Chapter III


IN the foregoing pages the available methods of criticism have been considered separately, and thus stock has been taken of the instruments within reach for the performance of this very delicate work. It remains to inquire how these instruments are to be used in the actual prosecution of criticism. Each method makes its own promises and attains for us its own results. But we must not permit ourselves to be satisfied with results obtained by one method only. The best criticism is rather that which makes the fullest use of all the methods, and checks and conditions and extends the results of each by the results of all. The value of combination of the methods is twofold. We thus obtain a system of checks: we may test the results obtained by one method by the results obtained by another, and by repeated trials preserve ourselves from error. And we obtain what may be called a system of relays: where one method fails to give a confident verdict, another may be called in, and thus their combination may enable us to carry criticism several stages farther than would be possible by one method alone. The effect of using a variety of methods, therefore, is both to extend the sphere which our criticism is able to reach and more firmly to settle the text over its whole extent. The first rule for the application of these methods, therefore, is to apply them all. Let no one be slighted; let each be used carefully and independently, and the results obtained by each carefully compared together. When the findings of the various methods agree the conclusion is certain, and we may feel sure that we have attained the autographic text. When they disagree, opportunity is given for review and revision of the whole process, with the not infrequent result of the discovery of an error, the correction of which will harmonise the evidence. By this repeated and, if need be, again repeated verification of our processes, our conclusions attain ever firmer standing; and it is very seldom indeed that the verdicts of the different kinds of evidence may not be brought into agreement. Until they agree some doubt continues to cling to our conclusions; and the canon may safely be formulated that no reading can be finally accepted against which any form of evidence immovably protests.

Experience further indicates to us that it is not a matter of entire indifference in what order we use the various methods of criticism. Certain of them are more liable than others to be swerved by the mental state of the critic, and it is a good rule to begin with the most objective. Certain of them yield at best only probable results, and it is a good rule to begin with the most decisive. Certain of them are largely negative in their findings, and it is a good rule to begin with the most positive. For each of these reasons it is safest to begin with the external evidence, and only when its bearing has been at least provisionally determined, to proceed to the internal evidence of readings. To begin with internal evidence of readings, especially with intrinsic evidence, runs very great risk of so filling the mind with the feeling that such or such a reading ought to stand in the text, that we may end by unconsciously making it stand there, against the evidence. The best procedure, and that most likely to issue soundly, is to begin with the consideration of the genealogical evidence, and when its results are obtained, to proceed to internal evidence of groups, and thence to internal evidence of readings,—usually in the order of, first, the transcriptional, and, secondly, the intrinsic evidence. When genealogical evidence speaks with force, it yields a testimony which ranks above all others in ease and certainty of interpretation, and consequently, by beginning with it, we consider, first, the surest evidence, and gradually proceed to that of more doubtful interpretation, although of no less finality when its meaning is certainly attained. After the evidence is all in, our next duty is to compare and harmonise the several results. When they are finally and hopelessly discordant, nothing is left us but to consider whether the oldest transmitted text may not itself be corrupt, and thus differ from the autographic text.

Perhaps the best way to exhibit the right procedure in criticism is by means of an example or two. Let us look at the famous reading in Acts xx. 28, where we have the following variations :—

A, C*, D, E, 13, 15, 36, 40, 69, 110, 118, and eight others; g of the Old Latin, Memphitic, Thebaic, margin of the Harclean Syriac, Armenian, Irenseus (Latin), (Athanasius), Didymus, Jerome, etc.

#eou, B, N, 68, lectionary 12, and twelve others; Vulgate Latin, (Peshitto), text of the Harclean Syriac, Epiphanius, Basil, Theodore of Mopsuestia(Latin), Cyril of Alexandria, etc.

Xpio-rou, -^Ethiopic, perhaps the Peshitto, m of the Old Latin (Jesu Christi).

Kvpwv Ko.i Oeov, C3, H, L, P, most minuscules, Slavonic,
Theophylact, etc.

