Codex Bezae and Latin-Trained Syriac Scribes

Last week I presented a paper at the Eleventh Birmingham Colloquium on the Textual Criticism of the New Testament, as announced in my previous post. In the paper, I examined five classic objections to the theory that Codex Bezae’s Greek text form has appropriated text forms from the Latin version, concluding that none of these objections entirely succeeds in excluding the possibility that Bezae’s collection of distinctive Latin parallels might have arisen as readings in the Old Latin version.

In the discussion following my talk, a sixth objection was raised, namely, that parallels between the Old Latin and Old Syriac suggest that a common source lies behind them both, presumably resembling Bezae’s Greek text.

In my answer, I noted that Bezae’s text form is only distantly related to the Old Syriac, at least in Mark.1 In Mark, for example, Bezae’s text is more closely aligned with the text of Codex Vaticanus. So Bezae is not a good candidate to represent the source of the Old Syriac, at least in this gospel.

Moreover, in my paper, I pointed out that Bezae is unlikely to represent the source of the Old Latin version, which requires no Bezan source to explain its development. Its text forms, where they differ from Greek witnesses, likely arose through translation and inner-versional copying.

So if the Old Latin and Old Syriac require a common source, which is doubtful, a text like Bezae’s is probably not it.

Syriac Traditions in a Greek Text that Borrows Latin Readings

Nevertheless, Bezae’s text form does have a relationship with Syriac traditions, apparently dating to the time after Aphrahat (d. c. 345), as I argue in my essay on Codex Bezae’s Lukan Genealogy, which depends on a list of names found only in Aphrahat. In this case, it is clear that Bezae is not the source of the list in Syriac, because Bezae develops Aphrahat’s list of names, even introducing an erroneous name in the process. So Bezae’s contact with Syriac traditions likely occurred in the latter fourth century, after the time of Aphrahat.

But how are we to explain Syriac contact in a Greek manuscript, whose text form seems to borrow from the Latin version?

One explanation relates to contact between the Greek and Latin languages in the East, where Bezae was likely produced.2 There was little if any true bilingualism in the East at the time of Bezae’s production. The situation is best described as diglossia, the side-by-side coexistence of languages in separate domains, where the use of Latin was confined primarily to official contexts, namely, the military, the civil administration, and the law.3

If Bezae was produced in such a context, it is unlikely to have been prepared for a bilingual congregation. Even in centers of Roman culture in the East, such as Berytus, the churches were Greek speaking.4

It is likely then that the Latin version of Bezae’s Latin parallels and Latin column was approached as a written source, much like the Latin legal documents of the law school. Indeed, Bezae’s legal script supplies a valuable clue that its production took place at or near a legal center, such as Berytus, or among individuals who had received their training at such a legal center.5

John of Palestine: A Latin-Trained, Syriac-Speaking Scribe

In his account of the Christian community at Berytus in the latter fifth century, Zacharias of Mytilene describes a certain John of Palestine, a monk at the monastery of Saint Jude, just outside of Berytus, who

after training in the law, had consecrated himself to God in that church [of Jude], adopting the philosophical life-style … [and had] benefited many of those studying law in that city, … because of the (collection of) Christian books that he owned, which he shared and gave out.6

The fact that John had amassed a collection of books suggests that he was a scribe. After all, an individual ascetic would have had little other means by which to acquire books.

As a well-educated native of Palestine, we can assume that John was fluent in both Greek and Aramaic, while as a former law student, he would have been literate in both Greek and Latin. So in John of Palestine we find a Latin-trained, Syriac-speaking scribe.

