while A stands alone, is at best an accident; and it is clearly unfair, on account of thia accident in copying or in the preservation of copies, to allow B twelve repeating votes to A's single voice. It is obvious rather that the whole group Bab 1—9 constitutes but one witness though they count up twelve codices, and that A by itself in point of originality balances the whole array. At one sweep, therefore, we lay aside all the codices a, b, 1—9, with all their various readings, and are enabled to confine our sole attention to A and B—the only two independent witnesses we have. This is an imaginary result in our present schedule, but in the codices of Cicero's "Orator," as worked out by Dr. Heerdegen, it actually occurs: one whole rather numerous class are codices (the codices mutili, as they are called), of for swept critical purposes into the waste-basket at once, because the source of them all, Codex Abrincensis, is still extant and in critical use.

Let us, however, come back nearer to the facts of our present case. B, a, and b, are lost, and we have just ten codices, we shall say—A, 1, 2, 3, 4, 5, 6, 7, 8, 9. How is the matter affected? If, before, B, a, b, 1—9, twelve codices, constituted but one witness, surely 1—9, nine of the same codices, have not become more than one witness by the destruction of three of their companions. This were to emulate the Sibyl and estimate value in inverse proportion to number. No more, then, in this case than in the preceding, can we allow equal weight to each codex—to A, say, and to 9. Plainly 1—9 are here combined, but one witness still, and must be counted as but one in opposition to A, which in point of originality is still able by itself to balance the whole array 1—9. Now, however, we are not able to neglect these codices; they are our only extant representatives of B, and taken together constitute B. But we must not treat them as nine separate witnesses, or even, because they obviously form two groups, 1—6 and 7—9, as two separate witnesses. We must treat them as together constituting only one witness, and we must so marshal their testimony as to eliminate the errors that have been introduced into them since B, before we match them against A. In other words, we must reconstruct B from them, and only then seek from A and recovered B their common original, the autograph. The effect of the classification on these ten codices, A, 1—9, is, therefore, to reduce the ten apparent witnesses to two, —to eliminate the large body of variants that exist among 1—6, or 7—9, as too lately introduced to merit our notice,—and so in a great number of places to fix the text absolutely.

Thus far we have proceeded as if the ten codices were found already classified to our hand. Lot us suppose, now, that they are simply handed to us as ten codices. Are we justified in assuming that each is independent of all the rest, and so beginning our textual criticism with an apparatus of ten witnesses 1 Certainly not. The fact that we receive them unclassified does not alter the fact that they actually bear such relationship to each other as is expressed in this classification. We must begin by a close examination of the codices with a view to tracing their affiliations. And, so beginning, we should note, first, that codices 1—6 are very closely alike, and that 7—9 draw likewise close together, leaving A standing apart; and then, secondly, that the group 1—6 is much more closely related to the group 7—9 than either is to A, and that the two groups contain even obvious errors (not found in A) in common. Whence it will be clear that while 1—6 come from a different proximate ancestor from that of 7—9, yet the groups unite in an ultimate common ancestor which is co-ordinate with A. This reached, the classification is complete, and we may proceed with our criticism of the text.

If we may assume that the validity and importance of the genealogical method has been thus made apparent, we may next investigate this process of criticism in its use. We have arranged our ten MSS., A, 1—9, in their genealogical relations. What have we gained as an instrument for settling the text t First of all we are enabled to attack our problem in detail. It is easier to reconstruct B from 1—9, and then the autograph from A and B, than it is to reconstruct the autograph from A, 1—9, directly. But, far above this, the classification of the codices actually gives us an instrument of criticism that settles much of the text of B, or even of the autograph, for us at a glance. For example, if one reading is supported by 1, 7, 8, 9, while 2, 3, 4, 5, 6 each give a divergent reading, it is clear beyond a peradventure that the first stood in B. For this combination of documents, 1 + 7, 8,9, cannot occur unless 1 inherits from a, and 7, 8, 9 from b, exactly the same reading, which, because in both a and b, must also have been in B. Again, if 1,2, 3, 4, 5, 6 present one reading, 7 another, and A, 8, 9 another, this last with absolute certainty must have stood in B and in the autograph. For 8, 9 cannot agree with A except by having inherited this reading from their common ancestor, and this involves its presence throughout the whole line of descent—i.e., in b and in B; it was, therefore, the reading of both A and B and of their common ancestor, the autograph. In cases of simple genealogy, therefore, the rale is obvious and exceptionless (in all such cases as cannot be accounted for as merely accidental conjunctions) that attestations including documents from two groups demonstrate the presence of the reading so attested in the common parent of these groups. All readings supported by A and any descendant of B (accidents excluded) were consequently in the autograph; all supported by any descendant of a and any descendant of b in common (accidents excluded) were in B. So far our results are certain. When A and restored B agree, the reading is, of course, that of the autograph. When they differ, in a case like the present, where we have but two primary witnesses, we are thrown back on the character of the witnesses to determine the probability of rectitude between them. Hence, we call in "internal evidence of classes," as we shall call it, to distinguish it from the same process when dealing with chance groups, instead of, as here, genealogically determined ones. In other words, we collect the various readings between A and the group 1—-9 considered as a unit, and that is as much as to say B, and try the relative value of the two by internal evidence, just as we did in the kindred processes of internal evidence of documents and internal evidence of groups. The class which supports the greater proportion of approved readings is the better class. Had we three primary classes instead of two, this process would need calling in only in cases of ternary variation; whenever there were two classes arrayed against one, the reading would be settled on purely genealogical grounds.

The essence of this whole procedure may be reduced to two simple rules: (1) First, work out a complete classification of the witnesses to any text by means of a close study of their affiliations, and thus determine how many independent witnesses there are; and (2) Then by internal evidence of classes determine the relative value of these several independent classes. When these two processes are completed we have a method of criticism available which will, in all cases of simple and unmixed genealogies, carry us with the greatest certainty attainable to the text that lies behind all extant witnesses.

