Miscellaneous Extreme Literal Mistranslation : The Gǝʿǝz Text of Titus 3 : 1

The Gǝʿǝz translation of Titus 3:1, by following the Greek original literally—word-for-word and even case-for-case—yields a good, clear Gǝʿǝz text which the average Gǝʿǝz reader will take as having a meaning which is opposite to that of the Greek: not (1) ‘Remind them to be obedient to rulers’, as in the Greek (the normal Pauline message), but (2) ‘Remind the rulers to be obedient’. This paper reconstructs how this semantic inversion came about. The Gǝʿǝz word-sequence is syntactically ambiguous, allowing two different parses: one normal (2) and the other highly marked (1). The intended, marked parse (1), which would yield the Greek sense, is unlikely even to occur to the Gǝʿǝz reader. Indeed, the andǝmta to this passage presupposes the unmarked parse (2).


zäkkǝr-[omu lä-qäddämt wä-lä-mäk w anǝnt] kämä yǝtʾazzäzu bäk w ǝllu mǝgbarä śännay wä-yǝkunu ṭǝbuʿanä botu
'Remind [them, (i.e.) the nobles and rulers,] that they should be obedient in every good deed, and let them be firm in this.' This is the way every Ethiopian I have consulted immediately understands the passage, and it is clearly the sense in which the author of the andǝmta (below) understood it. For the moment I will assume that this is the correct understanding-an assumption, however, which is not absolutely compelling, as the passage can in principle also be parsed quite differently. I will examine this issue critically below.
There are two serious problems with the above translation. First, it is exactly the opposite of Paul's injunction elsewhere in the New Testament, where he repeatedly and clearly says that you (i.e. good Christians) should be obedient to the secular rulers, for instance Romans 13:1 (see below), or 1 Peter 2:13. Here he is instead telling the rulers that they themselves should be obedient. Obedient to whom? The text does not say. An obvious interpretation is: obedient to God. But this is just an interpretation, with no support in the text itself. We might hope for help from the andǝmta commentary to this passage, which says (in Amharic), ቀሳውስት ዲያቆናትን ምከራቸው አስተምራቸው በበጎ ምግባር ኹሉ ጸንተው ይታዘዙ ዘንድ ዘክሮሙ በዚያው ጸንተው ይኖሩ ዘንድ ዘክሮሙ qäsawǝst diyaqonat-ǝn mǝkär-aččäw astämr-aččäw bä-bäggo mǝgbar ḵullu ṣäntäw yǝttazzäzu zänd zäkkǝromu bä-zziyaw ṣäntäw yǝnoru zänd zäkkǝromu 'Exhort and teach the priests and deacons. 4 Remind them that they should be obedient and stand firm in all good deeds; remind them that they should be firm in this.' 4 Some readers have suggested the alternative translation 'O priests, exhort and teach the deacons'. Arguing against this is the fact that the imperative verbs are both in the singular (though with a plural object suffix) while 'priests' is plural, as well as the fact that both the Greek and the Gǝʿǝz texts involve a coordinated noun phrase (X and Y). Leslau presents asyndetic (zero) coordination as a normal possibility in Amharic (Leslau 1995, 725), a general statement to which informants react with skepticism. Significantly, Leslau's examples involve pairs of nouns which are natural partners (salt and pepper, by foot and by horse), which is also true of the andǝmta example (priests and deacons); this No hint here of 'obedience to God'. Moreover, the temporal 'nobles and rulers' of the Gǝʿǝz text-powerful figures of high political rank-have now been recast as churchmen, indeed merely middle-level churchmen (priests, deacons). The andǝmta writer seems to have been puzzled at the idea that the top political figures should be 'obedient', and accordingly changed the wording and the reference so as to 'rescue the text' and create an interpretation which can make sense. After all, priests and deacons do have superiors, to whom they should be obedient.
The second, and much more serious, problem with this passage is that it clashes flagrantly with the Greek. The Greek text reads as follows: The message of this Greek verse (which is of course the original) is the opposite of its Gǝʿǝz counterpart. Unlike the Gǝʿǝz, it 'says what it should say', in harmony with Paul's message elsewhere. It does not explicitly mention who should be obedient, but clearly the intention cannot be the 'rulers and authorities'. Presumably the reference is to good Christian people.
