1. A Global Census?

Few historical claims in the canonical Gospels have excited such puzzlement and even derision among commentators as Luke’s assertion that Jesus was born amid an Empire-wide census: ‘And it happened in those days that a decree went out from Caesar Augustus that all the inhabited world (πᾶσαν τὴν οἰκουμένην) should be enrolled. This was the first enrolment, when Quirinius was governor of Syria’ (Lk 2.1–2).Footnote ¹ LukeFootnote ² is of course right to say that Quirinius undertook a census shortly after taking office, though he also seems to misdate this event to the reign of King Herod (d. 4 bce, cf. Lk 1.5), rather than (correctly) after the deposition of Herod’s son Archelaus in 6 ce.Footnote ³

Interpreters have ventured some plausible speculations regarding how Luke could have arrived at this confused dating of the regional census.Footnote ⁴ The scope of the census is an altogether different matter, however. Most simply throw up their hands in bewilderment in the face of Luke’s apparently wild assertion that this was a global undertaking. Bart Ehrman’s comments are representative: ‘We have relatively good records for the reign of Caesar Augustus, and there is no mention anywhere in any of them of an empire-wide census for which everyone had to register by returning to their ancestral home.’Footnote ⁵

James Edwards makes some progress toward an explanation by connecting Luke 2.1 with Acts 11.28–30, which describes a similarly unsubstantiated ‘great famine in all the world (λιμὸν μεγάλην…ἐφ᾽ ὅλην τὴν οἰκουμένην)’. He too, however, only infers from this that Luke used the word to refer to any event that ‘covered much of the Roman empire’.Footnote ⁶ The problem with this interpretation is not only that there is no external evidence that either the census or the famine covered ‘much of the empire’. More importantly, this reading is directly contradicted by Acts 11 itself, which makes clear that the famine did not even affect Antioch (which was in a position to send an offering of support to the beleaguered Judean church, 11.29), much less ‘much of the empire’.

This paper argues that any attempt to make sense of Luke’s use of οἰκουμένην in either Luke 2.1 or Acts 11.28 is simply a forced answer to a false question. This is because, as Charles Torrey proposed in 1936, the Greek versions of these passages which have come down to us almost certainly include a fateful mistranslation of their Semitic sources (Hebrew in the case of Luke 2 and perhaps Aramaic in the case of Acts 11). Unfortunately, Torrey’s Our Translated Gospels and his other works on the Semitic origins of the Gospels are out of print,Footnote ⁷ and largely neglected today even by advocates for a Semitic approach to the Gospels.Footnote ⁸ My aim in the remainder of this article is to bring fresh attention to Torrey’s brilliant but neglected solution to the puzzle of the census’ extent, and to point out some additional evidence in support of his hypothesis.

2. A Semitic Approach: ‘All the World’ or ‘All the Land’?

Let us begin with Torrey’s evidence for the mistranslation of the phrase πᾶσαν τὴν οἰκουμένην in Luke 2.1. He summarises the problem along the lines I sketched above: in both Luke 2.1 and Acts 11.28, events are described as affecting ‘the whole inhabited world’ which we know from evidence both external (Josephus on the census) and internal (Acts 11.29 on the famine) to have affected only Palestine.Footnote ⁹ Torrey explains both passages at a stroke: ‘As for the confusion of Palestine with “the world”, no one who is acquainted with Hebrew-Jewish usage can fail to see the explanation of Lk.’s Grk [sic]. The same word which means “the earth” and “the world” means also “the land”.’Footnote ¹⁰

That is, πᾶσαν τὴν οἰκουμένην in Luke 2 could easily have arisen as a mistranslation of the Hebrew, כל־הארץ, or the cognate Aramaic expression, כלה ארעא, each of which could mean either ‘all the earth’ (Gen 1.26), or ‘all the land [of Israel]’ (as in Jos 2.24, 9.24; 1 Sam 13.3; 2 Sam 24.8; 2 Kgs 17.5; Isa 10.23; Jer 23.15; etc.).Footnote ¹¹ Indeed, when the phrase ‘all the land’ appears without further qualification in the Hebrew Bible (as opposed to, e.g. ‘all the land of Egypt,’ Gen 41.19), it uniformly describes the land of Israel. Unsurprisingly, in the Old Greek versions of these texts, הארץ is overwhelmingly rendered with γῆν. Nonetheless, in Isaiah 10.23, it was rendered with οἰκουμένην, thus establishing a narrow precedent for Luke’s mistake in 2.1.