The limitation "in all cases of simple and unmixed genealogies " was not unintentionally introduced into the last clause. Normally we may expect each document to be made simply and without intentional alteration from a single pre-existent document; and when this has been the actual course that has been taken, all documents, each having a single parent, arrange themselves in a simple genenlogy. It is possible, however, that a given document may not be thus simply copied from a single exemplar, but may have two or more parents. The scribe may place two copies (which may as well as not be of different types) before him, and make his new copy by following now one, now the other, either capriciously or with a conscious effort to act as editor. Or again, a scribe accustomed to a strongly marked type of text, when called upon to copy a codex of another type, may consciously or unconsciously allow his teeming memory to introduce into the new copy readings drawn not from the exemplar before him, but from the type of text to which he has been long accustomed. The result, in either case, is a document which is not a simple copy of a single exemplar, but which rather will be more or less intermediate between two types, and will therefore refuse to take its place in any scheme of simple or

unmixed genealogies. There is yet a third way in which this "mixture," as it is technically called, is introduced into texts, and this is doubtless the way by which, in actual fact, most mixed texts have been formed. The student will remember that it was customary of old time, more or less completely, but usually very incompletely, to correct codices in the text or margin by other codices with which the owner chanced to become acquainted. All of our great codices have been so corrected, and often the process has been repeated several times. Thus we distinguish between N, N", Nb, and between B, Bs, B3, etc. Now, suppose a codex which has been thus corrected by a divergent type of text to be used as copy for the production of other codices. The scribe does not know what corrections are merely marginal readings and what are really corrections; he inevitably adopts some or perhaps all of them into his text as he writes it out. And the result is a "mixed text," having for its parents the original codex and all the divergent codices, readings from which had been written on the margin. A very interesting example of such a mixed text is furnished in Codex E of Paul,—Codex Sangermanensis. This MS. is recognisably a copy of the Codex Claromontanus (D2), but it does not give the original text of D, but that text as corrected by the several hands which had diligently ornamented its margin with readings from other codices. The result is that E is a mixed text. Of course, if the corrections had all been taken from a single simple codex, and the correcting had been thoroughly done, and the scribe in copying from the MS. nad noted and adopted them all, the result would not have been a mixed text, but a text of the type of the document to which the original had been conformed. But this completeness is not to be expected, and the result is, therefore, always a more or less mixed text.

Now, it is obvious that the effect of mixture is to confuse genealogies. Wherever it has entered, and in the proportion in which it has entered, the arrangement of the documents in their true genealogical relations is rendered difficult, as also the interpretation of the evidence, after it has been arranged. The detection of the fact of mixture is generally, however, easy, and when it is once detected it can be allowed for; so that it will only force us to apply genealogical evidence with more care and discrimination, rather than render it inapplicable. Suppose, for instance, that in undertaking to determine the mutual relations of a body of five witnessing documents, we find that they separate easily into two pairs, each a representative of a marked type of text, while the fifth witness is intermediate between the pairs. Whether this intermediate position is due to mixture or not is usually possible to determine by the character either of the intermediate readings themselves or of the whole mass of readings furnished by the intermediate witness. If any of the readings are themselves composite readings, uniting the readings characteristic of the other types—" conflate readings" as they are called—and especially if many such readings occur, mixture may be assumed to be proved. If, again, in looking over the whole mass of its readings we find the intermediate witness to follow arbitrarily first one and then the other of the two pairs in their obvious errors, and especially if this is true of the obvious errors of a separate document from each (or either) pair, while its own obvious errors can be traced back by transcriptional evidence with equal arbitrariness now to the one and now to the other, mixture again may be assumed. The fact of mixture having been thus determined, it may be allowed for, and the elements in the witness under investigation be separated and placed in the genealogy accordingly.

Some such state of things as we have thus assumed seems actually to occur in the witnessing documents to the "Two Ways," or first section of "The Teaching of the Apostles," the scheme of which is apparently as follows :—

Original Text.

Here the extant witnesses are a, b, forming one pair, and 1, 2, forming another, together with 3, which proves to be a descendant of a lost d mixed with 2. A glance at the table will show the effect of the mixture. Without it, the combination 2 3 would necessarily determine both what was in c and d, and hence what was in B. But owing to mixture of 3 from 2, the combination 2 3 may be only a corrupt reading peculiar to 2; and 1 may preserve the true reading of c, while the reading of B may be that of c now extant in 1, or the lost one which stood in d before mixture with 2 displaced it from its descendant 3. So, again, without mixture, such a combination as b 1 against 2 3 would have been impossible. For b and 1 could not agree (accidents apart), unless this reading had been inherited from their common ancestor, and this would imply its presence in all the links between that ancestor and each document—i.e., in A and in B and in c. But, again, 2 and 3 could not agree unless in like manner that reading stood in every link between each and their common ancestor—i.e., in d, c, and B. Thus both readings would have to stand in B and in c as well to allow this division of evidence. With the mixture, however, this combination is very possible; for though b 1 implies that the reading so supported stood in c and B, 2 3 need not imply anything beyond the presence of its reading in 2 itself, whence it may have been borrowed by 3. A division or attestation of this kind is called a "cross attestation," and "cross attestations" are among the surest proofs that mixture has taken place. Go back to the diagram from Catullus, for instance. If we find A, 1, 2, 3,4, 5 supporting one reading, and 6, 7, 8, 9 another; or A, 8, 9 one, and 1, 2, 3, 4, 5, 6, 7 another; or 1, 2, 3,4, 5, 7 one, and 6, 8, 9 another,—we may be certain (accidents being excluded) that mixture has taken place. For each of these divisions is such as cannot occur in a simple genealogy, inasmuch as it springs across from one group to another, and hence presupposes that its reading was in the parent documents.