OeOV KO.I KVptOV, 47.
KVplOV OfOV, 3, 95**-

If we should undertake to estimate the relative weight of these groups of testimony by the weight of the separate codices included in each, we might well despair of ever reaching a conclusion. The best uncials are for ®eov, the best minuscules and versions for Kvplov, the most witnesses for Kvpiov Kal ®eov. Fortunately there is a better way. Beginning with the genealogical evidence, we sift out all readings but Kupt'ov and ®eov in sifting out the Syrian evidence. We observe next that the typical Western document D stands on the side of Kupiou, and the typical Neutral B on the side of ®eov, and considering the other testimony for each, we see that this much is certain: ®eou is the Neutral reading, and Kvptov the Western. The most constant representatives of the Alexandrian class stand by the side of D and the Western witnesses, in support of Kupi'ou; here are A, C, 13, 36, 40, 69, 110, Memph., TLeb. Were not all these documents full of Western readings, we might find the Alexandrian reading in Kupi'ou, but this is not presumable in the mixed condition of all these documents, and internal evidence of classes gives us no ground to believe that the union of the Western with the chief Alexandrian documents is a union of the two classes. We must treat this reading, therefore, as a case in which the Western and Neutral classe.5 oppose one another, and internal evidence of classes forces us to accept in such cases the Neutral reading as presumably right. Thus the genealogical evidence supports ®eou. On turning to internal evidence of groups we obtain the same result. The high character given to B N by this process, whether it stands alone, or in whatever combination with other documents, affords strong ground for pref jrring ®eou, especially as it has the important further support of the Vulgate Latin and Cyril of Alexandria. This result is cumulative to the former, so that the external evidence throws a very strong cumulative probability in favour of ®eou.

We next appeal to the transcriptional evidence. The three readings Kupi'ou Kiu ®eou, ®<ou Kiu Kupi'ou, and Kupi'ou ®eou, are clearly all connate readings, and presuppose the previous existence of both the others. They are, therefore, out of consideration. Xpio-roS is easily accounted for either as a substitution of a synonym for Kupi'ou or ®eou (for whichever word was used, Christ was the person meant), or a misreading of an abbreviation, Ky or By being taken for XY, or even perhapsjipY (cf. Kpn, 1 Cor. i. 1 of Codex Augiensis) for XPY (D- ssepe: cf. Rom. vii. 4 in N). In either case it is a derivative reading and may be neglected. The problem of transcriptional evidence, then, is to decide between the relative originality of Kv/h'ou and ®eov, the difference between which again concerns only a single letter: Ky and 6y. As a mere blunder, either might equally easily pass into the other. They are equally brief. Either reading would be characteristic enough; the phrase "Church of God " is as common as the phrase "the blood of the Lord." But it is undeniable that ®eov is the more difficult reading, and this commends it to us as probably genuine. If ®eoti were original, it is easy to see that it would be startling, and that the scribe's mind working upon it might (scribe-like) intrude its mental explanation into the text; so that the very unusual character of the phrase here becomes, transcriptionally considered, its strongest commendation. On the other hand, if Kiyu'ov were the original reading, there is no jag in the phrase to catch the mind of the scribe and throw it off its balance; he would write smoothly on and find full satisfaction in the language as it stood. It seems, indeed, impossible to find any reason for altering Kupiou into ®eoC except a dogmatic one, and if dogmatic considerations be brought into the case they certainly authenticate ©eov rather. For a dogmatic alteration of Kupi'ou into ®eov could have no incitement except a cold determination to manufacture a proof text: there is nothing offensive to any one in the reading Kvpiav, and nothing that could suggest alteration. But ®eov might give offence to many: to extreme Arians, and to the orthodox antiPatripassians alike, and even to simple orthodox souls whose philosophical way of looking at theological language would be offended at this sharp paradox. Like language horrified Athanasius himself (Cont Apollinar., ii. 11, 12, 13). If dogmatic alteration has taken place, therefore, it certainly has softened the original ®eov into the less startling Kvpi'ou. And from every point of view the transcriptional evidence supports ®eov.

Does intrinsic evidence unalterably oppose this conclusion, commended alike by genealogical evidence, internal evidence of groups, and transcriptional evidence? For this is the way in which this branch of evidence may be fairly approached, seeing that it delivers negative judgments with far more force than positive ones. It is difficult to see how the reading ®eov fails to accord with the contextual flow of thought or the rhetoric. There is rather a fine propriety in it, and a solemn and moving motive lies beneath it. Paul incites the elders to more heedful attention to their duties to their flock by the considerations—(1) that it was the Holy Ghost who made them bishops, and (2) thai; it was the blood of God Himself that bought the flock now placed under their care. It is said, however, that it is un-Pauline to cull Christ God. The argument is a merely verbal one, and hence of small weight. And it is easy to point to Rom. ix. 5 and Titus ii. 13, where Paul does call Jesus God; and when it is objected that these are disputed passages, it is just to remind the objector that this will exclude his original statement as well as our rebuttal of it. Apart from such passages, however, it is very easy to show that Paul held a very exalted doctrine of Christ's person, and might as well as John (John i. 1) have given Him the name which his descriptions imply; and this is enough to set aside the force of the objection that the unwontedness of the phrase is fatal to its genuineness. This very unwontedness is from the transcriptional point of view its best proof of genuineness, and it is not the part of intrinsic evidence to pare down the unusual. The phrase would oppose its own genuineness only if it contradicted Paul's otherwise known opinions, or at least were not only unexampled but inexplicable. But since this same Paul has elsewhere declared that Christ was begotten before every creature, we need find nothing to stumble at in his applying to Him here, where the context bids us look for a solemn enhancing of the greatness of the gift of His blood, the name which is elsewhere implied. The effect of these considerations is not merely negative; it is corroborative of the other evidence. And since all forms of evidence unite to commend ©tou here, their cumulative effect makes it certain that this is the original reading.

