Textual Evidence

Sanders on Scribal
Errors in Codex W

Excerpt from: H.A. Sanders, THE WASHINGTON MS OF THE FOUR GOSPELS, (1912)

Page Index

Sanders: - Washington Codex (W):
    Homoioteleuton - Haplography examples in Gospels
    Dittography - example in John
    Scribe of MS - omissions, and master-copy/exemplar
    Corrections - by first hand

    The Problem of the Text - Codex W & the Versions
       Von Soden's K-text
       Bilinguals and Trilinguals
       Early Versions - c. 150 A.D
       Tatian's Diatessaron - & the Old Syriac

Return to Index

Sanders on
Codex W

Excerpt for review from:

Headings have been added for clarity and navigation purposes.



Examples of Homoioteleuton

Omissions by carelessness or because of like endings, [homoioteleuton/arcton] which can be definitely assigned to our scribe, are a few:

(1) Matthew 4:21-22 (214-220 letters omitted)

Example column width: 24-25 chars

..τα δικτυα ηκολουθησαν αυτω

και προβας εκειθεν ειδεν αλλ-
ους δυο αδελφους ιακωβον τον
του ζεβεδαιου και ιωαννην τον
αδελφον αυτου εν τω πλοιω με-
τα ζεβεδαιου του πατρος αυτων
καταρτιζοντας τα δικτυα αυτων
και εκαλεσεν αυτους οι δε ευθ-
εως αφεντες το πλοιον και τον
ΠΑΡ αυτων
ηκολουθησαν αυτω

και περιηγεν ολην την γαλιλαι-

(2) Matthew 15:18 (44 letters)

(3) Matthew 16:2-3 (36 letters)

(4) Mark 6:23 (36 letters)

(5) Mark 7:13 (30 letters)

(6) Mark 11:15 (14 letters)

(7) Luke 8:31 (17 letters)

(8) Luke 15:19 (27 letters)

(9) Luke 15:24 (22 letters)

(10) Luke 17:35 (65 letters)

(11) John 5:11-12 (69 letters) (between 1st/2nd quires, perhaps from parent)

(12) John 21:4 (49 letters)

Examples of Dittography

In John 6:56 there is a repetition of 5 whole lines (139 letters) not discovered by scribe or subsequent correctors.

The regular scribe is very free from such errors, if we consider the rapid style of his script and the length of the ms. Neither can the peculiarities of spelling, forms, and construction, as a rule, be referred to our scribe, but rather mark the character of the text tradition and its locality and age.

Singular AND Unique Features

Only in the case of those characteristics which run through the MS without change can we assume our scribe responsible. Among these I venture to draw deductions only from those pronounced characteristics which are rare or non-existent in other Biblical mss.

In this class we may enumerate: tendency towards aspirated consonants, αλλα before all vowels, κα for και, and the decided tendency towards Attic or other old forms. The most of these find their nearest parallels in the early papyri and the oldest uncials of Egyptian origin, thus confirming the supposed Egyptian origin and suggesting an early date.


(1) Order, Omissions, Crowded Writing

The MS once contained the whole of the four gospels in the order, Matthew, John, Luke, Mark. This is the order known as the Western, of which the best known examples have been the MSS. D, X, 594, Old Latin: (a b e f ff1 q), and the Gothic (c. 375 A.D.).


As noted above, there are two lacunae caused by the loss of leaves. These cover John 14:25 (ο δε παρακλητος) to 16:7 (including ελευσεται) and Mark 15:13 (οι δε παλιν) to 15:38 (including εσχισθη εις δυο).

The remainder of the ms is so perfect that there is rarely a letter missing or indistinct.

On the preceding page I have listed 12 cases of longer omissions by our scribe ;
9 of these were due to like endings [homoioteleuton] and
3 to like beginnings of successive phrases [homoioarcton].

Column Width of Exemplar

We may assume that these omissions would more easily occur if the parallel parts stood at the beginnings or ends of neighboring lines, and thus may draw inferences as to the length of line in the parent ms.

The three omissions in Matthew are respectively 214, 44, and 36 letters long, indicating a line of either 20 or 40 letters in the parent. As W has about 30, it seems quite certain that the parent did not agree.

In Mark the three omissions are of 36, 30, and 14 letters each. These lengths might be consistent with a line length similar to W, but seem to point to a line of about half the length (15).

In Luke, the lengths of the four omissions are 17, 27, 22, 65 letters, which would seem to suggest the short line attributed to the parent of Matthew (e.g., 10 or 20).