The evidence for mistranslation in Luke 2.1 and Acts 11.28 is further strengthened by comparison with Luke 23.44. This verse describes how, at the crucifixion, σκότος ἐγένετο ἐφ᾽ ὅλην τὴν γῆν ἕως ὥρας ἐνάτης.Footnote ¹² Assuming Markan priority to Luke, Luke’s ἐφ᾽ ὅλην τὴν γῆν is drawn from Mark 15.33, which correctly uses the biblical idiom for Palestine. Luke’s troubles with the expression πᾶσαν τὴν οἰκουμένην seem to be of his own devising, whatever their precise origin.

If there are good reasons to think that Semitic Vorlage underlie at least part of Luke’s Infancy Narrative and Acts 11, then Torrey’s mistranslation hypothesis would solve the mystery of ‘global’ census: the earliest versions of these texts, which were subsequently mistranslated in Greek Luke, were in broad agreement with Josephus that Quirinius’ census was restricted to the land of Israel.Footnote ¹³ Luke or whoever translated his Semitic sources seems to have misconstrued the biblical and Targumic idiom for describing Israel in two of its occurrences in his sources, and the result has been pervasive confusion among his commentators for most of the last two millennia.

3. Identifying Translations from Hebrew or Aramaic: Two Test Cases

But there’s the rub: is it reasonable to think that at least Luke 2 and Acts 11 might be translations of Semitic originals? This question really divides into two: first, can we establish with any reasonable confidence, in the absence of clear manuscript evidence, that an ancient text is a translation into Greek from a Semitic language? And second, if that prior burden can be met, is there sufficient evidence to warrant the judgement that Luke-Acts is, at least in part, a translation into Greek of Hebrew or Aramaic sources?

Regarding the first question, the answer is unambiguously, ‘Yes’. At the very least, such judgements are common in studies of ancient texts outside the New Testament. We can illustrate this point by reference to contemporary views of two ancient texts which survive solely in Greek, but which were, by the near-universal consensus of their specialist scholars, composed in Syriac or Hebrew. First, consider the Cologne Mani Codex, which was recovered in Egypt and is now housed at the University of Cologne. Measuring only 4 x 4 centimetres, it is, according to its editor and translator, Albert Henrichs, ‘the smallest codex known from antiquity.’Footnote ¹⁴ This postage-stamp of a book contains a Greek life of Mani, the third-century founder of Manichaeism, and naturally is of great interest to students of that now-extinct religious community. Most significant for our purposes, however, is that there is wide agreement that the text is a translation from Syriac, the language in which most of Mani’s own works were composed.Footnote ¹⁵

Like Torrey with Luke, Henrichs points to apparent mistranslation as the key evidence that the Mani Codex is a translation from Greek, as the following passage underscores:

What is most striking about this passage is that Henrichs warrants his translation-hypothesis by appealing to a single apparent mistranslation, alongside an unspecified number of Semitisms. As we will see further below, applied to Luke-Acts, this standard would readily confirm each text as at least partly a translation from a Semitic source.

Like the Cologne Mani Codex, 1 Maccabees survives only in Greek, but the critical consensus strongly supports its having been written in Hebrew. Daniel Schwartz’s treatment of the issue in his recent (2022) commentary for the Anchor Bible series is representative:

He goes on to catalogue a number of salient examples, including the prevalence in 1 Maccabees of paratactic syntax, typical of Hebrew rather than Greek; the frequency of Semitisms (e.g. ‘by the mouth of the sword’, 1 Macc 5.28, 51); and mistranslation, as in ‘the case of “thirty” in 6.37, where the number evidently derives from a misreading of consonantal ŠLŠYM as šelōšîm (“thirty”) instead of šālīšîm (“officers”, as in Exod. 14.7)’.Footnote ¹⁸ This style of argument is common sense in Manichean or 1 Maccabean studies;Footnote ¹⁹ why should the same standard not apply to the study of the New Testament?