Our next example shall be the very important variation that is found at John i. 18. Here the chief rival readings are :—

vtos: A, Cc, E, F, G, H, K, M, S, U, V, X, P, A, A, H, and all minuscules except 33; the Old and Vulgate Latin, the Curetonian Syriac,

the text of the Harclean Syriac, the Jerusalem
Syriac, the Armenian in Platt's edition [Irenseus
(Latin)], Eusebius, Athanasius, Theodore of
Mopsuestia, Chrysostom, etc.

povoyevrrs Oeos: N, B, C*, L, 33 (33 prefixing o); the Memphitic, Peshitto Syriac, margin of the Harclean Syriac, the Valentinians [Irenseus (Latin)], Clement, Origen, Epiphanius, Didymus, Basil, Gregory of Nyssa, Cyril of Alexandria, etc.

Genealogically, it is to be noted that 6 /tovoyev^s Dios is the reading of the Syrian class, and when the Syrian testimony is sifted out, of the typical Western witnesses. D is defective here; but the union of A X, Old Latin and Curetonian Syriac, cannot well have more than one meaning. On the other hand, the Neutral documents (B, S) unite with the most constant Alexandrian documents (C, L, 33, Memphitic), and the Alexandrian fathers, for fMvoyevys ®eos, which thus seems to have the combined support of the Neutral and Alexandrian classes. Internal evidence of classes very strongly commends the Neutral-Alexandrian readings, and genealogical evidence thus gives a very strong verdict for /xovoyev^s ®eos. Internal evidence of groups casts its weighty vote in the same scale,—as B , supported by an additional body of important witnesses, advises us. So that again external evidence is cumulatively set in favour of one reading,—/Aoyoyevijs ®eos.

The chief divergent words in the two readings differ from one another in this case, too, by a single letter, iince they stand in the MSS. Yc and 9c; and transcriptionally either one of these might very readily pass into the other by a mere scribe's blunder. The case is complicated, however, by the connection of the insertion or omission of the 6 nine letters back with the variation in the main word. This seems to exclude a mere error of the eye as the cause of the change; and dogmatic considerations stand in this case just as in Acts xx. 28. The insertion of ®eos for dogmatic reasons would be a barefaced manufacture of a proof text, as the reading vtos could give offence to no one, while, on the other hand, the reading ®eos might be an offence to a great body of readers. If dogmatic considerations, therefore, are responsible for either reading, surely they have produced the softening vios, and not the startling ®eos. The canon that the harder reading is to be preferred, again, commends ®eds. If o... vids stood here originally, there would be nothing to attract a scribe's attention or to suggest a change. "The only-begotten Son " is a sufficiently common phrase in John to give itself readily to the pen when ftovoyevqs is being written. On the other hand, " only begotten God" is unique; if the scribe observed it, his mind might unconsciously transmute it into the more familiar phraseology, and if he merely glanced at the phrase he might readily take it for the more familiar "only begotten Son." In every way, thus, transcriptional evidence commends /Aovoyevi^ ®eo's.

Intrinsically, either reading, had we known it alone, would be satisfactorj enough. "The only begotten Son" is a Johannean phrase, and John might be expected to use it here too. But to call the Logos "God " is also Johannean, and "only begotten God" only unites hero the two predicates which had just before been assigned to the Logos (®€os ver. 1, and /iovoyevrjs ver. 14). When the sequence of the thought in the prologue is carefully examined, a fine appropriateness for "only begotten God " just here emerges, which goes far towards authenticating that reading. John describes to us, first, the Word in His eternal relations (verse 1 ) ; then, the Word in His relations to creation (verses 2 — 13); and then the revelation of God through the Word (14 — 18) — culminating with putting into words in verse 18 what was already implied in the facts, that the Word was God (ver. 1), and yet Himself became flesh (ver. 14), — viz., that this revelation was self -revelation. If no one has seen God at any time, who is His revealer if not the Word who was God (ver. 1), and only begotten (ver. 14) — God only begotten (ver. 18)? The intrinsic evidence, thus, not only fails to oppose the reading commended alike by genealogical evidence, internal evidence of groups, and transcriptional evidence, but even corroborates it. And again we may accept the fourfold support as giving us a reading which is certainly the original one.85, 86; [Memphitic], [Thebaic], [Peshitto],
Harclean Syriac's margin, Gothic, [^Ethiopic],
[Armenian], [Origen] Epiphanius, (Theodore of
Mopsuestia), etc.