In John there are two omissions of this type; one comes between the first and second quires and is 69 letters long ; the other, at 21:4, is 49 letters long. We are also assisted by a repetition 139 letters long, covering five lines in the repeated form and five lines and eight letters in its first form. If we may unite the evidence of these three, the parent MS [for John] would seem to have had a line from 23 to 25 letters in length, i.e. again a different length, and so indicating a different parent.

The average amount of text written on a sixteen-page quire of the MS is ten and one-half pages of the Oxford 1880 edition. Yet the first quire of John has about eleven and one-half pages, and the last two full quires of Luke (crowded writing noted above, p. 7) contain nearly twelve pages of text each. It is easy enough to explain large quires toward the end of a gospel, if crowding would have saved an extra small quire, but such is not the case here, as Luke ends in a four-page quire. This looks like a hint that the parent ms had larger quires. The larger first quire of John suggests a similar guess for that gospel as well. We shall find this thought confirmed in our study of the text affiliations later.

2. Corrections

There seem to be four well-defined groups of corrections to the MS.

(i) First Hand

There are 78 cases where the scribe corrected his own blunders. Only rarely is there doubt as to the author of the correction. The original scribe uses a full round dot above a letter to delete it. The dot is made as dark and heavy as his ink allowed. He erases only rarely, preferring to wash or wipe off the still moist ink. The example which makes the delete dot sure for the first hand is in Luke 17:35, where we find kai imokpiQkvrkk kkyov of verse 36 standing before verse 35, though it follows in its regular place ; cf . above, p. 26. Similarly deleted errors are : Matthew 17, 25 6 i? (also deleted by second hand); Mark 10, 35 (see under third hand); 15, 43 o; Luke 6, 26 v/jllp; 17, 20 work; 19, 23 fiov; 20, I avTo}; 24, 14 irkpL nain-iov; John 10, 30 fiov; 17, 22 SeScDKaq (8 also deleted by second hand) ; 19, 9 kid.


A few of the corrections by the original scribe are well-established variant readings. Their appearance as corrections made by first hand seem to indicate that they stood in the parent MS as glosses either between the lines or in the margin, and so were not always seen by the copyist at first.


Summing up this evidence we may note that in Matthew the first written forms agree in all four cases with the version tradition,^ while the corrected form is each time the same as the Antioch and Hesychian * recensions.

Practically the same condition holds for Luke 8:13 to end, for of the seven corrections six agree with the Antioch recension, usually supported by the Hesychian , while one agrees with the Hesychian alone. The forms first written agree with the version tradition. The Hesychian and Antioch recensions are found supporting these readings only once each.

Also in Luke i -8, 12, both of the corrections are from the text found in bilinguals or lectionaries to the Hesychian and Antioch recensions.

In Mark one correction is from the Antioch recension supported by lectionaries and some versions to the Hesychian supported by part of the version tradition. The other is from the Hesychian and Antioch recensions to the version tradition.

In John two of the corrections are from the version tradition to the Hesychian and Antioch recensions, and one is the opposite. Even from this fragmentary evidence it seems likely that the different parts of the parent MS had been corrected to agree with different text traditions.

The remaining 33 corrections by first hand are mostly due to errors of eye or memory ; all are given in the collation, so I shall omit them here.



The solution of the text problem of W has been much impeded by the inadequacy of the textual material in the critical editions and the impossibility of explaining its peculiarities on the basis of the text theories generally accepted. It was a common occurrence to find in the Tischendorf apparatus al 2, al 3, etc., as the chief authorities for noteworthy variants of W. This seemed at first much more disconcerting than to find no authorities cited, yet in the end I found that the two conditions were often not different, for Tischendorf might have taken from the older editions cursive ms authority for many readings, which he left unmentioned.

The inadequacy of any one critical edition, and the danger in omitting from consideration the minuscule MSS and the variants in the versions, is well illustrated by the fact that a comparison of W with the apparatus of Tischendorf left unexplained nearly five hundred important variants in the gospel of Mark alone.

Through the use of the minuscule MSS and the early versions this number of unsupported readings has been reduced by nearly three-fourths.

Furthermore, a comparison of the readings of W with von Soden's results, as shown in his prolegomena, convinced me that Tischendorf and Westcott & Hort had built on a false foundation.

Von Soden's K-text Vindicated by W

Von Soden's earliest form of the Antioch recension (K^) pointed so plainly in Matthew to W as its oldest and best representative, and his Hesychian recension (H) agreed so closely with W in Luke 1-8, 12, that I could not hesitate to accept his results, at least up to that point.

The matter was made more certain by the fact that the corrections of first and second hands showed plainly that these recensions had been corrected into an older style of text in our MS or its parent (cf. pp. 31 and 36).

That there was another, probably older, recension connected in some way with Origen is also likely, but that assumption does not seem sufficiently to explain all the divergences of the "Western Texts" of Westcott & Hort.

