James R. Davila

University (Home Page)

(c) 2003:  reproduction beyond fair use only on permission of the author.

Apocrypha Group, International SBL, Cambridge, 23 July 2003

An earlier version of this paper was posted here as an online lecture for my Old Testament Pseudepigrapha course on 1 March  in the spring semester of 2002.  This is still very much a work in progress!  A handout for this paper can be found here.


Many of the works in the Old Testament Apocrypha and Pseudepigrapha survive only in Greek (or in Greek plus in other, more exotic church languages such as Syriac and Ethiopic).  The Greek tends to have a biblical or septuagintal flavour, which has not infrequently let scholars to allege (or in some cases, less charitably, to assume) that the works in question have been translated from a Semitic language – either Hebrew or Aramaic.  Often this linguistic appraisal has formed part of the process of arguing (or again, less charitably, asserting) that the work was originally a Jewish composition.  This paper is a survey of some of the significant methodological issues that arise when we try to decide if a Greek Apocryphon or Pseudepigraphon is really a translation of a Semitic original.

The study of the possible Semitic linguistic background of Greek Pseudepigrapha has two possible agendas.  One may aim either for “retroversion,” an attempt to work backwards from the surviving Greek to reconstruct the actual wording of the original Hebrew or Aramaic document or, less ambitiously, for establishing “Semitic interference”:  showing that the Greek text must have been translated from a Semitic original, but not attempting a global reconstruction of that original.  The basic premise for both agendas is that Semitic languages have a different sound system and different vocabulary, morphology, and syntax from Greek.  Therefore, in theory, translation from a Semitic language into Greek, to the degree that the translation is literal, should reflect these differences by producing “funny,” i.e., incorrect or idiosyncratic Greek.

Over the last century or so three streams of scholarship have engaged with the problem of Greek translations of Semitic originals.  First, there are studies of the translation technique of the LXX, and attempts to work out principles for retroverting the original Hebrew or Aramaic.  Second, there is research on the question of the original languages of the traditions preserved in Greek in the New Testament Gospels and, again, attempts to establish the existence of or even retrovert the alleged Semitic (generally Aramaic) traditions or documents behind the Greek.  Third are studies of the translation technique of Greek Jewish Apocrypha and Pseudepigrapha whose Hebrew or Aramaic originals survive at least in part, as well as studies attempting to retrovert the Semitic original of specific Apocrypha or Pseudepigrapha.  In this paper I draw on all three streams of research to formulate some preliminary methodological principles for attacking the problem.

The basic approach when searching for Semitic interference or attempting retroversion is to look for signs of improper or unusual Greek in the text.  Ideally, one might find transliterations of Semitic words, nonsensical Greek that suddenly makes sense when successful retroversion shows the Greek to be an obvious mistranslation, or dual translations of a Semitic original in different Greek manuscripts.  The most common argument is to advance cases where the Greek shows non-Greek but Semitic syntax.  An example would be when the Greek shows a high level of “parataxis”:  connection of clauses with the word “and” rather than subordination, use of participles, or use of other particles.  Other arguments are also advanced, such as use of Semitic poetic parallelism and the alleged recovery of wordplays and puns in the retroverted original.  However, these types of arguments in themselves are not entirely adequate for retroversion or establishing Semitic interference.  Numerous problems, including the following, must be faced when trying to do this sort of analysis.


The inadequacy of the term “literal.”

Analysis of the LXX has shown that the types and degrees of literalness must be determined not just for the LXX as a whole, but on a case-by-case basis for individual biblical books and sometimes even for parts of these books.  Translation techniques vary widely from book to book in the LXX.  Often retroversion would be impossible for some or many specific details.  For example, if a work is free in its preservation of word order or in its consistency of lexical equivalents, one would not be able to reconstruct accurately these elements of the original.  More on this below.