It is natural to take as our next example the famous reading in 1 Tim. iii. 16. Here three variations demand our attention :—

o: D, Zahn's Codex (Supplementum Clementinum, p. 277), Old Latin, Vulgate, [Peshitto], [Harclean Syriac], [Memphitic], [Thebaic], [^Ethiopic], [Armenian], Latin fathers, etc.

The greatest difficulty that faces the critic here lies in the uncertainty that attends so much of the evidence. Expert palseographers differ diametrically as to what the reading of A is, whether ec or oc (®eos or os), and in the present worn state of the MS. decision by renewed examination is impossible. The same kind of controversy has been held as to the reading of C. although apparently with much less reason; and although we have inclosed C also in doubting parentheses we entertain no great doubt as to its support of os. A large proportion of the versions so deliver their testimony as to make it indeterminable whether they read os or o; they have been placed in both lists inclosed in square brackets. Codex 181 has also been inclosed in brackets, as its existence has been doubted. Codex 73 has been personally examined by Dr. Schaff, and certainly reads os.

On applying genealogical considerations to this evidence, all the testimony that is at all certain for ®eos sifts out with the sifting out of the Syrian testimony. This reading appears in no father until late in the fourth century, in no version until at least the seventh century, and in no MSS. until long after the Syrian text had become everywhere the virtual textus receptus. If A be adjudged to read ®«>s the determination of its Syrian character would not be affected; and the very late character of all other witness for it is itself an argument against the likelihood of either A or C having ever had this reading, and much more against both having it. On genealogical grounds, thus, ®eds is at once set aside, and the choice rests between os and o. It can scarcely be doubted that o is Western; while the attestation N (A) C 17 gives os the appearance of having the support of the Neutral and Alexandrian classes. The doubt that hangs over the testimony of the versions is of the less moment because of the certainty of the Latin reading, which enables us to identify the Western type; and the absence of B is here of no importance, as its presence on either side would not affect our determination. Genealogical evidence thus verypointedly commends^os. Internal evidence of groups corroborates this finding. N A C or N C alone is one of the best groups attainable in this part of the New Testament, and although the absence of B disturbs us here, yet the transcriptional evidence comes to our help by making it improbable that o can be the correct reading, and hence enabling us to account all the testimony for both os and o combined against that for ®eds. The result is to condemn ®eos hopelessly.

The transcriptional evidence is thus in a true sense the key to the problem. As between os and o, the succession of round letters, ioNocec|>A, would render the change easy either way, whether by mistaking the c for the succeeding e, or the already written c for the half-finished e. Unless, however, os were original, it could never have been written except by a mere blunder, and could scarcely escape the eye of the "corrector "; while o could easily be passed over on account of the easy sense which it introduced, and would be apt to be written by the scribe after the neuter antecedent (jLvcrnqpiov. As between os and ®eos the same canon of the harder reading decides for os. Here the difference is only in the fine lines that distinguish the o from 6 and mark the contraction: ec and oc; and thus one reading may easily pass into the other. But again, as ®eos is grammatically easy, forming a proper apposition for ^va-r^piov, while os is grammatically hard, nothing but a mere blunder could have originated os, while the difficulty of the sense would have operated as an incitement to the conscious or unconscious transmutation of os into ®ebs.

Unless, then, intrinsic evidence immovably protests against os it is to be accepted as the true reading. It is indisputable that it introduces a difficult reading, and the difficulty seems to disappear with the change to o or ®eds; on these facts the transcriptional evidence founded its preference for os. But does the difficulty rise to so high a pitch that os is impossible t The difficulty is wholly grammatical, and the grammar is not made intolerable by os, but only relatively hard. Moreover, Qeo's, while apparently reducing everything to an easy smoothness, introduces difficulties of its own. It accords well with the first of the following clauses, but immediately becomes an unnatural antecedent to the next, and continues so throughout. It is thus a fair sample of scribes' work, and combines the surface appearance of fitness with a real unfitness for its place. When, next, the antithetic and rhythmical character of the succeeding phrases is observed, suggesting that we have here a fragment of a hymn, which would allow us to suppose that the grammatical antecedent to os is to be sought in the hymn rather than in this context, or, better, that the first clause is the subject followed by five predicates; the intrinsic evidence, so far from immovably opposing 05, appears to be slightly in its favour. No doubt, o would be intrinsically unobjectionable, but it is not preferable to os save in the strict and narrow grammatical sense; and intrinsic evidence readily gives way here to transcriptional evidence in its strong preference for os. In this reading, therefore, difficult as it at first seems, all varieties of evidence come finally to agreement upon a single reading os,—which we may, therefore, confidently accept.