The effect of mixture, then, on genealogical evidence is to limit the sphere of its application. Thus, in our present illustration, we no longer know at sight what 2 3 means. It may be c + d, and hence carry us back to B, or it may be only 2 + 2, and so leave us at 2. Even 123 may be nothing but a corruption introduced by c. In all cases in which A and B differed, 13 is the only combination that we can be sure will take us back to B. But mixture does not affect the validity of genealogical evidence wherever it can be applied. Thus, again, in our present illustration, a (or b) 1, or a (or b) 2, or a (or b) 3, all alike carry us back to the common original of all our witnesses despite the mixture of 3 from 2, and in general every combination of a or b with a descendant of B still settles the original text with certainty. We gain somewhat fewer results from genealogy than we should have attained, had there been no mixture; but what we do gain are equally sound in this case as in that. The actual instance of mixture which we have been studying is no doubt a very uncomplicated one. It sufficiently illustrates, nevertheless, its effect, its dangers and its difficulties; and the most complicated case imaginable would differ from it only in degree. The one principle that unties, as far as may be, all the knotty problems that mixture sets for us, is that mixture acts simply like marriage in real genealogies, and we must allow the possibility of each combination of documents, into which it enters, meaning as many diverse things as there are diverse ways of tracing up their inheritance to a common original. Thus, the common original of 2 3 may be found at 2, or if it is the other element of 3 that here unites with 2, not until we reach B.

As mixture operates in a directly opposite direction to pure genealogy,—tending to bring together whereas it tends to separate the texts, to compress all lines of descent into one composite line whereas it broadens them out more and more, like a fan,—it is not strange that it introduces some paradoxes into.criticism. One of these it is worth while to call attention to. Where mixture has been at work, it is often discovered that a group is weakened instead of strengthened by the addition of other witnesses. For example, in our illustration, 1 3 is a strong group; its readings must take us back at least to B, the common original of this whole class. Add 2 to this group and at once its value is lowered. For 1 3 (2 dissenting) must be a combination of 1 descended from c and of 3 in that part of it which descends from d, inasmuch as the dissent of 2 proves that this is not the part of 3 that comes from 2. But 1 2 3 is a combination of 1 and 2 descended from c and 3 in a part that may well have been borrowed from 2, and hence which also may descend from c. Hence, while 1 3 must be at least B, the larger group 123 may mean only c, and is therefore a weaker group. Analogous findings crop out in the New Testament. For example, internal evidence of groups proves that B D in Paul is a better group than B D G, or than B D G -f- most uncials and most minuscules. Again, N A C in Paul is a better group than N A C D G. The explanation of it lies in this: some mixture has taken place that makes B D analogous to 1 3 in our diagram and B D G analogous to 1 2 3.

The application of genealogical evidence to the New Testament has proved to be exceptionally difficult. Not only has the critic to face here an unheard-of abundance of matter, all of which has to be sifted and classified; but the problem is complicated by an unparalleled amount of mixture, which has reigned so universally that it has left scarcely a half-dozen witnesses entirely unaffected by it. The task of working out the genealogy of the New Testament MSS. has, therefore, been the labour not of one man, nor of one age, but of a succession of generations. The first dim signs of classification were mistily seen by Mill (1707) and Bentley (1720); the genius and diligence of Bengel (1734) and Griesbach (1775—1811) drew the lines of division with some sharpness; and Dr. Hort, in our own day, has at last so far perfected the details that this method of criticism can now be safely used for the settlement of much of the New Testament text. The multifarious abundance of mixture in our witnesses complicates and limits the use of genealogy sadly; but, as elsewhere, leaves the soundness of its results unaffected wherever it can be applied. Genealogy, thus, does less for us in the New Testament than could have been hoped, but it does much for us nevertheless. In particular, the results attained by it so fully explain those reached by internal evidence of groups, which it is to be remembered is an entirely independent process, and those attained by that process so fully accord with those attained by this, that the two methods actually prove the soundness of each other, and place the text obtained by both combined in a very unassailable position.

It does not fall within the plan of this primary treatise to enter fully into the details or the justification of the genealogy which Dr. Hort has worked out for the New Testament witnesses. For this the student must be referred to the full exposition and proof which Dr. Hort has himself given in his epochmaking "Introduction" to the Greek Testament, which was published by Dr. Westcott and himself in 1881. Here, it must suffice to set forth only so much as will enable the beginner to make intelligent use of the method.

At the root of all genealogical investigation lies \ the classification of the documents according to their I affinities; and Dr. Hort has shown that the documents representing the text of the New Testament part into four great and well-marked classes, which he would somewhat conventionally designate the Syrian, Western, Alexandrian, and Neutral. Next the difficult "problem of the relation in which the several classes stand to one another is unravelled. And here, first, it has been shown that the Syrian/ class is not an independent witness to the text of the New Testament, but is rather the result of a critical editing of the New Testament text which was accom plished probably in Syria at some time not earlier than the last half of the third century. The evidence that proves this is of three kinds. First, the distinctive readings of the Syrian text, although common