4. Semitic Vorlage in Special Luke

Let us turn to consider the evidence for Semitic Vorlage in the Gospel of Luke in particular. We need to be clear at the outset, however, about the scope of this problem: I do not intend to argue here for the Semitic origins of the ‘triple’ (i.e. Markan) tradition in Luke, or even for the ‘double’ tradition, consisting of pericopae shared by Matthew and Luke but absent from Mark.Footnote ²⁰ I am interested here only in establishing the likelihood of Luke’s having relied on Hebrew and Aramaic sources for at least parts of his Sondergut, ‘Special Luke,’ and at least parts of Acts.

On its face, there is nothing implausible about this hypothesis. Luke himself tells us that he knew of ‘many (πολλοí)’ prior attempts ‘to compose a narrative (ἀνατάξασθαι διήγησιν)’ about Christ’s life, and also that he had access to eyewitness sources (οἱ ἀπ᾽ ἀρχῆς αὐτόπται) of his own (Lk 1.1–2). Why should we assume that all of Luke’s sources were transmitted in Greek? After all, few today doubt that Jesus taught his disciples in Aramaic, not least because of the many Aramaic expressions placed on his lips throughout the Gospels (cf., e.g. Mk 5.31, 7.34, 14.36, 15.34; Matt 5.22, 27.46; Jn 20.16). Moreover, as James Edwards observes, in Palestine between the Maccabean and Roman revolts, ‘Hebrew…remained in use as a written language, and particularly in the writing of sacred texts.’Footnote ²¹ For instance, ‘about 75% of the Dead Sea Scrolls are written in Hebrew…These general proportions also apply to the documents of Wadi Muraba’at, Nahal Hever, the Bar Kokhba Letters, and Jewish ossuary inscriptions in Palestine.’Footnote ²²

Moreover, we also have external reports going back to Papias (fl. 115 ce) associating the apostle Matthew with Hebrew or Aramaic narratives about Jesus: ‘Matthew arranged the discourses of the Lord in the Hebrew language, and each one translated them as he was able.’Footnote ²³ (NB: We will not here consider the relationship of this alleged Matthean composition to the canonical Greek Gospel of Matthew.) This theory is repeated by Irenaeus,Footnote ²⁴ while Eusebius wrote that the Alexandrian Pantaenus (d. ca. 210) had reported seeing a copy of Matthew ‘in Hebrew characters’ on a trip to ‘India’,Footnote ²⁵ and Jerome claimed that a copy of ‘Hebrew’ Matthew was still present in his day (ca. 400 ce) in a library in Caesarea.Footnote ²⁶ As B. Ward Powers has observed, this link between Matthew and early written narratives about Christ makes good sense: Matthew was the only one of the Twelve who definitely had a ‘white collar job,’ as a tax collector, and so who would have been, not only literate, but presumably accustomed to writing for a living.Footnote ²⁷ If Matthew or others similarly equipped had in fact recorded reminiscences of Jesus in Hebrew or Aramaic, the real surprise would be if none of them had made their way into the canonical Gospels, including Luke.

Moreover, there is particularly strong evidence that ‘Special Luke’ is a translation from Hebrew, as James Edwards has argued with great force. He observes that there is ‘almost a 400% increase of Semitisms in Special Luke over Semitisms that Luke shares in common with Matthew and/or Mark.’Footnote ²⁸ Edwards’ discussion is much too rich to summarise here, but a few examples will serve to illustrate the point. For instance, he notes that ‘there are sixty-nine occurrences of καὶ ἐγένετο, ἐγένετο δέ, and ἐγένετο in the Third Gospel’, each of which ‘repeats in Greek the same function that initial ויהי (‘and it happened’ or ‘it came to pass’ does in Hebrew.’Footnote ²⁹ This Hebraism is distinctively Lukan: ‘Of its sixty-nine occurrences in Luke, only two are shared with Matthew and only five with Mark. Thirty-eight (=55%) of Luke’s uses of ἐγένετο occur in Special Luke.’Footnote ³⁰