Our next example shall be one of those few readings which affect large sections of the New Testament text: Shall we insert or omit the famous pericope of the adulteress, John vii. 53 —viii. 11? The evidence is as follows:—

Insert: D, F, G, II, K, TI, T (also E, M, S, A, II, etc., with asterisk or obelus), more than three hundred minuscules; many codices known to Jerome; the Latin MSS. b, c, e, ff2, g, j, 1; the Vulgate Latin, Jerusalem Syriac, ^Ethiopic; "Apostolical Constitutions," Nicon, Euthymius, Ambrose, Augustine, Jerome, and later Latin fathers.

Omit: N, (A), B, (C), L, T, X, (A); codices known to Jerome, 22, 33, 81, 131, 157, and many other minuscules; the Latin MSS. a, f, q, rhe, and others known to Jerome' and Augustine, Curetonian, Peshitto and Harclean Syriac, best MSS. of the Memphitic, Thebaic, Armenian, Gothic; (Origen), (Eusebius), (Theodore of Mopsuestia), (Apollinaris), Chrysostom, etc.

On sifting out the Syrian witnesses, the testimony for insertion plainly becomes merely Western, including D and the European Latin; but not certainly the African Latin, although e contains it, inasmuch as the early Latin Fathers are strangely silent about this passage. The testimony for omission includes everything typical in both the Neutral and Alexandrian classes. The only difficulty that meets us in determining the genealogical classes arises when we try to trace the Syrian class. Most of the later documents contain the section, but it cannot be traced in the Antiochian and early Constantinopolitan fathers. Whence it seems that this pericope found no place in the Syrian revision, but has passed into the Syrian text from the Western, say, at some time about the seventh century. Whatever its relation to the Syrian class, however, the section is strongly discredited by genealogical evidence. The finding of internal evidence of groups, which is very strongly given, is in the same direction. So that the external evidence is solidly arrayed against the genuineness of the section.

Transcriptional evidence is generally ambiguous in readings of great length; insertion or omission must have been alike a mere blunder. It seems difficult to account for such a blunder as its omission, however, except by some such accident as the loss of a leaf or two from the exemplar. Mr. J. R. Harris has shown that the matter of this section corresponds, in extent, very exactly to two leaves of what seems to be a form which might very well belong to an ancestor of B. But he also shows that it would not all have fallen on four pages, if belonging to the present place in John. On the other hand, its insertion may readily be accounted for as an incorporation into the text of an explanatory gloss drawn from some extraneous source. When we add that some codices place it at the end of John's Gospel and some after Luke xxi., instead of here, it becomes still more probable that we are dealing with phenomena of insertion rather than of omission. On the whole, the transcriptional evidence, while able to accept the passage if otherwise commended, is itself rather in favour of its omission.

Intrinsic evidence is more strongly so. For the fact that the story is worthy of our Lord and bears every mark of historic truth has no bearing on the question whether it is part of John's Gospel; any true story of Jesus would be beautiful, especially if it came ultimately from the apostolic circle. While, on the other hand, the style and diction are very unlike John's writing elsewhere; several words are used which seem strange to his vocabulary; and some matters of detail fit ill with the context,—e.g., Jesus is left alone with the woman at verse 9, and yet addresses "them" at ver. 12, and the Pharisees answer at ver. 13. This last fact might be of small moment, except that in these very matters verses 12 and 13 fit on directly with verses 45—52 of the seventh chapter, and so the omission of the disputed verses restores verses 12 sq. to a context with which they seem to belong. Nor is this close connection of verses 12 sq. with the seventh chapter merely verbal; the presence of the pericope of the adulteress seriously disturbs the progress of a discourse the order of which would be admirable without it. This intrinsic evidence is so strong that it would almost cast doubt on this section of itself; and in union with the external evidence, and with the allowance of the transcriptional, it forces us to omit the passage. Here too, therefore, we may feel that we have attained the original text.