The immediate question, and the central concern of this paper, is how this mistranslation could happen at all. It is utterly improbable that it was intended as a deliberate (and huge) theological change, violating the Greek text and reversing Paul's message elsewhere. Rather, the explanation is technical and syntactic in nature. It emerges transparently if we do a word-for-word comparison of the beginning of the verse in Greek and in Gǝʿǝz. For the sake of clarity, the schematic comparison will be presented here via the English glosses. natural pairing could also explain why, in the andǝmta, only the second of the two conjoined nouns (deacons) takes the accusative marker -(ǝ)n, as if the two nouns together formed a single sense unit, a sort of quasi-compound. The string of words is well-nigh identical in Greek and Gǝʿǝz. 5 Even the case marking is the same, insofar as the two languages allow it: the Greek datives correspond to Gǝʿǝz prepositional phrases lä-Noun, 'to/for Noun'. 6 The explicitly accusative 'them' of the Greek corresponds in Gǝʿǝz to a pronominal suffix, which does not unambiguously indicate case.
What does differ, in a major way, is the syntactic analysis of this near-identical string of words in the two languages. In the Greek, the rulers and authorities (in the dative) belong to the embedded clause: they are the object of the infinitive 'to be obedient (to)' (their linear position preceding the infinitive is a normal possibility in Greek syntax). In the Gǝʿǝz, as understood above, the rulers and authorities belong to the main clause; they are an example of the very common 'analytical object' construction, whereby the suffix -omu, 'them', which is the object of the verb 'remind', is restated appositionally as a noun with the preposition lä-,'to': 'Remind them, namely (to) the rulers and authorities'. 7 I want to stress the absolute normality of the analytical object construction in Gǝʿǝz. It is found throughout the entire recorded history of the language, including inscriptional Gǝʿǝz. Bulakh's study of word order in epigraphic Gǝʿǝz gives, 8 in passing, (at least) two 'undamaged' examples of the analytical object syntagm that do not involve any reconstruction of broken text: 9 ʾiyǝṭǝʿǝmā la-mot 'he will not taste death'; 10 smʿk k-ʾgzbḥr knnh ldḥr = samaʿku k-ʾǝgziʾabǝḥer k w annanahu la-dḥr 'I heard that God judged DḤR '. 11 There are also examples from ʿEzana's inscription RIÉ 187, though with some of the text broken: 12 RIÉ 187, line 9: ʿaraznahomu la-ʾAg w ezāt 'we plundered the ʾAg w ezāt'

annannāhomu la-[sar]//[we] maḥazā wa-la-nagaśta sarāwit
'we ordered the troop of Maḥazā and the leaders of the troops …' Of course, the prepositional phrase with lä-enjoys considerable positional flexibility in the analytical object construction, and need not be adjacent to, or even following, the suffix. But impressionistically, based on my experience, by far the commonest construction does indeed place the lä-phrase directly after its coreferential suffix. This is what we have, or seem to have, in Titus 3:1. By literally following the words and even the cases of the Greek original, the Gǝʿǝz translation has apparently inverted the sense of the sentence. And it inverts it in a particularly egregious way, producing a statement which runs counter to a 10 Bulakh 2012, 145; RIÉ 232, lines 9-11. Transcription of this and the following four examples follows the conventions of Gǝʿǝz, not Amharic. 11 Bulakh 2012, 146;RIÉ 192, lines 17-18. A characteristic of this inscription is that the third-person suffixes typically appear with h, thus not third person masculine singular -o (as in classical manuscript Gǝʿǝz) but h(u); see Bulakh 2012, 138, n. 3. This is what enables us to see the suffix in this unvocalized inscription. 12 Pace Schneider: 'Les inscriptions d'Axum ont été examinées, mais […] on n'y relève pas d'exemple de la construction analytique avec suffixe et préposition' (Schneider 1959, x; again on p. 72).
basic tenet of Pauline theology (obedience to secular authority), a tenet which is rendered correctly elsewhere in the Gǝʿǝz New Testament. Literalism was a favoured technique of translation in antiquity, 13 as realized most notoriously in Aquila's translation of the Old Testament into Greek. But usually, when a literal translation misfires, it is because it has the effect of producing incoherence or grammatical bizarreness or stylistic crudity, not because it produces a clear and linguistically correct translation that effectively inverts the intended sense of the original, which is what we have here. It is at this point that we must return to the possibility, mentioned near the beginning of the paper, of understanding the Gǝʿǝz text in a different way. 14 The text itself reads, ዘክሮሙ ለቀደምት ወለመኳንንት ከመ ይትአዘዙ zäkkǝr-omu lä-qäddämt wä-lä-mäk w anǝnt kämä yǝtʾazzäzu.