So too, ‘ἐνώπιον occurs twenty-two times in the Gospel of Luke, but never in Matthew or Mark and only once in John (20.30). Ἐνώπιον occurs some 540 times in the LXX, most frequently as a translation of לפני. Sixteen (=73%) of the occurrence of ἐνώπιον in Luke occur in Special Luke.’Footnote ³¹ In a lengthy appendix, Edwards catalogues a total ‘of 703 Semitisms’ in Luke, of which ‘653…are unique to Luke’, appearing ‘either in Special Luke or as Lukan additions to material shared in common with Matthew and/or Mark’.Footnote ³² Moreover, he notes, ‘Semitisms occur in much higher frequency in Special Luke than in passages that Luke shares in common with Matthew and/or Mark.’Footnote ³³

Luke’s Semitisms are of course commonly described as ‘Septuagintisms,’ as deliberate imitations of ‘biblical Greek’. Edwards offers several strong objections to this explanation. He points out, first, that while Luke’s Semitisms are invariably ‘unqualified Hebraisms’, they are at best only ‘qualified Sepugantism[s], for [they are] instance[s] of Luke imitating only those parts of the LXX dependent on a Hebrew Vorlage’.Footnote ³⁴ And second, he notes that ‘several…Lukan Semitisms are either difficult or impossible to explain as Septuagintisms’.Footnote ³⁵ For instance, ‘In only one instance is בלבו translated as ἐν ἑαυτῷ in the LXX (Esth 6.6). In order to maintain the Septuagint hypothesis in this instance, one would have to argue that Luke used an expression five times on the basis of one occurrence in all the LXX. That contradicts the definition of a “Septuagintism”, which is the conscious imitation of a pattern or characteristic style in the LXX.’Footnote ³⁶

We need not canvass all of Edwards’ extensive refutation of the ‘Semitisms-as-Septuagintisms’ theory here, but it would be worthwhile to consider one objection to a Semitic approach which he does not address, namely that (as Chang-Wook Jung has pointed out) the correspondence of Luke’s biblical quotations in 1.15 and 2.23–4 to the LXX might suggest that Luke’s Infancy Narrative was composed in Greek.Footnote ³⁷ This strikes me as a bruised reed on which to rest the case against a Semitic origin for the Infancy Narrative, however. After all, the same logic would suggest that we should argue from the presence of the line, ‘Blessed is the womb which bare thee, and the paps which thou hast sucked’ in Pevear’s and Volokhonsky’s recent translation of The Brothers Karamazov to Dostoevsky’s dependence on the King James Version of Luke!Footnote ³⁸ The translator’s practice of rendering quotations from an authoritative text in the target language’s received version is too common for the presence of the Greek Old Testament in canonical Luke to tell us anything about a particular pericope’s language of composition.

Another possible alternative to (parts of) Luke being a translation from a Semitic original is that Luke’s first language was Aramaic (or, less plausibly, Hebrew), and that he could write in Greek only with a great deal of interference from his Semitic mother-tongue.Footnote ³⁹ In this case, the Third Gospel would still be at least partly a translation from Semitic sources, but one which took place, so to speak, in the mind of the author as he composed it in Aramaic or Hebrew, and then translated it into Greek on the page. This is not impossible, but we have strong grounds for rejecting it in Luke’s Prologue (Lk 1.1–3), which is famous as a piece of polished and idiomatic Greek.Footnote ⁴⁰ As Edwards wonders, ‘Why would an author craft a prologue in cultivated and literary Greek and then adhere to a particular style in the half of his Gospel not paralleled by Matthew and/or Mark, characterised by perceptible and sometimes awkward Semitisms?’Footnote ⁴¹ Torrey makes much the same point: ‘The style [of Luke 1–2], which is homogenous throughout, is as far removed from that of the preceding prologue [sc. Luke 1.1–3] as the east is from the west…It is not a matter of occasional or frequent Hebraisms, the style is one continuous Hebraism.’Footnote ⁴² The (partial) translation-hypothesis fits these data better than does bilingual interference: if Luke were incapable of writing anything but pidgin Greek, heavily accented by his Semitic mother-tongue, then the Prologue should have been well beyond his reach, whereas even a brilliant Greek stylist might well be prone to errors of grammar or style when attempting to produce a literal translation of a Semitic original.