Peter of Iberia’s Recruitment of Latin-Trained Monks in Palestine

Apparently, John was not alone among the Latin-trained monks of Syria and Palestine. In his account of Peter the Iberian, John Rufus describes Peter’s deliberate recruitment of law students to his monastic community:

When he [Peter] came to Beirut, he was recognized immediately by the young lawyers, his acquaintances from Palestine and from Alexandria … Some of them while he was still alive and some after his departure would reject the vanity of the world and run to him.7

Among Peter’s law recruits identified by Rufus (who was also trained at Berytus) were Theodore of Ascalon, Anastasios of Edessa, and Elisha of Lycia, all former students of the law school at Berytus. Rufus describes how Peter had received a vision instructing him specifically to recruit lawyers to his monastic community:

God wanted that there should be offered to him also from [among] the lawyers rational sacrifices and whole burnt offerings, able to carry his cross and follow him, like Basil and Gregory and John and Arsenius, and those who are like them in zeal.8

But every one of these recruits must have known Latin, because Roman law was promulgated in Latin, even in the East, down through the time of Justinian. So we cannot assume that Latin was wholly unfamiliar among the monks of Syria and Palestine. A Greek text such as Bezae’s, accompanied by a Latin column, would not have been entirely inaccessible in the Syriac world among those trained in law who inhabited its monastic communities.

Bezae’s Text and Miaphysite Monasticism in Syria, Palestine, and Egypt

Therefore, when we encounter errors in Bezae suggesting that its producers were not fully competent in literary Latin and Greek, a natural inference is that their native language was neither Greek nor Latin, but a third language, such as Syriac.9 Here we find a second clue as to Bezae’s origins, linking it potentially with the monastic communities of Syria and Palestine (and, ultimately, Egypt) among those trained in Roman law.

As the Chalcedonian settlement split apart the Eastern church, it is not difficult to imagine such a text ending up among the miaphysite communities of middle Egypt, where something like it appears in Codex Glazier, and still later in the Harklean marginal notes of Acts, produced outside of Alexandria by the Syrian miaphysite, Thomas of Mabug. So exchange between the monastic communities of Syria, Palestine, and Egypt supplies one explanation for the appearance of Bezan text forms in Syriac and Coptic. This more refined theory is at least as plausible as the usual theory that Bezae’s text form represents the original source behind both the Latin and Syriac versions.

Ephrem’s Typological Genealogy of Mary

In two recent posts, I noted the possibility that Justin Martyr and Celsus understood one of the two genealogies of Jesus given in the gospels as Mary’s genealogy. While such a view is at best implied by these early writers, Ephrem is remarkably explicit in his apparent attribution of the Matthean genealogy to Mary:

“Matthew wrote down the genealogy of Mary from whom our Lord was born and, because of this, he began from David and from Abraham, according to the promise revealed, ‘Not to you and to your descendants as though unto many, but to you and to your descendant, which descendant is Christ’ (Gal 3:16).”1

While Justin and Irenaeus imply that it is the Lukan genealogy that belongs to Mary, Ephrem somewhat surprisingly designates Matthew’s royal genealogy “the genealogy of Mary.”

Now there are some puzzling aspects about this passage in the Syriac Commentary on the Diatessaron. For one, Ephrem is citing the evangelist Matthew in a Diatessaron commentary. This is odd. Since Tatian had combined the four gospels into one, it is remarkable that Ephrem calls out one of the four evangelists by name.

But perhaps this is because of a second puzzle: the Diatessaron purportedly contained no genealogy of Jesus.2 It is remarkable that Ephrem senses a need to comment on a passage that was apparently excluded from the work upon which he is commenting. Indeed, Ephrem’s explicit reference to Matthew (and later to Luke) appears to support the notion that the Diatessaron lacked the genealogies.

But the third puzzle is Ephrem’s surprising attribution of the Matthean genealogy to Mary.

This attribution is all the more surprising because just two paragraphs earlier Ephrem asserts that Mary’s family line is nowhere to be found in the scriptures and, moreover, he implies that it is inappropriate to look for it:

“Scripture is silent [about Mary] since it numbers and counts the generations of men. If Scripture were accustomed to represent the family through the women, it would be appropriate to inquire after the family of Mary.”3

But apparently, in Ephrem’s view, it is not appropriate. Otherwise, why bring the matter up? But if not, why does Ephrem later apparently contradict himself by explicitly ascribing the Matthean genealogy to Mary?

If this is in fact a contradiction, it will not have been the first incongruity to have been discovered in the Syriac Commentary. Similar incongruities of thought have led to the conclusion that the commentary preserves multiple layers of tradition, ultimately compiled by a “school” of Ephrem in the latter fourth century.4

But perhaps this is not an incongruity.