Harris,* Chase,* and especially Hoskier* have, I believe, started on the right path here. Enough has been done so that it may be considered as settled that the peculiarities of the so-called Western text (von Soden's I-text ) are closely allied to the early versions.

My comparisons with the text of W, especially in Mark and the early part of John, establish this intimacy most clearly.

Only rarely did I fail to find authority for W's " special " variants in some one of the versions, Syriac, Latin, Coptic, Gothic, Armenian, or Ethiopic. In the cases where I failed I generally attributed it to the inadequacy of the textual apparatus in the versions or to the insufficiency of my own acquaintance with all these languages.

Bilinguals and Trilingual MSS

Harris (op. cit) first showed the extent of Latin reaction on the Greek text in the bilinguals ;

Chase (op. cit) followed with a more elaborate proof of the Syriac influence, which though adequate on the main theme claimed too exclusive an influence of Syriac.

It remained for Hoskier to reconcile the conflicting ideas by assuming the early existence of trilinguals, in which there might be influence of more than one version on the Greek text at the same time.

In my article in the Amer. Jour, of Phil. vol. 33, pp. 30 fl., I hesitated to accept this theory in full on the ground that it was barely possible to explain the textual conditions on the basis of bilinguals alone, and that no proof of the early existence of trilinguals was known to us. Yet even in the few weeks of study since that article new evidence has come to view, and doubtless more lies hid in the Church literature.

We may note first Auxentius, 305-306 (Streitberg, Gotische Bibel, xvi), concerning Ulfilas :

Grecam et Latinam et Goticam linguam sine in termis stone in una et sola eclesia Christi predicavit . . . qui et ipsis tribus Unguis plures tractatus et multas interpretationes volentibus ad utilitatem et ad aedificationetn sibi ad aetematn memoriant et mercedem post se dereliquit.

This does not state that Ulfilas completed or used a trilingual version, but that he used all three versions. That these were, however, combined in a trilingual may now be assumed, and I feel sure that with the new material available scholars will be able to prove that the Gothic version was made from a Greek-Latin bilingual and that it existed for a long time parallel to a Latin version at least, of which the best-preserved example is Old Latin f.

I quote from Professor MacDonald's translation * in Estudios de Erudicion Oriental, 1904, p. 386:

" I have seen in Cairo a codex of the Psalms in three columns, Coptic, Greek, and Arabic, and in Damascus also a codex of the Psalms in three columns, Syriac, a transliteration of Greek, and Arabic."

On page 385 Ibn al-Assal mentions a Greek-Arabic bilingual of the Gospels, and on page 387 refers to a Coptic-Arabic Bible. On page 389 he states that his translation has a Coptic interlinear over all words which are doubtful or difficult.

Furthermore, there is now on exhibition in the British Museum (Harl. 5786) a trilingual Psalter, Greek, Latin, and Arabic, of a date before 1153. Mr. Hoskier writes me that the Greek forms the first column and the Latin the middle, and that the two correspond line for line. This evidence does not, to be sure, prove that there were trilingual mss of the Gospels; but the fact that a trilingual of the Psalms still exists and that such mss were perhaps frequent in the 13th century, when they must have been relatively unnecessary, and that the known examples included Syriac-Greek as well as Greek-Latin columns, gives us a most convincing suggestion as to what must have been the condition in the earlier times when the peoples of the East were bilingual or even trilingual.

We know that bilingual (Greek-Syriac) inscriptions were common in Syria (cf. Mommse'h, Prov. Rom. Emp. vol. 2, p. 96), and Latin also must have gained a foothold in the larger cities and garrison towns.

Early Versions: circa 150 A.D.

In Egypt the conditions are known to have been quite similar. The absolute necessity of having Syriac and Coptic versions of the New Testament in spreading Christianity among the peoples of those regions will be felt by any one who has ever observed the enthusiasm with which a speaker in their own tongue is greeted by the Germans in Wisconsin, the French in Quebec, and still more the common people in Wales, though all of these understand English.

We may be sure that the rapid spread of Christianity over the Roman world was caused by or accompanied by the translation of the Gospels into the chief languages of the Empire. Yet Pliny, Ep. 10, 96, tells us that Bithynia was overrun by it before 111 A.D., and Tacitus, Ann. 15, 44, that it was widespread in Rome before 65 A.D.

The question of the date of the earliest translations of the New Testament is still undecided, but I have no hesitation in taking the side of those who claim the earliest date.

That the translations into Syriac and Latin were the earliest has been generally conceded, but some have tried to put the date late in the second century. The real difficulty with such an assumption, aside from its incompatibility with the rapid spread of Christianity before that date, is the impossibility of explaining the age, frequency, and wide distribution of N. T. text corruptions, which are best assigned to bilingual or translation influences.