zäkkǝr-omu [lä-qäddämt wä-lä-mäk w anǝnt kämä yǝtʾazzäzu].
That is, rather than 1) remind [them = to-rulers and to-authorities] that they.should.obey 'Remind (to) the rulers and authorities that they should obey', one might parse it as 2) remind them [to-rulers and to-authorities that they.should obey] 'Remind them that they should obey (to) the rulers and authorities'. I will call the bracketed part of pattern (2) the 'Arg kämä V' pattern (Arg = Arguments).
There seems to be no grammatical barrier to such an interpretation. Although Gǝʿǝz usually exhibits VO word order, the flexible word order of the language also allows the verb to be clause-final, so that the object can precede its verb; and the clause-medial position of the subordinating conjunction kämä, separating the clause-final verb from its arguments, though not common, is certainly attested (see below). The parsing ambiguity arises because, in such a sequence of words, there is nothing that explicitly marks the boundary between main and embedded clause-the Nahtstelle (lit. 'location of the seam'), to use Sima's apt term. 15 It should be noted that in Amharic, in contrast to Gǝʿǝz, this 'Arg kämä V' pattern is the normal construction (with ǝnd/ǝndä instead of kämä), and Amharic substratal influence will readily explain its presence in relatively late Gǝʿǝz. Thus Sima, examining a late Gǝʿǝz text (the sälamat verses of the Synaxarion, ultimately of unknown age but attested only after 1581), 16 shows how the grammar of this Gǝʿǝz text is truly describable as 'Amharic syntax in Gǝʿǝz'-including the 'Arg kämä V' construction at issue here. 17 The presence of this construction in the Gǝʿǝz sälamat is certainly due to an all-pervasive Amharic substratal influence on the text; the author was exploiting and privileging an already-existent (albeit uncommon) structural possibility of Gǝʿǝz.
But the existence of this marked construction in Gǝʿǝz is not just a recent phenomenon. Weninger does not address the phenomenon as such, 18 nor have I seen any discussion of it in the literature; but among dozens of examples of embedded kämä clauses in Weninger's book, a huge majority of which have clause-initial kämä, I found three that involve the 'Arg kämä V' construction, one from the Fisalgos and the other two from the Bible. 19 Particularly telling is this example from the very old Abba Gärima I text of The fact that the Gǝʿǝz independent pronoun antä, 'you' (absent in the Greek) is a nominative form shows unmistakably that it belongs to the subordinate clause 'that you should come' and has not been 'raised' to become 15 Sima 2010, 96. 16 Sima 2010, 91. 17 See Sima 2010, 97, exx. 33, 38. 18 Weninger 2001 These are Weninger's exx. 688, 851, 862 (respectively Weninger 2001, 175, 198, 200); ex. 688 is from the Fisalgos, the other two from the Bible. 20 Weninger 2001, 198, ex. 851. an object in the matrix clause; thus there is no doubt that kämä here is truly clause-medial. 21 The 'Arg kämä V' construction is thus a possible parse for Titus 3:1. But if so, it is a highly unusual piece of Gǝʿǝz. There are no examples of the 'Arg kämä V' construction anywhere else in Timothy or Titus, as a close examination of all the manuscript variants in Nebeyou's critical apparatus shows. Thus this was evidently not a routine syntactic tool that the translator used freely wherever he might wish. Then why use it precisely here? The only plausible (and obvious) answer is, in order to follow the Greek word-for-word.
If parse (2) truly was the translator's intention, we cannot say that he was literally guilty of a linguistic 'mistake'. He understood the Greek, and meant to say what it said. But if not 'wrong', the translation is certainly inept, because it creates major ambiguity. Both the main verb zäkkärä, 'remind', and the embedded verb taʾäzzäzä, 'obey', take a prepositional object with lä-. In this case, the use of the 'Arg kämä V' syntagm would have the unintended effect of positioning the object of the embedded verb ('to-rulers and to-authorities') directly adjacent to the -omu suffix of the main verb, thereby setting up the configuration of the analytical object construction, that is, the alternative reading (1). The translator's 'mistake' lay in shutting his eyes, wittingly or unwittingly, to this possibility-which would be the normal, unmarked parse for these words in Gǝʿǝz. To employ a Gǝʿǝz sentence that lends itself to two interpretations, one (1) obvious and normal and the other (2) unusual and marked, and to expect the reader unproblematically to understand the sentence precisely in the unobvious, grammatically marked sense (2), seems to me so infelicitous and misleading a translation technique as to merit being called a 'mistake' in its own right. How could the translator expect the reader not to take the Gǝʿǝz passage in its normal sense (1)? How could the reader possibly know that he was intended to understand it in the marked sense (2)? And can we really believe that the translator never once reflected upon his own text so as to become aware of the alternative parse (1), with its radically different meaning?