The problem of distinguishing Semitic and Greek morphology and syntax.   The whole basis of the approach is to make distinctions between Semitic and Greek morphology and syntax, but much of their syntax is the same, and it is often difficult to be sure if a particular construction is Semitic rather than Greek.  Also, the available grammars of Hellenistic Greek are old and a definitive one still remains to be written.  Nevertheless, the effort must be made to distinguish Greek grammar from Semitic grammar.  Crucial work on this has been done by Raymond A. Martin, who has proposed seventeen syntactical criteria for isolating Greek that has been translated from Hebrew or Aramaic.[4]   These features appear frequently in the verifiably translated Greek of the Septuagint but are rare in works composed in Hellenistic Greek.  He has applied these criteria mainly to New Testament texts, but also to part of 1 Maccabees , a work in the corpus under consideration here.  Any attempt to argue for Semitic originals of works surviving in Greek must start from the basis of Martin’s work.  Nevertheless, some of the remaining problems complicate and indeed hinder the use of his criteria.

The problem of interference from the LXX.   The LXX was widely available in Jewish circles during the Hellenistic period and its presence had an effect on Jewish Greek (much the same way as the King James translation of the Bible has affected subsequent English).  Therefore, apparent Semitisms in Greek works could be stylistic features imitating the LXX.  So expressions found frequently in LXX Greek as well as direct allusions to specific LXX passages are best discounted when we look for Semitic interference.  Martin claims that his seventeen criteria are “not the kinds of syntactical features which would be readily chosen if a writer should seek deliberately to imitate translation Greek style,”[5] yet all of them seem to me to be readily imitable by a writer who knows the LXX well and who strives to write Greek in a biblical style.

The problem of bilingual interference.   If a native speaker of Hebrew or Aramaic were to compose a text in Greek, it is entirely possible – likely, even – that the writer would produce a Greek text containing elements of Semitic interference purely because he or she thought in a Semitic language.  Ideally, one would have to show mistranslations, misunderstanding of homonyms, or dual translations in the Greek to prove that the Semitic features were not due to bilingual interference alone.  Moreover, the Hellenistic Greek nonliterary papyri from Egypt are often compared to New Testament Greek, but it has been pointed out by Elliott C. Maloney that they are likely to suffer from Egyptian interference – assimilation of features from the spoken Egyptian language that has much in common with the Semitic languages.[6]   So if an apparent Semitism appears in a Greek work from the New Testament or the Apocrypha or Pseudepigrapha but also in the Egyptian literary papyri, it does not automatically follow that it is an internal Greek usage.  Once again, Martin asserts that his seventeen criteria are not “the aspects of the language which would colour the speech or writing of a person who is thinking in Aramaic or Hebrew as he writes Greek.”[7]   I am not sure how he can be so confident of this.  Someone thinking in one of these Semitic languages might well use awkward Greek with many of the features in these criteria.

The problem of interference from liturgy and testimonia.   Even rare expressions from the LXX can affect early Jewish and Christian compositions if they are from scriptural passages commonly used in liturgy or in collections of scriptural testimonia.  We must look closely at possible Semitisms to see if they appear in LXX passages used elsewhere in this way.[8]

Semitic stylistic features and poetic forms must be used very cautiously, if at all, to argue for Semitic interference.   Semitic poetic canons in this period were loose and flexible and did not include rhyme.  Alliteration, assonance, and puns are subjective features and were not formal parts of Semitic poetry.

Reconstructed Hebrew or Aramaic must be in the dialect from the right time and place.  For example, attempts to reconstruct the putative Aramaic behind parts of Mark’s Gospel should be based on the vocabulary, morphology, and syntax of Qumran Aramaic rather than the Aramaic of the later Targums.