Matthew: A Typological Genealogy?

In the same context, Ephrem refers to the genealogy of David as a “type” of the true genealogy of Mary’s child:

“The family of David continued as far as Joseph who had married her, and her child was through the type of the men, for the sake of the family of David.”5

Is Ephrem spiritualizing the Matthean genealogy?

According to Ephrem, the genealogies of the men establish a pattern of the true genealogy, which is Mary’s genealogy, because she is the one who gave birth to Christ, the seed promised to Abraham and David, as Ephrem reminds us:

“He is the son of Mary, however, and not the son of Joseph.”6

Ephrem says that because “Matthew wrote down the genealogy of Mary,” therefore “he began from David and Abraham.” So according to Ephrem, Matthew’s intent was to set forth the spiritual promise to Abraham and to his seed, Christ, and not to record a literal genealogy.7

Yet even to Ephrem, Matthew’s genealogy is not totally spiritualized. It contains three crucial names in the physical lineage of Jesus: Abraham, David, and Mary, “of whom Jesus was born.”8 If we follow this spiritual path, the genealogy becomes Mary’s genealogy, because, after all, the child is hers. And the seed is literally from David because, as Ephrem insists, Mary was from David:

“take note that [the evangelist] has said [elsewhere], concerning Joseph and Mary, that They were both of the house of David.”9

Here again Ephrem reaches outside of the Diatessaron, referring to Luke 2:4, which in the Old Syriac asserts that both Mary and Joseph were of the house of David.10

Whose Names?

So according to Ephrem, whose names are in the Matthean genealogy, Mary’s or Joseph’s?

Apparently, neither, at least after David. We know that Ephrem does not consider the Matthean names to belong to Joseph, because he later assigns the Lukan genealogy explicitly to Joseph:

“Luke however [was concerned] only with Joseph, husband of Mary, and [went back] as far as Adam who is from God.”11

So Ephrem chooses not to interpret the Matthean genealogy in a literal sense. Rather he interprets it typologically, as a figure of the true but unknown genealogy of Mary.

Is this Ephrem’s own solution to the discrepancies between the two genealogies of Matthew and Luke? It is difficult to say. But Ephrem is certainly not the first — nor the last — early Christian writer to spiritualize a difficult passage. Typological interpretations became one defense against the literalistic readings of certain heretics, such as Marcion, Ephrem’s arch-heresiological foe.12 From Ephrem’s perspective, Marcion’s literalistic interpretation ended up pitting one testament against the other and ultimately splitting the godhead into demiurge and stranger. Considered from this perspective, Ephrem’s pious spiritualization of the Matthean genealogy does not seem terribly shocking.

How old are the Old Latin gospels?

In the second edition of his Introduction and Appendix (1896), B. F. Westcott (1825-1901) suggested a novel criterion to assess so-called external evidence, namely, that the convergence of the Old Latin manuscript 1 (k) with the recently-discovered Sinaitic Syriac palimpsest (sys) could signal a text form of greater antiquity than the combined witness of Codices Vaticanus and Sinaiticus:

“The discovery of the Sinaitic MS of the Old Syriac raises the question whether the combination of the oldest types of the Syriac and Latin texts can outweigh the combination of the primary Greek texts. A careful examination of the passages in which syr.sin and k are arrayed against אB would point to the conclusion.”1

Christened a “methodological formula” by A. Vööbus and, rather dramatically, a “deathbed confession” by W. L. Petersen, Westcott’s “criterion” gained some acceptance, judging from the range of scholars who cited it.2 Yet we wonder whether — with F. J. A. Hort (1828-1892) no longer in the picture as a voice of restraint — Westcott’s remarks reflect a degree of sensationalism following Agnes Lewis’s recent discovery of the Sinaitic Syriac palimpsest in 1892.