Conceivably the translator might have been aware of the ambiguity, but deliberately chose to ignore it. This seems to me unlikely. What is being translated here is the holy Bible, the Word of God, which the translator undoubtedly revered and whose message he would not wish to distort. The ambiguity is a distortion because it embodies two opposite possibilities; and the grammatically privileged possibility (interpretation (1)), the interpretation that almost any reader would automatically take, is the wrong one.
Actually, it almost does not matter whether the Gǝʿǝz text technically counts as a 'mistake' or not. Either way, by virtue of following the Greek exactly, the translator produced a text which was almost sure to convey the wrong message to readers.
We are left with the question of why the translator proceeded as he did. Here it is easy to raise questions and offer speculative suggestions, much harder to say anything definitive. Very plausibly the translator's command of Greek grammar was weak. Maybe he somehow really understood the Greek to mean, 'Let the rulers obey'? If so, his way of making sense of the Greek must have been to blindly translate each successive Greek word into Gǝʿǝz and then put the Gǝʿǝz pieces together following Gǝʿǝz grammar. Or could it be that, in the translator's conception of what it meant to produce a 'good' translation, word-for-word faithfulness to the Greek took priority over conveying the actual message? 22 Yet it should be noted that the translator of the Pastoral Epistles does not always follow the Greek slavishly. The Gǝʿǝz text shows many passages where the word order has been changed. Indeed, a change of word order in Titus 3:1 (notably, putting the conjunction kämä at the beginning of the embedded clause) would have easily enabled the translator to convey the message of the Greek unambiguously and with no confusion. On the other hand, if we assume that the translator in fact intended to follow 'Arg kämä V' syntax, could he truly have been totally unaware that his translation could be parsed and understood differently? Or perhaps he was aware of the ambiguity but it simply did not bother him? Finally, we should not forget that the text as we have it undoubtedly reflects centuries of copying, rethinking and rewording, a process which we cannot recover and during which the text could have undergone almost any kind of transformation.
The present case is not just an isolated fluke. There are other examples where such syntactic literalism, combined with lack of awareness of grammatical differences, leads to mistranslation, though seldom with such devastating consequences. A good example is Romans 13:1, already referred to above: centuries. Hence it is somehow strange that this mistranslation of Titus 3:1 has been taken up uncritically and accepted in the biblical canon-witness the andǝmta writer, who clearly takes this to be the 'true' biblical text which he is seeking somehow to explain.
I do not know how common such radical syntax-based misunderstandings of the text are in the Gǝʿǝz Bible. 25 The few comments I have found about the technique of Greek-to-Gǝʿǝz translation point in different directions. In his edition of the Gǝʿǝz Book of Acts, Niccum notes that 'a poor command of Greek best explains the idiosyncrasies of the Ethiopic Acts', observing in particular that 'frequently the [Greek] case ending of a noun is ignored resulting in quite unusual (and sometimes impossible) renderings'. 26 On the other hand, with regard to the Versio Antiqua (A-text) of the Synoptic Gospels, Zuurmond says that 'mistranslations resulting from evident lack of knowledge of Greek syntax or vocabulary are rare'. 27 In his study of the Gǝʿǝz Book of Esther, Miles says, Our heuristic expectation throughout is that in a mechanical translation, the translator will always choose that construction in the translation language which is formally closest to the construction he faces in the original. When he does not do this, there is always a reason: something would happen in the translation if he did. 28 But the opposite is true in the passage from Titus: 'something happens in the translation' precisely because he does choose the construction which (apparently) is formally closest to the Greek.
I am not a New Testament scholar, and hence I am in no position to assess the phenomenon of extreme literal mistranslation in the New Testament as a whole. But surely a number of potential and actual theological misunderstandings over the Bible would evaporate if Ethiopian theologians paid more attention to philology and linguistics. 25 Focused studies of translation syntax in Gǝʿǝz seem to have concentrated more on translations from Arabic into Gǝʿǝz, rather than from Greek into Gǝʿǝz; see e.g. Kropp 1986, Weninger 2014. 26 Niccum 2014 see also examples in his Appendix A. 27 Zuurmond 1989, 50. 28 Miles 1985, 5;underline in original.