It is very difficult to distinguish translated Hebrew from translated Aramaic, since the vocabulary and grammar of the two languages are so similar.   It is true that there are some points of distinction that may carry over into a translation.  Hebrew has a construction involving the infinitive construct with a preposition and this construction is lacking in Aramaic but is often reflected in Greek translations of Hebrew.  A Greek translation with examples of this construction would be unlikely to have come from an Aramaic Vorlage .  Likewise, R. H. Charles successfully deduced that much of the Book of the Watchers was composed in Aramaic, in part because transliterated words in the Greek translation reflected the long “a” of the emphatic (definite) ending, found in Aramaic but not Hebrew.[9]

Methodology and Prospects

It is fair to say that the prospects for retroverting a Greek text into anything close to its original Hebrew or Aramaic are quite poor.  To do so would require a translation that was consistently literal in all the aspects discussed above.  Emanuel Tov has shown that there are many elements of the LXX which we cannot reliably retrovert, such as the connective conjunction waw / kai , singular vs. plural forms of nouns and verbs, the presence or not of possessive pronouns, active vs. passive forms of verbs, equivalents for prepositions, and the appearance or not of the definite article.[10]   Since translations often translate a given lexeme in different ways according to sense and context, we would frequently be unsure of the wording of the original, and since translations rarely preserve the segmentation of lexemes or the word order of the original with care, we could not consistently reconstruct these elements either with any confidence.  These factors have led Benjamin G. Wright to be sceptical of our ability to reconstruct the parent text of Ben Sira where it is no longer extant.[11]   Likewise, regarding the Greek translation of 1 Enoch , Erik Larson judges “that the translator made some attempt to produce a literary and idiomatic version of the Book of Enoch and [the features that indicate this] put this translation well into the category of freely translated books of the LXX such as Job, Proverbs, and Isaiah.”[12]   Accurate retroversion of such a translation would be very difficult.

            Nevertheless, it presumably remains possible to produce a retroversion of a Semitic original of many works which is something like what an ancient writer might have written in Hebrew or Aramaic.  Moreover, with the advent of hypertext technology, one could envisage producing a reconstructed text that aimed to provide all possible Semitic retroversions of each word, phrase, sentence, etc. of a translated Greek work, with filters that would weed out impossible combinations.  Such a “brute force” retroversion would be immensely complex and would require a huge amount of work, but somewhere in its myriad possibilities would indeed lurk something very like the original Hebrew or Aramaic text.

If we move from retroversion to the comparatively more tractable problem of establishing translational Semitic interference in a Greek text, a strict methodology would need to adhere to the following steps.  One needs to look at all possible linguistic and stylistic features (such as vocabulary, morphology, syntax, word order, and poetic elements) for apparent Semitisms – Semitisms that are compatible with the Hebrew and Aramaic of the right time and place.  Then one must eliminate all those that are shared with Greek, that appear frequently in the LXX, and that are found in LXX passages used in liturgy and testimonia.  If the nonliterary papyri are used for comparison, one should not eliminate apparent Semitisms found in them which are also found in Coptic, since they may be due to Egyptian interference in the Greek.  Poetic and stylistic features should be advanced – if at all – only as ancillary evidence.  If possible, and it may well not be, one should distinguish Hebrew Semitisms from Aramaic Semitisms.  Finally, one would need to show mistranslations, mishearing of homonyms, or dual translations in the Greek to establish that we are dealing with Semitic interference from a translated original rather than bilingual interference from the composer’s native language.  This approach follows, with some nuancing and development, Maloney’s Semitic Interference in Marcan Syntax , the most fully methodologically aware treatment of Semitic interference in a Greek text of which I know.  If a cumulative case for a high density of Semitisms remains after such an analysis, the probability that the work was translated from Hebrew or Aramaic would be very high.