in the later fourth century and all subsequent fathers, cannot be traced in ante-Nicene patristic quotations; so that, journeying backwards in time, the favourite text of Chrysostom and his age has disappeared entirely from use by the time we reach Origen. Secondly, the distinctively Syrian readings, when tried by internal evidence, betray themselves as inferior to, and, when tried by transcriptional evidence, as derived from, those of the other classes. And, thirdly, this culminates in the presence among the Syrian readings of a body of "conflate readings," the simple elements of which occur in the other classes, so that it is certain that in some of its parts this text was made out of the Neutral and Alexandrian, or the Neutral and Western, or the Alexandrian and Western. When all the phenomena are closely scrutinised, it is made out positively that the Syrian text was made by a revision out of the other three classes, and preserves nothing from antiquity not already in them. In the presence of the other three classes its testimony is, therefore, collusive testimony, and is simply to be neglected. The case with reference to it is precisely similar to that with reference to the codices mutili of Cicero's "Orator," or the printed editions of the New Testament. We should have much the same warrant for introducing Westcott and Hort's Greek Testament among our witnesses that we have for introducing the Syrian text; in both cases the valuelessness of the text as a witness-bearer depends on the fact that it represents not testimony—i.e., inheritance, but the opinion of editors—i.e., revision. Setting aside, then, the documents containing the Syrian text, we are left with only three classes representing the New Testament text. That the Western class is an independent class is easily proved; and its character is so strongly marked that it stands quite apart from all other types. The Alexandricin is more difficult to deal with. Although there is much that would lead us to assign an independent position to it, too, on the whole it seems to be the truer disposition to join it with the Neutral, and arrange these two as two great subclasses of a greater class, including them both and standing over against the Western. With this disposition, the New Testament genealogy will have a form of descent worked out for it which is very closely analogous to that for Catullus, which we have used as a sample genealogy; and it may be graphically represented as follows:— Original Text.

Western Text.

Neutral Text. Alexandrian Text.

Had no complications of mixture entered into the descent of the various documents which at present represent these three classes, this genealogical scheme would teach us that a combination of the Western text with either the Neutral or Alexandrian would necessarily take us back to the common original of all. On the other hand, wherever each text appeared a,s sponsor for a different reading, or the Neutral and Alexandrian stood opposed to the Western, the bearing of the external evidence could be settled only by calling in internal evidence of classes. This last named process proves to speak with no doubtful voice. It condemns the Western text as the most corrupt of all known forms; it commends the Neutral as the most correct of all forms; and it assigns a character somewhat intermediate between the two to the Alexandrian. The observed characteristics of the various classes account for this verdict. The licence which teems to have characterised the scribes whose copyings formed the Western text may be almost described as audacity: paraphrase, assimilation, modification, elaboration, extensive interpolation, abound everywhere, and result in the most corrupt text which has ever been current. The Alexandrian text is characterised rather by workmanlike and even scholarly corrections of forms or syntax, and petty modifications, which might easily creep in where the scribe was also partly editor. While honest and careful copying, with only the intrusion of the errors incident to all copying, seems to be the characteristic of the Neutral text. The Syrian text, formed on the basis of these preceding types, appears to have been an effort to replace by a purer and smoother text the corrupt Western type, which had been at that time, for probably a century at the least, practically the Textus Seceptus of the Christian world. As such it was eminently successful; and gave to the Church for the next millennium and a half a textus receptus that is practically free from the gross faults of the Western text, that is noble and attractive in form and worthy in diction, and peculiarly suited for the cursory perusal of the closet or reading-desk. Considered as a representative of the New Testament, it is competently exact for all practical purposes; considered as an effort to reform a corrupt textus receptus, it is worthy of great admiration when the narrow opportunities of the time when it was made are kept in view; but, considered as a witness of what was in the original New Testament, it passes out of court simply because it is a good editoriallyframed revision of the text, and not a simple copy of it.

It will scarcely need repeating at this point, however, that mixture, so far from being absent from, has been specially active among New Testament MSS. To such an extent has it ruled, that we have perhaps only four codices that have escaped it altogether, to which may possibly be added one version. Codex B in the Gospels, Acts, and Catholic Epistles (not in Paul), seems to be purely, or all but purely, Neutral; D, D2, G3, seem purely Western everywhere, and to them may possibly be added the African Latin version. No extant document presents an Alexandrian text unmixed; both Western and Neutral admixtures have entered even C, L, A (in Mark), and the Memphitic version, the most constant representatives of this type of text. It follows, therefore, that a combination of the Western and Alexandrian documents need not be a combination of these two texts, and therefore will not overbear the testimony of the Neutral class; and internal evidence of groups proclaims the Neutral usually the better reading in such cases. To B, D, D2, and G3 there need be added only some small fragments such as T, H, to complete the list of New Testament MSS. which have not received mixture from the Syrian text. -iB has a Western element in Paul's epistles mixed with its Neutral base, but apparently has nowhere received Alexandrian admixture. N has a Neutral base, but has received both Alexandrian and Western elements by mixture, although these elements are unequally distributed, being most abundant in the Gospels (especially in John and parts of Luke), and apparently in the Apocalypse, and least abundant in Paul. Among the versions the African Latin seems purely Western, and the Curetonian Syriac predominatingly so; while the Memphitic and Thebaic, though betraying some Syrian admixture in their extant forms, were originally probably Neutral-Alexandrian with a Western admixture,—largest in the Thebaic. All other documents have a larger or smaller Syrian element, and thus present very composite texts. A is fundamentally Syrian in the Gospels; but in the other books has only a Syrian admixture on a base fundamentally Neutral, with Western and Alexandrian elements (the latter especially in the Acts and Epistles). L is AlexandrianNeutral with Western admixture. A is fundamentally Syrian (probably as copied from a MS. fully corrected by a Syrian codex) everywhere except in Mark, where it is very largely Alexandrian-Neutral. Among the codices which have a Syrian element such MSS. as C, L, P, Q, 11, Z, T, A (in Mark), 33, 81 (= 2^), 157 in the Gospels, A,C,E, 13, 61 in Acts and the Catholic Epistles, A, C, M, H, P, 17, 67** in Paul, and A, C, P in the Apocalypse, preserve the largest proportion of pre-Syrian readings.

The effect of this state of things on the genealogy of the MSS. of the Gospels, say, for example, may be roughly represented to the eye by the following diagram, which does not aim to arrange the MSS. in anything like their actual relations to one another, but only to represent in the simplest way the general effect of mixture.

Original Text.