No doubt a certain mystique surrounds the early versions. This aura of antiquity can be felt in their weighty titles, Vetus Latina and Vetus Syra. It is only enhanced by the abundance of free text forms they attest, thought to evoke the earliest stages of the New Testament text. Indeed, the Latin and Syriac versions are cited in support of a variety of seemingly ancient traditions, from the pericope adulterae to reports of a light or fire on the Jordan to a range of other traditions. Of course, since the attesting manuscripts tend to be much later than the period of interest, it is routinely noted that the dates of manuscripts matter less than the dates of the texts they contain, which of course may be several centuries older.

But how certain can we be that these particular texts as found in the Old Latin and Old Syriac gospels are very much older than their manuscripts? How reliable is the impression of age of their text forms? Can they somehow transport us behind the Greek tradition?

Examining the evidence, it is difficult to see how. Bobbiensis’s text likely dates to a generation or so before Cyprian, perhaps to the 230’s, as Tertullian betrays no knowledge of it.3 The Sinaitic Syriac text is difficult to date, but while perhaps influenced by earlier material seems itself not much earlier than the fourth century — not much earlier indeed than the Greek witnesses it is said possibly to preempt.4 And turning to the manuscripts themselves, we find in the case of the Old Latin that they date mainly to the fifth century:5

MS/Nr Name Contents Date Place
e 2 Palatinus Gospels V Italy
b 4 Veronensis Gospels Late V Italy
d 5 Bezae Cantabrigiensis Gospels,Acts c. 400 Berytus
ff2 8 Corbiensis Gospels V Italy
h 12 Claromontanus Matthew Late V Italy
n 16 Sangallense Gospels V Italy
i 17 Vindobonensis Luke,Mark Late V Italy
t 19 Bernense Mark Late V Italy (?)

Now at the end of the fourth century, the diversity of the Old Latin tradition is mentioned by more than one writer. Jerome writes in the 380’s:

“the Latin manuscripts of the Scriptures are proved to be faulty by the variations which all of them exhibit.”6

And again:

“We must confess that as we have it in our language it is marked by discrepancies, and now that the stream is distributed into different channels we must go back to the fountainhead [i.e. the Greek].”7

Writing about a decade later in 397, Augustine makes a similar observation, also conceding the priority of the Greek:

“As to the books of the New Testament, again, if any perplexity arises from the diversities of the Latin texts, we must of course yield to the Greek.”8

Augustine tries hard to cast these divergent translations in a positive light:

“the great number of the translators proves a very great assistance, if they are examined and discussed with a careful comparison of their texts.”9

But while Jerome was apparently content to make his own revisions (at least in the gospels), Augustine urges anyone with the requisite skills to correct texts:

“For those who are anxious to know, the Scriptures ought in the first place to use their skill in the correction of the texts, so that the uncorrected ones should give way to the corrected.”10

So as the fourth century drew to a close, we find a context in which various correctors are at work on the Old Latin texts, attempting to correct them to a Greek model.

Yet while Augustine and Jerome could at least agree that Latin texts should be corrected to the Greek, their opinion was by no means the consensus view. Ambrosiaster argues the contrarian position that it is rather the Old Latin that preserves the most ancient available text form, because, as he argues, it agrees with the earliest Christian writers in the Latin tradition:

“it is well-known that very long ago native Latin speakers translated the text we now have from ancient Greek codices, which the innocence of former times has safeguarded and now certifies to us without corruption. … For the text retained today in the Latin codices is found the same in the ancients, Tertullian, Victorinus, and Cyprian.”11

The picture emerges then of a tradition still in flux at the turn of the fifth century, whose text had lost the confidence of its readers, who at the same time sought to affirm its value through their corrections. But with no common criteria for “correcting” the text, the instability of the tradition as a whole could only have been increasing at this time. Given that most of our best witnesses to this tradition date from after this period of instability, we are left wondering how we are to distinguish between ancient traditions and recent corrections. But if this is so, how can these traditions offer us any useful glimpse behind the Greek text?

Has Aramaic influence on Codex Bezae been disproved?