One could argue that this is an excessively rigorous methodology that is really designed for the analysis of pericopes of a few verses in the New Testament; that it is overkill to insist it be applied to whole books; and that applying Martin’s seventeen syntactical criteria should be adequate to establish Semitic interference in the case of whole books.  I have some sympathy with this view, especially since it would at present be difficult perhaps to the point of impossibility to apply the more rigorous criteria to a substantial Greek work in the Apocrypha or Pseudepigrapha, both because we lack a modern, comprehensive grammar of Hellenistic Greek from which to establish the norm and departures from it and because a morpheme-by-morpheme analysis of the sort envisaged would best be done with a tagged, searchable text interfaced with the comprehensive grammar.  Fortunately, the Open Text Project, created by Brook O’Donnell, Stanley E. Porter, and Jeffrey T. Reed, is in the process of producing just such a Greek grammar and text archive and it is hoped that it will be finished within the next five years or so.

Meanwhile, even applying Martin’s criteria in a thoroughgoing way to texts in the Apocrypha and Pseudepigrapha would be a significant step forward.  Martin has done so with the first five chapters of 1 Maccabees and the high level of correspondence of its Greek to the Semitic features makes a persuasive case that it was composed in Hebrew or Aramaic.[13]   It is fairly difficult to argue that the correspondences are due to influence of the LXX or bilingual interference in this case.  Nevertheless, comparable cases for other works remain to be made.  It has been argued, for example, that the Greek book of Baruch was translated from Hebrew and efforts have even been made to retrovert the Hebrew original.  But the high density of septuagintalisms in the work at least leaves open the possibility that it was actually composed in a deliberately biblicizing Greek.[14]

Unambiguous cases where whole works correspond closely to Martin’s seventeen criteria are likely to be more persuasive than not.  Ambiguous cases, where the evidence could be read either as indicating a free or paraphrastic translation of a Semitic text or some effort to write in septuagintal Greek or that the writer was thinking in a Semitic language while composing in Greek, would not persuade.  These would need to be analysed according to the more rigorous criteria outlined above, and even then the result might not always be decisive.

Establishing the existence of a Hebrew or Aramaic original of Pseudepigrapha that survive only in secondary and tertiary translations (e.g., 4 Ezra and 2 Baruch ) would be correspondingly more difficult and would depend on their being translated literally into Greek and then translated slavishly enough into the secondary language – Latin, Syriac, or whatever – that the Hebraisms or Aramaisms still show through.  I confess myself sceptical of the confident and widely accepted claims to have established the original languages of such works.  Retroversion of the originals of these works would, of course, be vastly more difficult.  Establishing other kinds of linguistic interference (e.g., translation of Greek into Latin or Syriac into Arabic) would have different sets of challenges, some of which would be analogous to those considered in this paper.


Many of the tools needed to establish Semitic interference in or attempt retroversion of the Greek Apocrypha and Pseudepigrapha already exist, having been developed by specialists in the LXX and the Gospels.  For the most part these have not been properly deployed in previous analyses of these works.  A fully rigorous philological analysis of the Greek Apocrypha and Pseudepigrapha must await computing tools such as those currently being developed by the Open Text Project, but significant progress could be made in the interim by a cautious use of Martin’s methodology.  Most of these previous attempts at retroversion or establishing Semitic interference need to be reconsidered in this light.