As the quotation just above indicates, Torrey himself, though he generally favoured an Aramaic Approach to the Gospels and Acts, argued that Luke’s Infancy Narrative in particular – including our crucial passage regarding the census – was a translation from Hebrew. He notes another apparent mistranslation in Luke 1.39 in particular, as strongly indicating a Hebrew Vorlage for the Infancy Narrative: Luke describes how Mary went εἰς πόλιν Ἰούδα, which ‘cannot mean “to a city of Judah (or Judea)”’, since that would have been εἰς πόλιν της Ἰουδαίας, as in verse 26 of this chapter. The only permissible reading, Torrey notes, ‘is “to the city [named] Judah”, but there was no city of that name’.Footnote ⁴³

Here again is a mystery best resolved, Torrey suggests, by ‘go[ing] back of the Greek to the underlying Semitic…What the author of the narrative wrote was אל מדינת יהודה, and the translation should have been εἰς την χώραν της Ἰουδαίας, “to the province of Judea”.’Footnote ⁴⁴ A confusion which is incomprehensible in Greek is readily intelligible in Hebrew: ‘In the first century ad the use of מדינת in the signification “province” was practically obsolete, having been supplanted by the meaning “city.”…The translator of Symmachus even “corrects” the χώρα of the older Greek versions of the Old Testament to πόλις in 1 Kings xx. 14, Dan. viii. 2’, the latter of which offers ‘a perfect parallel to the one in Luke, for the Hebrew reads בהמדינת אלם, which Symmachus renders “in the city (!) Elam”.’Footnote ⁴⁵ Torrey takes this apparent mistranslation to betray a specifically Hebrew Vorlage for at least Luke 1–2: ‘I believe that the original document was Hebrew, not Aramaic: for the Aramaic יהוד למדינת would hardly have been rendered by εἰς πόλιν Ἰούδα. The word יהוד could not well have been misunderstood; moreover, it does not look like the name of a town, nor would it have been transliterated by Ἰούδα.’Footnote ⁴⁶

These additional instances of apparent mistranslations from Hebrew into Greek in Special Luke provide independent and converging lines of evidence in support of Torrey’s proposal that the ‘global’ census in Luke 2.1 is in fact the result of a mistranslation of the original text’s כל־הארץ.

5. The Witness of the Lewis Palimpsest

Moreover, though I do not believe Torrey ever drew attention to it, we find substantial support for his mistranslation thesis for the interpretation of Luke 2.1 in the Lewis Palimpsest, the Syriac (i.e. E. Aramaic) version of the Gospels discovered by Agnes Lewis at St. Catherine’s Monastery in the Sinai Peninsula in 1892.Footnote ⁴⁷ This important manuscript was likely copied in the fourth or fifth century,Footnote ⁴⁸ but the translation itself doubtless originated earlier still. (We will return to how much earlier in a moment.)

In the Peshitta (‘Simple’) version of Luke translated in the fifth century to standardise the text of the New Testament across the Syriac-speaking East,Footnote ⁴⁹ the command in Luke 2.1 is for the enrolment of ܕ݁ܐܘܚܕܢܗ ܥܡܐ ܟ݁ܠܗ, ‘every people of his dominion’ – this represents an idiomatic rendering of the Greek οἰκουμένην, as indicating an Empire-wide undertaking. In the Lewis Palimpsest, by contrast, the command is for a census of ܟܘܠܗ ܐܪܥܐ, ‘all the land’ – precisely what Torrey suggests was the original reading of Luke 2.1, which was garbled in our Greek versions.

This reading of Luke 2.1 is highly significant, because the Lewis Palimpsest is among our earliest witnesses to the Gospels, and in some pericopes seems to reflect the most primitive reading of any surviving version,Footnote ⁵⁰ as in John 4.24, which omits the expression, ‘God is Spirit’,Footnote ⁵¹ or Matthew 1.16, which reads ‘Joseph…begot (ܐܘܠܕ) Jesus’ (Matt 1.16), and also omits the observation that Joseph ‘did not know (οὐκ ἐγίνωσκεν)’ Mary until Jesus was born (Mt 1.25). This does not necessarily reflect a denial of the Virgin Birth, since Lewis’ Matthew still conspicuously identifies Mary as ‘the Virgin (ܒܬܘܠܬܐ)’ (Matt 1.16), and narrates Jesus’ conception as having occurred ‘by the Holy Spirit (ܡܢ ܪܘܚܐ ܕܩܘܕܫܐ)’, before the marriage of Joseph and Mary (Matt 1.18). Given that, these readings seem hard to explain as, e.g. ‘Ebionite’ efforts to eliminate the Virgin Birth from Matthew. On the contrary, the Lewis Palimpsest seems to be more primitive in this instance than other versions of Matthew – if it is hard to see why later generations would have deliberately introduced the Lewis Palimpsest’s wording, it is easy enough to see why an anxious scribe might have ‘improved’ it, and so introduced an ‘orthodox corruption’ of the text.Footnote ⁵²