It is appropriate to draw our next example from the only other various reading that involves so large a section,—that which concerns the last twelve verses of Mark. The evidence may be stated as follows:—

Insert: A, C, A, D, X, S, *, T, etc., 1, 33, 69, and

nearly all minuscules; all Old Latin codices except k; the Vulgate Latin; the Curetonian, Peshitto, Harclean and Jerusalem Syriac; the Mernphitic, and Gothic; Justin, Tatian, Irenseus, [Hippolytus], Macarius Magnus; and post-Nicene fathers generally.

Omit: B, N, L, 22, 743 (on the authority of the Abbe Martin); codex k of the Latin; the Armenian, and ./Ethiopic; [Clement], [Origen], Eusebius, [Cyril of Jerusalem], and, among the post-Nicene fathers, the wrofecri's, Jerome, Victor of Antioch, Severus of Antioch. Also such minuscules as 15, 20, 300, 199, 1, 206, 209, which preserve knowledge

of the doubt.

Some words are necessary in explanation of this evidence. N simply omits the passage. B omits it, but leaves a blank space, which is apparently intended for it; this seems to prove that the exemplar from which B was copied lacked these verses, but that they were known to B's scribe. As the weight of B is due to the character of its exemplar, not to the knowledge of its scribe, this does not affect B's testimony. L closes at verse 8, but adds at the top of the next column: "These also are somewhere current: 'But all things that were commanded, they immediately announced to those about Peter. And after this Jesus also Himself, from the east even to the west, sent forth by them the sacred and incorruptible proclamation of eternal salvation.' These are also, however, current, after 'For they were afraid.' "... And then our usual twelve verses are inserted. The existence of this shorter conclusion (to which L gives the preference) is d, fortiori evidence against the longer one. For no one doubts that this shorter conclusion is a spurious invention of the scribes; but it would not have been invented, save to fill the blank. L's witness is, then, to MSS. older than itself, which not only did not have our twelve verses, but had invented another conclusion in their place. The Abb6 Martin tells us of another codex, which he numbers 743, that repeats the arrangement of L. Codex 22 closes the Gospel at verse 8, marking it as "The End," and then adds : " In some of the copies the Evangelist finishes at this point; in many, however, these also are current," . . . and inserts our verses 9—20, closing again with " The End." The Old Latin MS. k contains the shorter conclusion only, and hence is a specially strong witness to the omission of our twelve verses. The Thebaic version might possibly be added to the witnesses for insertion, but we have from it only a mediocre paraphrase of verse 20, and it cannot be confidently determined what disposition was made of it.

Proceeding now to estimate the evidence, we note first that the Syrian text inserts the passage, and, when the Syrian witnesses are sifted out, it is left with Western (D, Latin, Curetonian Syriac), and apparently Alexandrian (C, A, 33, Memphitic) witnesses only, and since all Alexandrian witnesses are full of Western readings, this means with Western witnesses only. For omission we have the Neutral witnesses (B, N) with L, 22, and other support. Where the Alexandrian reading stands we cannot discover ; but on appealing to internal evidence of classes the apparent conjunction of Western and Alexandrian witnesses is discredited, and we must decide that the genealogical evidence is in favour of omission. L may represent the Alexandrian text and k the primitive Western; and in the case of either of these hypotheses, the verdict for omission receives additional strength. Internal evidence of groups, which throws strong favour on B N, only confirms genealogical evidence, and we have the whole weight of external evidence for omission.

The transcriptional evidence leads to the same conclusion. No good account can be given of the omission of these verses. To suppose that they were omitted in a harmonic interest is to presuppose a freedom and boldness in dealing with the Gospel narratives never elsewhere experienced, and that to serve a purpose far more easily attained. To suppose the omission to have arisen from the misunderstanding of a note placed here to mark the end of a liturgical lesson is to assign a greater age to the present lessonsystem and to this method of marking MSS. than can be proved for either. To suppose that a leaf was lost from the end of the Gospel, containing these verses, will best of all account for their omission, but will not account for its wide distribution, nor for the failure of the beginning of the next Gospel, on the other side of the leaf, to get lost too. Mark stands very rarely in Greek MSS. at the end of the book of the Gospels, and the loss of a leaf early enough to affect the ancestors of N, of B, of L, and of Western k, must have affected nearly all MSS. as well. On the other hand, the insertion of such an ending is transcriptionally easy to account for. The abrupt ending of verse 8 demanded something more. That the scribes felt this is evidenced by their invention of the certainly spurious shorter ending. Why should not other scribes have sought and found another tolerably fitting close for the Gospel? And that this ending does not belong here, but fits its place only tolerably, is clear on careful examination. The tear at verse 8 is not mended by verses 9—20. Only Matthew and Luke tell us what actually happened after verse 8. And if verse 8 demands a different succeeding context, verses 9—20 no less need a different preceding one from that here furnished them. Jesus is presumed to be the subject in verse 9; but the subject that would be taken over from verse 8 is the women. The "but" that opens verse 9 does not introduce anything adversative to verse 8. The new specification of time in verse 8 is surprising, after verse 2. "First " looks strange here. The identifying description of Mary Magdalene in verse 9 is very remarkable after verse 1. Every appearance, in a word, goes to show that the author of the Gospel did not write versos 9—20 as the conclusion of the narrative begun in verses 1—8. And if so, the transcriptions! evidence that makes an insertion here easier to conceive of than an omission has full play, and we can recognise verses 9—20 as only another way of filling up the gap left by the iinfinished appearance of verse 8. The intrinsic evidence is not fully stated, however, until we add that there are peculiarities of style and phraseology in verses 9—20 which render it easy to believe that the author of the Gospel did not write these verses.