  1. Ronald E. Hine, Origen:  Homilies on Genesis and Exodus (FC 71; Washington, D.C.:  Catholic University of America Press, 1982) 30-40.
  2. For the relationship of the Greek Testament of Levi to Aramaic Levi see M. de Jonge, The Testaments of the Twelve Patriarchs:  A Study of their Text, Composition and Origin (2nd ed.; Amsterdam:  Van Gorcum, 1975), 38-52; Robert A. Kugler, From Patriarch to Priest:  The Levi-Priestly Tradition from Aramaic Levi toTestament of Levi (SBLEJL 9; Atlanta, Ga.:  Scholars Press, 1996).
  3. The Typology of Literalism in Ancient Biblical Translations (MSU 15; G¿ttingen:  Vandenhoeck & Ruprecht, 1979).
  4. Syntactical Evidence of Semitic Sources in Greek Documents (Cambridge, Mass.:  Society of Biblical Literature, 1974); idem, Syntax Criticism of the Synoptic Gospels (Studies in the Bible and Early Christianity 10; Lewiston, New York/Queenston, Ontario:  Edwin Mellen, 1987).  The criteria are presented in detail in chapter 1 of Syntatical Evidence on pp. 5-43 and are, briefly, the relative frequency of eight prepositions in relationship to the preposition en ; the comparative frequencies of kai and de in coordinating independent clauses; the separation of the Greek definite article from its substantive; a tendency to place genitives after the substantive on which they depend; a greater frequency of dependent genitive personal pronouns; a tendency to omit the article on a substantive with a dependent genitive personal pronoun; a tendency to place attributive adjectives after the word they qualify; less frequent use of attributive adjectives; less frequent use of adverbial participles; and less frequent use of the dative case without a preposition.
  5. Syntactical Evidence , 2.
  6. Semitic Interference in Marcan Syntax (SBLDS 51. Missoula, Mont.: Scholars Press, 1980), 38.
  7. Syntactical Evidence , 2.
  8. Max Wilcox, The Semitisms of Acts (Oxford:  Clarendon, 1965), 56-86.
  9. APOT 2:172-77.  However, he incorrectly concluded that chapters 1-5 were composed in Hebrew.  Only Aramaic fragments of this material has been found at Qumran.
  10. The Text-Critical Use of the Septuagint in Biblical Research (2nd ed.; Jerusalem:  Simor, 1997), 154-62.
  11. No Small Difference: Sirach’s Relationship to Its Hebrew Parent Text(SBLSCS 26; Atlanta, Ga.:  Scholars Press, 1989), 231-50.
  12. “The Translation of Enoch:  from Aramaic into Greek” (Ph.D. diss., New York University, 1995), 349.
  13. Martin, Syntax Criticism , 142, 168-70, 189-91.
  14. Emanuel Tov, The Book of Baruch:  Also Called I Baruch (Greek and Hebrew) (SBLTT 9; SBLPS 6; Missoula, Mont.:  Scholars Press, 1975); David G. Burke, The Poetry of Baruch:  A Reconstruction and Analysis of the Original Hebrew Texts of Baruch 3:9-5:9 (SBLSCS 10; Chico, Calif:  Scholars Press, 1982).

  In translator circles, “literal” is generally used in contrast to something along the lines of “dynamic equivalence,” and implies a one-to-one correspondence between all the grammatical and lexical elements in the original language and the target language.  On the one hand, a very literal translation in this sense would be almost unreadable (an example is “Aquila’s” Greek translation of the Hebrew Bible at its worst).  On the other hand, a very free translation could still be very good to the degree that it is “dynamic,” that is, that it captures the meaning of the original on a thought-by-thought basis.  There was a wide range of translation techniques used for scriptural and quasi-scriptural literature in antiquity.  If we take the poles “literal” vs. “free” in their most general senses and on a very rough-and-ready level, Aquila’s translation belongs at the farthest end of the literal pole, although his techniques are not as consistently literal as is sometimes supposed.  The LXX of the Pentateuch moves more in the direction of the idiomatic, while remaining closer to the literal pole than not.  The LXX of Isaiah is freer still, moving at times into the realm of paraphrase.  Rufinus’s translations of the works of Origen into Latin may be regarded as still more free, moving beyond mere paraphrase and deliberately altering the text to suppress contradictions and to purge supposed heretical changes.[1]   Finally, the translation, if we may call it that, of Aramaic Levi now preserved in the Greek Testament of Levibelongs on the far end of the free pole.  The Greek text of the Testament of Levisometimes translates Aramaic Levi fairly directly but it not only paraphrases and makes theological changes, it reworks the earlier text into what amounts to a new composition.[2]   However, these two poles are entirely too simplistic, as has been demonstrated at length by James Barr in a brief but important monograph.[3]   More precision is needed when discussing literal vs. free translation.  Translations can be literal or free in various ways:  lexically, in their consistent or inconsistent use of equivalents for individual words, in the degree to which they retain the correct semantic range of translated terms, and in the degree to which they deploy true and false etymological relationships to translate homonyms and etymologically related words; morphologically, in their degree of representation of grammatical elements within words, to the degree they preserve the word order of the original, and to the degree they add or subtract elements to or from the original.  A typology of literalism in ancient translations such as that roughed out above would have to be nuanced with these elements in mind in order to be truly accurate and useful.