Peculiarities characteristic of the bilinguals headed by codex D and by the Old Latin and Old Syriac MSS are now found in W, an old Greek MS of Egypt, and have long been known in Irenaeus and other church fathers of his time and earlier.

The characteristic features of this type of text were well established and widespread before 150 A.D., and to those who find the most acceptable explanation in the use and influence of the versions, as I do, there can be no doubt about the early date of the first NT translations. Even in the case of the Coptic translations the trend is now towards the earlier date; cf. Bousset, Text. u. Untersuch. vol. 11, p. 95.

On pages 903 ff. of his Prolegomena von Soden states that MSS K and B show influence of the Sahidic translation and, while he assumes that the Bohairic version was made later, he notes instances where the Bohairic version shows variants plainly older than Sahidic and the related K and B.

Hoskier upholds the early date of both the Sahidic and Bohairic versions; cf. his Genesis of the Versions and Concerning the Date of the Bohairic Version, London, 1911.

The date of the Sahidic version has now been definitely placed before 300 by the discovery and publication of a MS of Acts in that version, which has been dated before 350 by Dr. Kenyon on the basis of a subscription in a cursive Greek hand of that date;

cf. page Iv in the introduction to Budge's Coptic Biblical Texts in the Dialect of Upper Egypt, London, 1912. Budge also notes that the version of Acts in question was not an original translation from the Greek, but because of transcription errors must be considered a later copy. We thus gain no definite date for the Sahidic version, but 300 may now be considered the terminus ante queni.

Tatian's Diatessaron & the Old Syriac

Another equally important and difficult question is the character and amount of influence of Tatian's Harmony of the Gospels. Did it influence or was it influenced by the Old Syriac version ?

Of recent works Burkitt, Evangelion da Mepharreshe, von Soden, Die Schriften des Neuen Testaments, p. 1536, etc., and Vogels, Texte und Untersuchungen, vol. 36, la, have made the earliest Syriac translation dependent on Tatian's Diatessaron.

Vogels even supposes a Latin version of Tatian, which influenced the Old Latin version or versions, and thus explains the close relationship of the Syriac and Latin translations. His work is able as well as elaborate and will be of value to scholars because of his long lists of " harmonistic " errors catalogued on pp. 63 to 106.

To me he seems both to exaggerate the harmonistic influence and to err in his fundamental assumption that all harmonistic errors must be referred to the influence of Tatian's Diatessaron. Every one knows how easy it is for us to remember the Lord's prayer according to Matthew, while few can repeat the original form in Luke. The ancient Christians, both readers and scribes, knew their Gospels far better than we, yet they also would have remembered the words of Christ and the story of his life in a form which omitted or harmonized the differences in the accounts given in the four Gospels.

Such a reader or such a scribe was sure to make corrections in his copy of the Gospels, especially in the period before the end of the second century, when the NT canon had not been formed and the written word was not yet so rigidly adhered to as in the Old Testament. The early established habit of collecting parallel passages for lectionary use aided this harmonistic tendency.

There can be little doubt that Tatian's Harmony had an influence on the separate Gospels in those regions of Syria where it was used, but it certainly was not the cause of all harmonistic errors in mss both east and west.

Hoskier, Genesis of the Versions, chapters iv and xii, has gathered many examples showing that the first Syriac translation of the Gospels was prior to Tatian,* and I shall note a few others in the discussion below. If this view is correct, the excessive exaltation of Tatian's Harmony rests on a very insecure foundation.

This outline of the current controversies on matters affecting the NT text does not aim or hope to settle the questions under discussion, but has been introduced in order that terms to be used later may be intelligible and the evidence of W placed on the proper side in these various controversies.

As it has already been seen that there are noteworthy differences in the different parts of W, I shall discuss the text of each Gospel separately. The proof that such a course was necessary will appear from the different results arrived at in the different Gospels.

Original Footnotes:

^ Codex Sangallensis, Cambridge, 1891; Study of Codex Bezae, Cambridge, 1891. > Old Syriac Element in the Text of the Codex Bezae, London, 1893 ; Syro-Latin Text of the Gospels, London, 1895.

' Genesis of the Versions, London, 1910-1911.

A much more decisive passage occurs in Ibn al-Assal's (1252 A.D.) introduction to his Arabic version of the Gospels.^

* My attention was called to this reference by Professor Worrell of the Hartford Theological Seminary.

1 From Brit. Mus. Orient. 3382.

* Cf. also Amer. Jour, of Phil. vol. 33, p. 35.

Return to Index