A few of the symbols of actual documents have been (very approximately) introduced into this diagram, in order to give point to its lessons. The letters w, n, and a are intended to represent respectively the Western, Neutral, and Alexandrian classes, each of which originated, of course, in a single copy, although it must be remembered that the peculiarities of each class grew progressively more and more marked, and took time and many copyings thoroughly to develop. In the lines of descent from w, n, and a, the single letters variously primed—e.g., w1, wm, n1, nlv, a1, aT—are intended to represent unmixed descendants, while the ordinary genealogical sign of marriage (=) is used to represent the union of two documents for the production of a third, the more or less composite character of which is indicated by the combination of letters which represents it,—e.g., wa, wn, an, wan, waan, waann, etc.

Now, the essence of the genealogical principle is that any combination of documents has weight in proportion to the distance from the autograph of the point in the genealogy at which the lines of descent of this combination unite. Assuming that the documents X, B, C, D, L, Old Latin, Memphitic, have been justly placed in the genealogy, it is possible to estimate the value of each combination of these documents by tracing them out in the table. For example, the line that connects B with the autograph and the line that connects D with the autograph do not come together until they reach the autograph itself; accidental conjunction in obvious corrections or unavoidable corruptions apart, therefore, the combination B D should be equivalent to the original text itself. On the other hand, since N traces back to the autograph through three different lines—viz., through w, n, and a—a combination of it with any other document, whether a Western one like D, or a Neutral one like B, or a prevailingly Alexandrian one like C. may, indeed, be a combination of classes, and so take us to their union; or it may be only a combination of documents within one class, and take us only to w, or to n, or to a. The combination D N, for instance, may be a combination of Western D with N in its Western element, and so take us only to w; or it may be with N in its Alexandrian or Neutral

element, and so take us to the original text. It will be remembered that the Western element in N is particularly large in the Gospels; hence D N here is apt to be only a combination of two Western witnesses; we shall not be surprised, therefore, to note that internal evidence of groups usually condemns this group. For the same reason, however, the combination B N, which might carry us equally easily to n, to X, or to the autograph through N's Western element, is most apt to do the latter; and herein we see the reason why internal evidence of groups gives such high character to B N. Let these instances suffice. The student will readily see that the genealogical evidence proper needs only supplementing by internal evidence of classes, by which we learn that w is a very corrupt and n a specially good line of descent, to make this distribution of the New Testament documents into their proper classes a very valuable engine of criticism.

The relative divergence of the three great classes from the line of pure descent is not illustrated by the diagram, and therefore it tells us nothing of the results obtained by the important process of internal evidence of classes. Perhaps even this may be roughly represented to the eye by a diagram of the following form. If x y be taken to represent the line along which all documents would have been ranged, had an absolutely pure descent been preserved and no errors introduced, z q may be taken to represent the actual line of descent which the Western documents have taken, k v that of the Alexandrian, and t s that of the Neutral; while w p will represent the line of

descent of the Syrian class. Along the line z q may be placed, therefore, the Western documents, each later one representing a greater divergence from the true text; along k v the Alexandrian documents, and along t s the Neutral ones. As N and C L are mixed, they may be assigned a more or less intermediate position, with dotted lines connecting them with their several sources. It is evident that the combination of any two documents will take us to the point in the descent of the text where their separate .^^ * * _ _* Tntjejet.

descents coincide. B, standing just beyond t on t s, is nearest the true text of all single documents. The two lines of B's and of D's descent can unite, when traced back, only at z, on the line of true descent, and at a point very far back in time, N draws a contingent from the Western text, and hence N D may only take us to some place on z q; it also draws an element from the Alexandrian text, and hence N D may take us to z on the line of true descent; and it also draws an element from the Neutral text, and hence again N D may take us to z on the line of true descent. Which of these is the true account can ba told in general only by internal evidence of groups, although in particular instances it may be discovered from the nature of the opposing party. For example, if N D stands opposed to B C L A in a passage in Mark, we can argue that the element of N represented here is neither the Neutral element (else would it stand with B), nor the Alexandrian element (else would it agree with C L A), but the Western element; and hence N D is here Western, and takes us only to some point on z q, off of the true line of descent.

This exposition of the genealogical method has been but little successful unless it has shown, along with the nature of genealogies in general, somewhat also of the effect of mixture on the genealogies of the New Testament, and of the methods that must be adopted to overcome the difficulties raised by it. There remains, therefore, only to give a more extended list of the documents which represent each class before we can proceed to study the application of this method to practical use. Let the student only remember that we must treat, here too, each section of the New Testament separately, and that by reason of mixture a single document may find place equally well in more than one class, and the following list will be useful to him.

The Neutral text is more especially represented by the following documents, viz. :—In the Gospels: B (purely), N largely, and then T, S, L, 33, A (in Mark), C,Z,R, Q, P, Memph. (Theb.) (Syrhier). In the Acts and Catholic Epistles: B (probably purely), N, 61, A, C, 13, P (except in Acts and 1 Peter), and less degree in A, C, E. In Paul: K, L, N, also in H, I2 M, O, Ob, Q, R, and in less degree in A, C. In the Apocalypse: B, and in large part P, and in less degree C, A.

The post-Nicene fathers generally, present a Syrian text in their citations, although Cyril of Alexandria, Apollinaris (xara /xepos Trams), and less markedly Epiphanius, and even John of Damascus, are to greater or less extent exceptions to this rule. The ante-Nicene patristic citations are prevailingly Western; this is true of those of Marcion, Justin, Irenseus, Hippolytus, Methodius, Eusebius, and even to some extent of Clement of Alexandria and Origen. A large nonWestern pre-Syrian element is found, also, however, in the Alexandrian fathers, Clement of Alexandria, Origen, Dionysius, Peter, and also in a less degree in Eusebius and others.