James R. Davila


Apocrypha Group, International SBL, Cambridge, 23 July 2003


Retroversion vs. establishing Semitic interference

Three streams of scholarship

Translation technique of LXX

Semitic/Aramaic background of the Gospels

Translation technique and retroversion of specific Apocrypha and Pseudepigrapha


           The inadequacy of the term “literal”

“literal” vs. “dynamic”


Aquila -LXX Pentateuch – LXX Isaiah – Rufinus/Origen – Greek T. Levi


Barr:  lexical :  consistent or inconsistent use of equivalents for individual words, retention of correct semantic range of translated terms; deployment of etymology; morphological :  degree of representation of grammatical elements within words; preservation of original word order; addition or subtraction of elements to or from the original.

           The problem of distinguishing Semitic and Greek morphology and syntax

Martin’s 17 syntactical criteria

  • The problem of interference from the LXX
  • The problem of bilingual interference
  • The problem of interference from liturgy and testimonia
  •  Semitic stylistic features and poetic forms must be used very cautiously, if at all, to argue for Semitic interference
  • Reconstructed Hebrew or Aramaic must be in the dialect from the right time and place
  • It is very difficult to distinguish translated Hebrew from translated Aramaic, since the vocabulary and grammar of the two languages are so similar

Methodology and Prospects

Retroversion – prospects are poor – hypertext “brute force” retroversion?

Establishing translational Semitic interference

Rigorous methodology:  perhaps excessive, currently difficult perhaps to the point of impossibility.  The Open Text Project (

            Martin’s criteria and its limitations

Pseudepigrapha that survive only in secondary or tertiary translations (e.g., 4 Ezra 2 Baruch )



A fully rigorous philological analysis must await tools like those being developed by the Open Text Project, but progress can be made in the interim by a cautious use of Martin’s methodology.


Barr, James.  The Typology of Literalism in Ancient Biblical Translations.  MSU 15.  Göttingen:  Vandenhoeck & Ruprecht, 1979.

Larson, Erik W.  “The Translation of Enoch: from Aramaic into Greek.”  Ph.D. diss., New York University, 1995.

Maloney, Elliott C.  Semitic Interference in Marcan Syntax.  SBLDS 51.  Missoula, Mont.:  Scholars Press, 1980.

Martin, Raymond A.  Syntactical Evidence of Semitic Sources in Greek Documents.  Cambridge, Mass.:  Society of Biblical Literature, 1974.

__________.  Syntax Criticism of the Synoptic Gospels.  Studies in the Bible and Early Christianity 10.  Lewiston, New York/Queenston, Ontario:  Edwin Mellen, 1987.

Tov, Emanuel.  The Text-Critical Use of the Septuagint in Biblical Research.  2nd ed.  Jerusalem Biblical Studies 3.  Jerusalem:  Simor, 1997.

Wilcox, Max.  The Semitisms of Acts.  Oxford:  Clarendon, 1965.

Wright, Benjamin G.  No Small Difference:  Sirach’s Relationship to Its Hebrew Parent Text.  SBLSCS 26.  Atlanta, Ga:  Scholars Press, 1989.

(c) 2003
Reproduction beyond fair use only on permission of the author.

Contact details

St Mary’s College
The School of Divinity
University of St Andrews
South Street
St Andrews
Fife KY16 9JU
Scotland, United Kingdom

Tel: +44 (0)1334 462850 
Fax: +44 (0)1334 462852