The Lewis Palimpsest is not only quite primitive, but also seems to reflect a distinctively Palestinian milieu. In ‘The Palestinian Origin of the Old Syriac Gospels’, Torrey canvasses a number of usages in the Lewis Palimpsest which are distinctive of W. (Palestinian) Aramaic rather than E. Aramaic.Footnote ⁵³ And Méténier observes, ‘Nous avons recensé dans la Vetus Syra vingt-six mots, groupes nominaux ou expressions ne relevant pas du syriaque Classique qui eut son épicentre à Édesse, mais qui l’on retrouve dans les écrits judéo-araméens.’Footnote ⁵⁴

Torrey notes an apparent textual link with the Targum Jonathan in the Lewis version of John 10.22,Footnote ⁵⁵ which is especially striking: ‘Then it was the [Feast of] the Renewal (τὰ ἐγκαίνια, i.e. Hanukkah) in Jerusalem.’ Where the Peshitta translates ἐγκαίνια straightforwardly (ܚܘܕ݂ܬ݂ܐ, renewal), the Lewis Palimpsest has the surprising, ܐܝܩܪ ܒܝܬ ܡܩܕܫܐ, ‘the glory of the holy house’. Torrey notes that with this usage, ‘we are carried back to the scene of the original dedication [of the Temple] described in 2 Chron 7.1–3, and more particularly to the Targum to these verses, where the words “איקר” and “בית מקדשא” occur and recur. Whoever will read this Targum will recognise at once the source of the Lewis text’.Footnote ⁵⁶ Torrey argues from this and other evidence for a second-century origin of the Syriac version of the Gospels preserved in the Lewis Palimpsest, perhaps by Aramaic-speaking refugees from the recurring wars in Roman Palestine.Footnote ⁵⁷

This translation of the Gospels into E. Aramaic is arguably early, primitive and distinctively Palestinian, and so it is particularly interesting that it includes a cognate Syriac expression to the Hebrew one which Torrey proposed lay behind the confused οἰκουμένην in Luke 2.1. How did this reading come to be in the Lewis Palimpsest? Was it sheer coincidence, a good guess at the biblical idiom misrendered in the Greek or (most interesting of all) an acquaintance with an untranslated, Semitic version of Luke’s Infancy Narrative?Footnote ⁵⁸ Whatever the case, the Lewis Palimpsest, here as elsewhere, sheds important light on the earliest form of Luke.

6 Aramaic Vorlage in Acts 1–15

Finally, and more briefly, let us consider the likelihood of a second Lukan mistranslation of ארץ/ארעא as οἰκουμένην, namely in Acts 11.28. Our earliest Aramaic version is the Syriac Peshitta, but its evidence is significant as well, since, like the Lewis Palimpsest of Luke 2.1, it describes the famine as taking place ܒܟ݂ܠܗ ܐܪܥܐ, ‘in all the land/earth,’ which likewise is precisely the expression that Torrey suggested is mistranslated in the surviving Greek versions of Acts. Interestingly, in his eighth-century Syriac commentary on Acts, Isho’dad of Merv glosses Acts 11.28 by noting that ‘the believers in Antioch…saw that there was famine in Judea’.Footnote ⁵⁹ This seems to indicate that he (correctly) interpreted the Peshitta’s ܒ݁ܟ݂ܽܠܳܗ ܐܰܪܥܳܐ not as a global famine (as in the Greek’s ἐφ᾽ ὅλην τὴν οἰκουμένην), but rather as ‘in all the land,’ viz., the land of Judea. This interpretation is simply unavailable for the reader of the received Greek text of Acts, but, as Isho’dad indicates, it was a natural and coherent reading of the E. Aramaic (Syriac) Peshitta, and perhaps also of the Aramaic Vorlage mistranslated in Greek Acts.