The ready application of the genealogical method to practical use in criticism will depend on our ability to read the digests of readings, where the evidence is expressed in terms of individual MSS., in terms of the classes of MSS., or, in other words, to translate testimony expressed in terms of individual MSS. into testimony expressed in terms of classes of MSS. The proper procedure may be tabulated somewhat as follows :—(1) First, sift out all Syrian evidence from the mass of witnesses recorded in the digest, and thus confine attention to the pre-Syrian testimony. If, on sifting out the Syrian evidence, only one reading is left, it is, of course, the oldest transmitted reading, and as such is to be accepted. (2) Next, identify the pre-Syriaa classes, Western, Alexandrian, and Neutral,

by separating the chief representatives of each from the body of the witnesses, allowing everywhere for mixture. (3) If, now, we have three readings, one supported by each of the pre-Syrian classes, the Neutral reading should have the preference. (4) If we have only two readings, that supported by the Neutral and Western against the Alexandrian is to be preferred; or that supported by the Neutral and Alexandrian against the Western is to be preferred; or (since all prominent Alexandrian documents have a large Western element) that supported by the Neutral against the Western and Alexandrian is to be preferred.

A few examples are needed to illustrate practice under these rules. The sifting out of the Syrian evidence is rendered necessary by the relation which the Syrian class bears to the others as dependent on them and made out of them, by which its evidence is made collusive and confusing. It will be sufficiently accurately accomplished at first by confining attention to the following documents, viz.: in the Gospels: X, B, C, D, L, P, Q, R, T, Z, A (in Mark), H, 33, Latin versions, Curetonian and Jerusalem Syriac, Memphitic, and Thebaic; in Acts, N, A, B, C, D, E, 13, 61, and the same versions (except the Curetonian Syriac, which is not extant here); in the Catholic Epistles, N, A, B, C, 13, the Latin versions, Memphitic and Thebaic; in Paul, N, A, B, C, D, G, 17, 67**, and the same versions; and everywhere the certain quotations of the anteNicene fathers. Any reading which has the support of no one of these witnesses may be safely set aside as Syrian or post-Syrian; and even if a few of these witnesses which contain a large Syrian element join with the mass of later witnesses against the body of those named here, the reading may still be safely neglected as Syrian. Not infrequently the reading is settled by the sifting out of the Syrian documents: when they are removed, the variation is removed too. An instance may be found in Mark i. 2, where " in the prophets" is read by A, E, F, G, H, K, M, P, U, V, r, II, many minuscules, the text of the Harclean Syriac, the Armenian according to Zohrab's edition, the -5Sthiopic, and some late fathers, including the Latin translation of Irenseus in opposition to the Greek elsewhere. Only P in this list occurs in the test list given above, and the whole support of the reading is, therefore, distinctly Syrian, so that when the Syrian testimony is sifted out we have left only "in Isaiah the prophet," supported by the whole pre-Syrian array—viz., B N 33, L A, D, about twenty-five minuscules, the Latin versions, the Memphitic, Peshitto, Jerusalem, and margin of the Harclean Syriac, the Gothic, and codices of the Armenian versions, with Irenseus and Origen among the fathers. In like manner the addition of lv ria if>avepia in Matt, vi. 4 and 6 is sifted out with the Syrian testimony, leaving the whole body of pre-Syrian witnesses at one for its omission. In such cases our work is oasily done, and the text is restored with the very greatest certitude. Any reading supported only by the Syrian class is convicted of having originated after A.d. 250.

Often, however, we seem no nearer our goal, after tte Syrian evidence has been sifted out, than we were at the start. Two or sometimes three readings may still face us, and our real task is yet before us. The next step is to identify the classes represented in the groups of witnesses supporting each reading, by attending very carefully to their constituent elements, whether pure representatives of any one class or mixed representatives of more than one. This is often a very delicate piece of work, but it is often also easy, and is generally at least possible. It is usually best to begin by identifying a class of which we have pure representatives, and to proceed thence to those the only extant representatives of which are mixed. In the Gospels it is nearly equally easy to identify the Neutral and the Western readings; in Paul we should begin with the Western; in Acts and the Catholic Epistles, again, we may almost equally well begin with either the Western or Neutral. Let us look at Mark iii. 29 as an example. Here the reading "judgment" sifts out with the Syrian testimony, and we are confronted with the pair of readings d/tapi-i^ai-os supported by N, B, L, A, 28, 33, 81 (= 2Pc), and a^apnas supported by C*"d, D. 13, 69, 346, Ath. The versions here can give but little help, and we omit them altogether. We note at once that purely Western D is united with a small body of adherents, all of which have Western elements, in support of djua/m'as, which we may thus recognise as Western. On the other side, the purely Neutral B stands in the midst of a group which therefore certainly embraces the Neutral class. Whether a/tapr^/xaTos is also Alexandrian is more doubtful, inasmuch as the Alexandrian documents supporting it have all Neutral elements. On the whole, however, this reading may be safely set down to the credit of both the Alexandrian and Neutral classes. But in either contingency internal evidence of classes determines for it as probably the true reading. A similar example may be found in the vivid insertion of To in Mark ix. 23, which has the support of B N, C L A, X T, involving the typical Neutral and Alexandrian witnesses against the omission by D, 13, 28, 69, 81 (= 2Pc), 124, 131, which is recognisably Western. In the next verse (ix. 24) the /hctoi ScucpvW is in the same way recognised as Western, supported as it is by D, N, X, r, the European, Italian and Vulgate Latin, Peshitto and Harclean Syriac and Gothic versions, while its omission is testified by B N, C* L A, 28, k of the African Latin, the Memphitic, Armenian and .ZEthiopic versions—i.e., by the combined Neutral and Alexandrian witnesses. A considerable insertion of the Western text is found in Mark ix. 45 and 46, supported only by D, N, X, F, Latin, Syriac, Gothic and ^Ethiopic versions, while the omission is supported by B N, C L A, 1, 28, 81(= 2i»), 118, 251, k of the African Latin, Memphitic, and Armenian. On the same kind of evidence Mark ix. 49, last clause, and xi. 26, are recognised as interpolations of the Western text. In all these cases we have proceeded by identifying and rejecting the Western reading, and the help in determining the text has been sure and immediate.