The combined force of external and internal evidence excludes this section from a place in Mark's Gospel quite independently of the critic's ability to account for the unfinished look of Mark's Gospel as it is left or for the origin of this section itself. The nature of the matter included in them, and the way they are fitted to the Gospel, seem, however, to forbid the supposition that these verses were composed for this place by any scribe. It is nearly as hard to believe that anybody wrote them for this place as it is that Mark did. They seem to be a fragment rather, adopted from some other writing and roughly fitted on to the end of Mark. This fragment is certainly as old as the first third of the second century, and may—as may also the pericope of the adulteress inserted into John—be taken from the book of illustrations of the Gospel narrative which Papias composed, apparently about 120 A.d. Neither is it necessary for the critic to be able to give an account of the mutilated condition of Mark's Gospel. To recognise that this fragment does not belong at the end of it does not make it any more mutilated than it was before. The evident incompleteness of verse 8 is evidence against the opinion that the Gospel was intended to close at that point; but no evidence that just this conclusion, —which does not fit on to verse 8 nor complete it, nor the subject then in hand,—was the conclusion intended. Why Mark's Gospel has come down to us incomplete, we do not know. Was Mark interrupted at this point by arrest or martyrdom before he finished his book? Was a page lost off the autograph itself? Or do all of our witnesses carry us back only to a mutilated copy short of the autograph, the common original of them all, so that our oldest transmitted text is sadly different from the original text? There is room for investigation here; but, apparently, no room for accepting this conclusion for the one that Mark wrote or intended to write.

We have purposely chosen all these examples of such a sort that the evidence can readily be seen to be harmonious through all the methods. But we have also purposely placed last among them a case in which the intrinsic evidence, while uniting with the other forms of evidence in determining this reading, is left still somewhat unsatisfied by its determination. It opposes the acceptance of the last twelve verses of Mark as genuine: but it no less opposes the acceptance of verse 8 as the end of the Gospel. It consents that this is not the limb that belongs here, but it no less insists that some limb does belong here. This may remind us that the work of the critic may not always be done when he has passed on all the readings which have been transmitted to us in our extant witnesses. It is at least conceivable that the oldest transmitted text may not yet be the autographic text, or in other words, that all our extant documents spring from a common original that is removed by a few copyings from the autograph, and may, therefore, contain some errors. Of course, this is not to be assumed to be the fact; but neither is it to be assumed not to be the fact. This, too, is to be settled only on trial and by the evidence. And here it will be of use to us to remember that the office of textual criticism is not merely to restore a text where it is known to be in error, but to examine all texts in every part in order to certify their correctness or discover that and where they are corrupt. Where the several documents give various readings the presence of error in some of them is already demonstrated, and the office of criticism is to determine which, if any, is right. But by this very act it contemplates the possibility that none of them are right, and it very frequently actually determines that the most documents may be in error. How narrow the chance that has preserved for us the true reading in all those cases in which we adjudge the palm to the few old documents as against the many! By the destruction of B and a half-dozen other docu ments we should destroy all extant evidence for several quite important readings which we now adjudge right; and in all these readings a false reading is prevented from standing in all texts without variation only by the accident of the preservation of these half-dozen documents. The possibility must be frankly confessed that other false readings may stand in all our extant documents. So that, even where there is no variation, criticism is still necessary to certify to us that the text is free from error or to correct it when in error.