But again, we need to assess the likelihood that at least parts of Acts are translations from Aramaic. Here too, Torrey offers a useful starting point: in his The Composition and Date of Acts (1916), he argued at length that Acts 1–15 was a translation from an Aramaic original. Our thesis does not require defending so strong a position as that; indeed, the evidence already canvassed for the likelihood that at least some of canonical Luke consists in translations from Hebrew might suffice on its own to warrant the possibility that at least some of Acts or its sources was originally composed in the same way. Nonetheless, Torrey surveys a number of other indicators, additional to Acts 11.28, for Aramaic Vorlage throughout the first half of Acts.

The best test, here as in the Gospels, is evidence of mistranslation, and Acts 2.47 provides an interesting early instance. Torrey notes that the conclusion of the verse (the Greek phrase ἐπὶ τὸ αὐτό, ‘together’) is so odd that most translations simply punt on it: e.g. in the RSV’s ‘And the Lord added to their number day by day those who were being saved’, ‘to their number’ is at best a generous gloss on the awkward, ἐπὶ τὸ αὐτό. But mistranslation of the Aramaic לחדא would neatly explain the Greek: in E. Aramaic, as Torrey notes, this expression consistently means ‘together,’ e.g. ἐπὶ τὸ αὐτό, ‘but in the Judean dialects of Aramaic, the usual meaning of לחדא is “greatly, exceedingly”, and this is precisely what is needed in the place of ἐπὶ τὸ αὐτό in Acts ii, 47. For example, in the Onkelos Targum, “לחדא” (properly “singularly, uniquely”) is the ordinary rendering of Heb מאד. Thus, Exod 19, 18, “And the whole mountain trembled greatly (לחדא)”.’Footnote ⁶⁰

Torrey identifies another striking instance of apparent mistranslation from Aramaic in Luke’s curious phrase ‘unto ordinances of angels (εἰς διαταγὰς ἀγγέλων)’ (Acts 7.53). As David B. Hart correctly observes in a note on this verse in his translation of the New Testament, ‘The phrase is strange.’Footnote ⁶¹ Here again, most English translations, including Hart’s, opt to quietly correct Luke’s grammar, rendering the phrase as ‘through the ordinances of angels’. And this, as Torrey notes, and as, e.g. Gal 3.19 suggests, was almost certainly the sense of the passage in its Aramaic original; Luke’s εἰς is likely misrendering the preposition ל, which can mean ‘to’ (i.e. εἰς), but can also be equivalent to the Greek instrumental dative (‘by means of’). The latter sense was surely intended; ‘Luke’s rendering here is not merely too literal, it is incorrect.’Footnote ⁶²

Another possible approach to identifying Semitic Vorlage in the Gospels is by way of textual variation in the manuscript tradition, in divergences which are best explained as independent translations of an ambiguous or polyvalent Aramaic (or Hebrew) word or phrase. In his An Aramaic Approach to the Gospels and Acts, Matthew Black assembles a wealth of such textual variants; of particular interest is his discussion of those apparently caused by the slippery Aramaic particle די, which can mean ‘who, which, when, because, so that,’ and more. Black cites one in particular from Acts 7: where most manuscripts read, ‘whom they did not want (ᾧ οὐκ ἠθέλησαν) to obey’ (Acts 7.39), Codex Bezae has, ‘because they did not want (ὅτι οὐκ ἠθέλησαν).’Footnote ⁶³ This variation is not easily explained in terms of a misreading of either ᾧ or ὅτι, but is readily explicable in terms of independent translations of די.

Again, there is neither space nor need to fully reproduce here either Torrey’s or Black’s survey of the evidence for Aramaic Vorlage in Acts, and nothing in our thesis depends on all or even most of Acts 1–15 having been originally composed in a Semitic tongue. But the foregoing three examples, set alongside the evidence for mistranslation in Acts 11.28, suffice to show that this is a compelling explanation for the idiosyncratic use of the Greek οἰκουμένην in Luke-Acts. More than that: so far as I know, it is the only plausible explanation for this usage ever offered.