In two recent posts, I have considered theories of the origins of Codex Bezae’s Greek text based on assimilation to one or more strains of the Old Latin version. But in his Textual Commentary, B. M. Metzger also lists several theories of Semitic influence, including those of F. H. Chase, J. Wellhausen, C. C. Torrey, A. J. Wensinck, and M. Black. [1]

One intriguing theory is that of C. C. Torrey, who argued that Bezae developed from a retroversion of an Aramaic Targum of the gospels. Torrey writes:

“There is very good reason to believe that an especially able and complete (also occasionally expanded) retroversion into Aramaic, an ‘original’ gospel very widely celebrated in its time (early second century?) and therefore translated into Greek with constant employment (from memory?) of the wording of the standard Greek text of that day, was the origin of our Codex Bezae and the ‘Western’ text.” [2]

Torrey’s theory is appealing for a number of reasons. For example:

  1. If we accept that Bezae’s features imply a Semitic influence, then a historical context among Aramaic-speaking Jewish Christians supplies a compelling motive for an initial translation that might explain some of Bezae’s variation.
  2. The expansive character of Bezae’s text is consistent with the interpretive and storytelling elements we expect to find in a Targum. [3]
  3. The existence of gospel materials in Aramaic is supported by the documentary evidence of later writers.
  4. The existence of written Targums is not without precedent in the time frame proposed by Torrey. [4]

Now it is often thought that J. D. Yoder’s study, “The Language of the Greek Variants of Codex Bezae Cantabrigiensis,” effectively excluded the possibility of distinctive Semitic influence on Bezae’s text. [5] By evaluating Bezae’s readings both for and against three broad categories of Semitic forms, Yoder found that the number of Semitic forms present in Bezae but lacking in Westcott and Hort (= WH) was generally offset by the number present in WH but lacking in Bezae, leading him to conclude that Bezae’s Semitisms are no more representative of its text than of the larger tradition.

But Yoder’s experimental design does not rule out a significant class of theories of Semitic influence in Bezae. Specifically, it does not rule out any theory (such as Torrey’s Targum theory) that involves initial translation into a Semitic language followed by retroversion into Greek.

Why is this the case? It is simply that any Semitisms that might have existed in the initial Greek Vorlage would have been subsumed into the text of the Aramaic translation. When the text was re-translated into Greek, these original Semitisms would now have become indistinguishable from the surrounding text. In the process, the Greek retroversion would likely acquire new Semitisms or new varieties of the original Semitisms. But unless copied from an existing Greek codex, we would not expect the original Semitisms to be restored with any consistency to their former places. In fact, the expected result if Torrey’s theory were true is remarkably close to Yoder’s actual result.

So what does this mean for Torrey’s theory?

It is important to realize that Yoder’s test does not disprove Torrey’s theory or any similar theory involving rewriting the Greek text from an intermediate Semitic stage. Yet neither is it a demonstration of such a theory.

Like nearly all theories that propose the early development of a “Western” text form, a weakness of the Targum theory is its reliance on gaps in our knowledge of the earliest period of Christianity. Yet while perhaps not altogether compelling, the Targum theory highlights a potentially significant aspect of the evidence that we cannot simply dismiss, namely, the presence of certain apparent Semitic influences in Bezae’s text that are not in the mainstream text.

If Yoder intended to show that Bezae’s apparent Semitic elements need not be considered in reconstructions of its textual history simply because they do not occur in other manuscripts, this objective was not achieved. So long as distinctive elements are found in Bezae but not in the remaining tradition, they must be accounted for in any comprehensive theory of its text.


[1] B. M. Metzger, A Textual Commentary on the Greek New Testament (2nd ed.; Stuttgart, 1994) 231-232.

[2] C. C. Torrey, The Four Gospels: A New Translation (New York, 1933) 282.

[3] Metzger’s comment that “such an hypothesis … offers no help in explaining how the Bezan text of Acts became nearly one-tenth longer” (p. 232) is difficult to understand. In fact, the Targum theory should be able to account for significant expansions to the text.

[4] We might consider, for example, the Targum of Job from Qumran.

[5] J. D. Yoder, “The Language of the Greek Variants of Codex Bezae Cantabrigiensis,” Ph.D. dissertation, Princeton Theological Seminary, 1958. See also, J. D. Yoder, “Semitisms in Codex Bezae,” Journal of Biblical Literature 78 (1959) 317–321.