Wherever, therefore, the evidence for any body of variations is so hopelessly in conflict that it cannot be harmonised, and in all that part of the text on which there are no variations, it is right to consider the text only provisionally determined, and to subject it to further criticism. In all cases of variation in which the evidence is in ineradicable conflict the high probability is that the oldest transmitted text is itself in error, and we may assume that here is .a case that needs further criticism. In all that part of the text on which there are no variations the strong presumption is that we have not only the oldest transmitted text (which is certain, since it is identically transmitted in all witnesses), but also the autographic text: but nevertheless this presumption may not be everywhere equally well grounded, and examination is necessary in order to conviction. Only in that part of the text which has been settled by the combined and harmonious testimony of all kinds of evidence may we confidently accept it as the autographic text. For in all these cases alike, the only evidence that is valid —whether to discover if the text be corrupt where no various readings occur, or to suggest the right reading wherever we know or suspect it to be corrupt—is internal evidence; and in all cases where the text has been already settled on the harmonious finding of all kinds of evidence, this has already spoken and has already been satisfied.

Before we close our discussion of the praxis of criticism, therefore, we must explicitly recognise the legitimacy and duty of examining the text of the whole New Testament with the most scrupulous care, with a view to discovering whether its transmission has been perfect; and of appealing to internal evidence to suggest and settle for us the true text in all cases of variation where the evidence is hopelessly in conflict, and in all cases where, in the absence of variation, an examination of the text has resulted in leading us to suspect corruption. It is evident that we are not here calling in a new method of criticism beyond those enumerated; but only extending the practice of criticism a step further than we had need to go in the examples which we have adduced. And it is further evident that the validity of this extension is involved in any use of internal evidence for settling readings at all. The technical name given to this extension of criticism is "conjectural emendation," which is meant to describe it as a process which suggests the emendation which the text is shown either by the presence of irreconcilable variations or by internal considerations to need, from the conjecture of the mind, working on internal hints.

The need of calling upon conjecture to aid us in determining the text of the New Testament depends on the provable presence of variations the evidence asi to which is in hopeless conflict, or of passages which, while without variation, are clearly corrupt. In dealing with this question of fact, the utmost tact, good judgment and candour are necessary. Two extremes are equally to be avoided. We must neither allow ourselves so to sharpen our acuteness that we discern an error in every corner, and lose the power to catch the plain intent of a plain man's plain speech; nor must we so blunt our minds, by attempting to explain as correct and good Greek what we could not tolerate in any other language, that no amount of evidence can convince us of the presence of a textual error. Licence has not been unknown in either direction. Some critics have seemed ready to cast the whole text into "pie," and set it up again to suit theiv own (and no one else's) conceits. Others have even savagely guarded each fragment of the transmitted text as if the scribes had wrought under Divine inspiration. The whole matter is nevertheless simply a matter of fact, and is to be determined solely by the evidence, investigated under the guidance of reverential and candid good sense. The nature of the New Testament as a Divine book, every word of which is precious, bids us be peculiarly and even painfully careful here: careful not to obtrude our crude guesses into the text, and careful not to leave any of the guesses or slips of the scribes in it.

Drs. Westcott and Hort enumerate in their edition some threescore or more passages in which they (or one of them) suspect that a "primitive error" is found in the text—i.e., an error older than our transmitted text, for the removal of which we are confined to conjectural emendation. Our own judgment would greatly reduce this number. Without discussing, however, the special cases, it is enough for our primary purposes to lay down two rules of action:

(1) Critical conjecture is not to be employed in settling the text of the New Testament until all the methods of criticism have been exhausted, and unless clear occasion for its use can be shown in each instance.

(2) No conjecture can be accepted unless it perfectly fulfil all the requirements of the passage as they are interpreted by intrinsic evidence, and also perfectly fulfil all the requirements of transcriptional evidence in accounting for the actual reading, and if variants exist also for them (either directly or mediately through one of their number). The dangers of the process are so great that these rules are entirely reasonable, and indeed necessary. The only test of a successful conjecture is that it shall approve itself as inevitable. Lacking inevitableness, it remains doubtful.

Few as the passages are that can be shown to need conjecture to settle their text, the passages in which successful conjectures have been made are still fewer. Perhaps no absolutely satisfactory one has yet been made. The best examples are probably two on Col. ii. 18, one by Bishop Lightfoot and the other by Dr. C. Taylor. Instead of the best attested reading, a edpafcev e/x/?areiW, the former scholar proposes eiapa or altupq. Keve^arevtav, which is attained by a change of only a single pair of letters, eo into 10). The latter scholar proposes depot Keve^arevoiv, which simply omits o. In such matters we may well listen to the advice of the Jewish sage and "be deliberate in judgment."