In such a reading, on the other hand, as the addition of /Vij/ta in Matt. v. 11, which is witnessed by C, r, A, Peshitto and Harclean Syriac, and Origen, against N, B, D, Latin, Memphitic, Jerusalem Syriac, and ^Sthiopic versions, and Cyril of Alexandria; or such an one as the addition of Tois dpxi"ots in Matt. v. 27, by L, A, 33, later Latin, Curetonian Syriac and Harclean Syriac versions, Irenseus, and Eusebius, against B N, DP, African and European Latin, Memphitic, Peshitto Syriac, Armenian, jSJthiopic, and Gothic versions, and Origen; we must proceed by identifying and rejecting the Alexandrian reading, which appears to be opposed by the combined Neutral (B, N, etc.) and Western (D, etc.) witnesses. In such cases the Alexandrian reading is identified by a process of exclusion: for example, in the former case C, A, are not Neutral, for they separate from the Neutral documents, and they are not Western, for they separate from the Western documents; they must be, then, either Alexandrian or Syrian, and the presence of the reading in Origen seems to point to the former. In these cases, too, the reading is settled securely by the combination of Western and Neutral witnesses.

Still another class of variations may be illustrated by the insertion or omission of " which art in heaven" at the opening of the Lord's Prayer in Luke's account of it. The insertion is supported by the Syrian text, and also by D, C, T, A, X, 33, etc., Old Latin codices, Curetonian, Peshitto and Harclean Syriac, Memphitic, and ^Ethiopic; and the omission by B, N, L, 1, 22, 57, 130, 346, Vulgate Latin, and Armenian versions, Origen and Tertullian. The Neutral text certainly is for omission (B, N, etc.), and the Western for insertion (D, Old Latin, Curetonian Syriac). But representatives of the Alexandrian text are on both sides: N, L, 1, 57-, on one, and C, T, A, X, 33, Memph., on the other. If we could be sure that this latter group represented the Alexandrian here, its union with the Western would carry our decision with it; but every single member of it is so strongly mixed with Western readings that it would be dangerous in the extreme to count it anything but Western here. So that we can only believe that we have here a case of Neutral versus Western, and follow the former accordingly. As for the Alexandrian reading, it is either lost or else represented by L, 1, 57. Internal evidence of groups not only supports this conclusion, but forces it upon us. Quite similarly " Let Thy kingdom come, as in heaven, also on the earth " is inserted at the end of the same verse by N, C, F, A, X, D, Old Latin, Peshitto and Harclean Syriac, Memphitic, and ^35thiopic, against the protest of B, L, 1, 22, 130, 346, Vulgate Latin, Curetonian Syriac, Armenian, and Origen and Tertullian. The transference of X, which has a very marked Western element in Luke, makes no essential difference in the testimony; every codex arrayed here with D has a large Western element, and the whole combination is explicable as a Western inheritance. So that again we treat the matter as an instance of Western versus Neutral, and decide accordingly, by internal evidence of classes, for the Neutral. A special but very small class of readings, called by Dr. Horfc "Western non-interpolations," deserves a separate notice. An example may be found in the odd insertion into Matt, xxvii. 49, to which attention was called when we were speaking of internal evidence of groups. The insertion is supported by N, B, C, T, U, T, 5, 48, 67, 115, 127, ^Ethiopic,—including the Neutral (n, B, etc.), and Alexandrian (0, L, F, 5, 48, 67, etc.) witnesses. The omission has the support of only D, A, E, F, G, H, K, M, S, V, A, II, most minuscules, the Latin, Peshitto and Harclean Syriac, Memphitic, Gothic, and Armenian versions, and the like,—which are easily seen to be Syrian and Western. Yet, as already pointed out, internal evidence of readings seems to forbid our accepting these words as genuine, and thus forces us to decide against the combination of the Neutral and Alexandrian and for the Western standing alone. In this reading, and possibly in some others like it (for each must be treated apart), we have the exception to the general rule that the Neutral-Alexandrian class is better than the Western, which the genealogical scheme on which we are working allows for and hence presupposes. If the Neutral and Alexandrian have been rightly accounted two branches of one stem set over against the Western, it would be difficult to understand how it could happen that the Western should be always wrong, without exception, and this stem always right. The process of internal evidence of classes, like internal evidence of groups and documents, determines only general and usual relations, and the exceptions to the general rule can be detected only by internal evidence of readings. If, for the moment, we conceive of the line xt in the last diagram as not the line of absolutely true descent, but the actual line of descent of codices, from which zq diverges when the descent becomes Western, k v when it becomes Alexandrian, and t s when it becomes Neutral, it will be evident to the eye that the Neutro-Aloxandriau descent coincided for the space represented by z k, after the separation of the Western descent had taken place, and hence it is to be expected that the combination Neutral-Alexandrian will testify to some errors introduced into their common stem during the series of copyings represented by the space z k. In other words, reverting to the former diagram, the very fact that the Neutral and Alexandrian classes are arranged, not as two independent classes co-ordinate with the Western, but as two sub-classes of X, which is coordinate with the Western, presupposes that they will combine against the Western in some errors. From all which we learn that textual criticism, even with the aid of the genealogical evidence, cannot, any more than in the case of other methods, be prosecuted mechanically; but each reading must be very carefully considered, separately, ere our conclusion concerning it be announced.

Procedure under the genealogical method in Paul's Epistles has enough of speciality to render it desirable to give some illustrations of it. It is a good practical rule to go by in the Gospels, to follow the group which contains B, at least provisionally. The best practical rule to go by in Paul is, to suspect the group which contains D, G, unless practically all the primary witnesses join with them. This difference of procedure results from the fact that B is purely Neutral in the Gospels, and hence forms there the rallying point for the documents of the best class to gather around. In Paul B has a Western element, and hence may stand with only Western documents—the worst class— around it. With no pure representative of either the Neutral or Alexandrian class, we are reduced in Paul to identifying, as our first step, the Western class by the aid of its pure representatives D and G, and this we identify only to reject, if it stands alone. And as all codices have a Western element, it follows further that any addition to D G need not alter its character as Western and probably corrupt. Hence A D G, B D G, N D G, C D G, or ABDG, A N D G, A C D G, BCDG, CDG, alike, need represent nothing better than a Western error. No d, priori reason exists why B N D G might not equally do so; but internal evidence of groups here steps in and proclaims this group so good that we are obliged to account it usually a union of Neutral (BN) and Western (DG) classes. This only shows that B and N, although both having Western elements, get their Western elements independently, and do not usually coincide in the same Western corruption; hence, while thoroughly consistent with the genealogical scheme, this finding is inconsistent with the supposition that these two codices come from a proximate original only a step or two older than themselves. The larger combinations, even, such as A C N D G, or A B C D G, may still be merely Western; and we are thus led to give the preference, on genealogical grounds, often to small groups which include only one or more primary uncials when opposed by a group including D G.

As an example, \ve may look at 2 Cor. ii. 9, where after the Syrian evidence is sifted out, we have ei, read by N, C, D, G, P, Latin versions, whereas rj is the reading of B, A, 17, 109. Here, although all the recent editors read el in their text (Westcott and Hort placing -g in their margin), the genealogical evidence is distinctly in favour of rj, the group N C D G P being distinctly Western. It may be added that the transmutation of ij into et either by itacism (ij, t, «,) or by mistake of the uncial letters (ei for H) is very easy and frequent: a case of it occurs in the neighbouring 2 Cor. iii. 1, where d py is read by A, P, and Syrian authorities, while T/ /at) stands in N, B, C, D, G, 31, 37, 67**, Latin, Memphitic, etc. Here we have a combination of the Neutral and Western at least, if not of all pre-Syrian classes against Syrian or possibly Syrian and Alexandrian, and easily follow this group even though it contains the ominous D G, since along with D G stands N B C, which is differentiated from other groups including D G, by a very emphatic verdict of internal evidence of groups. The complications that can arise by dividing the testimony a step further are well illustrated in 2, Cor. ii. 7, where jj.S.\\ov is placed before v//,Ss by N, C, L, P, Vulgate Latin, Memphitic, Harclean Syriac, Armenian, and Syrian authorities, after v/^Ss by D, E, F, G, 17, Goth., and omitted altogether by B, A, Peshitto Syriac, and Augustine. Tischendorf and Tregelles follow the first array, although Tregelles places "omit" opposite in the margin, and Westcott and Hort follow the last, placing /loAAov in their margin before v/u.Ss. Who is right? Primdfacie the first group is Alexandrian, the second Western, and the third Neutral; and were this the true finding it would be difficult to resist the combined evidence of the Western and Alexandrian texts in an insertion in which they did not stand in collusion. More likely, however, the insertion of yu.oAAov is Western, and the misplacing of it a later divergence; in which case Westcott and Ilort's conclusion will result. Another instructive reading occurs in 2 Cor. xii. 7, where N, B, A, G, 17, ^thiopic, insert a 810, which D, P, the Latin, Gothic, Syriac and Armenian versions and the Syrian evidence omit. The omission is here easily seen to be Western, while the insertion has the combined support of the Neutral and Alexandrian documents and on genealogical grounds is preferable. In Gal. ii. 12, where N, B, D*, G, 73, 45, Origen read r,\Oev against jj\Oov read by A, C, Dbandc,H, K, L, P, most minuscules, Vulgate Latin, Syriac, Memphitic, Armenian, Gothic versions and fathers, we have one of the rare cases in which N B together unite with D, G, in a Western corruption; for corruption this is certainly shown to be by internal evidence. Again, we learn that the rule ascertained by internal evidence of groups that N B is usually right is not exceptionless; and that though N and B do not usually unite in the same Western readings, they do unite in one occasionally. This is an example of this rarity.

The difficulty of dealing with variations on genealogical grounds culminates in that portion of the Epistles (Heb. ix. 14 to Philemon inclusive of the Pastoral Epistles) where B is lost. Shall we read» for instance, "priest" or "high priest " at Heb. x. 111 All three of the great editions read "priest," but Tregelles and Westcott and Hort put the alternative in the margin. For "priest" we have N, D, E, K, L, 17, 47, most minuscules, Old Latin codices, Vulgate Latin, Memphitic, the text of the Harclean Syriac, Chrysostom, Euthalius, Theodoret; while for "high priest" we have A, 0, P, 31, 37, 46, 73, 74, 80, 137, and sixteen others, Peshitto and Harclean Syriac (with asterisk), Armenian, ^Ethiopic, Cyril of Alexandria. We long for B: if B should stand by N, D, etc., we should have the approved group N B D=Neutral + Western; if it should take its place alongside of A, C, P we could recognise it as Neutral versus N, D, Western. Internal evidence of readings and a careful study of grouping inclines us to suppose the former most likely to be the right solution. The weight of genealogical evidence is more clearly traceable in the case of three interesting readings in the first verse of the same chapter, where N P adds avriav (after Owuk) which the Western class, A C D, omits; N C reads as against the Western class, D H L, which supports als; and NAC P 17 67** reads Swavrai against the Western DHL, supporting Swacrai. In no one of these cases would the presence of B on either side change the determination.

In the Apocalypse, finally, genealogical evidence can as yet be scarcely employed at all, without the greatest doubt and difficulty.