The Father William Most Collection
Did St. Luke Imitate the Septuagint?
[Journal for the Study of the New Testament 15 (1982) 30-41]
Which expressions in the NT are Semitisms? How extensive are they? Are they due to translation or other causes? These and similar questions have occasioned much ingenious research, but, unfortunately, generally accepted results are few.
To give a few examples: A. Deissmann noticed that NT Greek is not the same as fifth century Attic1: thus the need of looking for Semitic influence was reduced. Still, the search for evidence of translation from Aramaic continued, whether for large stretches of the Gospels and Acts 2 or for just the words of Jesus3.
A search for errors in translation from Aramaic produced much work, especially that of C.C. Torrey. This approach rested on the assumption that those responsible for the Greek text of the Gospels, especially Luke, mistranslated, since they were unable to learn Aramaic well, even though they lived with Aramaic speakers4. Yet it was thought that modern scholars, with far fewer sources, could learn Aramaic better than those who lived with natives. Further, many of these alleged instances of mistranslation "... are open to grave objection"5. Yet that did not stop Black from asserting his own claims that he had found mistranslations.
Others sought, and still seek, explanations not in translation, but in the assumption that there was a Jewish brand of Greek6.
Some of the most convincing work on Semitisms has been done by H.F.D. Sparks in his studies on Luke7. Sparks was right to focus on Luke, since it is admitted that Luke, though a gentile, shows far more Semitisms than do the native Semitic
writers of the other Gospels. In his 1943 study, "The Semitisms of St. Luke's Gospel", Sparks mentions three possible sources, direct translation from Semitic, use of Semitized Greek translations from Semitic, or conscious imitation of the Septuagint to give what we might call a "Biblical flavor"8.
In 1943, Sparks quickly disposed of the theory of direct translation from Semitic, saying that such a theory "is ruled out of court" 9 by the Two Source Theory, which he then considered solidly proved. However, later on, in 1951, he spoke differently: "For myself, I am not wedded to orthodox Synoptic criticism"10. Sparks showed foresight and prudent judgement; at present the number of attacks on the Two Source Theory is multiplying11.
Still another shift appears in his 1951 article. Formerly, Sparks was quite firm in claiming that Luke's Semitisms came from deliberate imitation of the LXX12. As part of the support for this theory of LXX imitation, he added that in Luke, "only two characteristically Aramaic expressions are at all common", whereas, "several {Semitisms] ... can be traced without question to Biblical Hebrew"13. This, he thought, pointed to imitation of the LXX, since it seemed unlikely that Luke made such a thorough study of Hebrew as to assimilate Hebrew expressions directly.
But in 1951 he could write: "We should ... be very chary of accepting only one" solution14.
Bruce Chilton, in a recent meticulous linguistically based redaction analysis of Lukan passages, has prudently rejected many proposals by others as insufficiently grounded conjectures, and has supported the sound work of Sparks, especially by showing that Luke followed tradition with great care at several points in the passages studied15.
We hope to build on this work of Sparks and Chilton by a study of one strictly Hebrew feature of Luke that occurs more frequently than other Semitisms: the apodotic kai.
Other Semitic features in Luke are not frequent enough to make conclusions firm. For example, Sparks noted that Greek enopion could reflect Hebrew lifne16. Yet he saw that it need not reflect it in every instance, since such uses of enopion are found at times even in secular Greek17. Again, en to plus the infinitive does correspond well to the Hebrew be with the infinitive18. But we cannot be sure of a Hebrew substrate, for that structure is known even in Classical Greek. Admittedly Classical Greek does not use it in a purely temporal sense, but it does use prepositions with the articular infinitive so often and in such a wide variety that it would not take much to induce a native Greek speaker like Luke to make the slight extension into the purely temporal use of en to19. It is only the frequency of the construction in Luke that does suggest Hebrew influence: Matthew and Mark use it only once John never employs it; but Luke has it 25 times20.
We turn now to the apodotic kai, a true hard-core Hebraism and a relatively frequent one21. This is the use of kai to connect the main clause to a preceding subordinate clause. Such a structure is almost unknown in Classical Greek, though it does happen to occur once in Homer, Iliad 1.47822. It does seem to occur before questions but then has a different sense23. It is also found in slovenly Greek prose24.
However the Hebrew equivalent, apodotic wau25, was very common in Canaanite26, and had a wide range of uses in Classical Hebrew: after conditional, temporal, relative conditional clauses, and after participles standing for clauses etc. It became rare in new Hebrew, and was rare in Aramaic. In both it is largely confined to use after conditional or temporal clauses27.
Before proceeding with our study of apodotic kai, we should notice that sentences that have it very often start with still another Hebraism: kai egeneto. This reflects Hebrew wa yehi, which is quite normal in classical Hebrew. However, in later Hebrew it retreats under Aramaic influence28, and disappears completely in the new Hebrew of the Mishna. The structure is not native to Aramaic, though an equivalent does appear at times in the Targums when they are translating the Hebrew expression29. Horton adds that wa yehi, when followed by apodotic wau seems to be absent from the colloquial Hebrew of the time of Christ30. He adds that the Greek parallel, kai egeneto with a following apodotic kai is not found in the Greek papyri from Egypt, though it could be found in religious works31. He assumes its appearance in these works comes from imitation of the LXX.
We turn now to Luke's use of apodotic kai. It is commonly ascribed to conscious imitation of the LXX32, or to translation from a document that imitated the LXX, or to the influence of synagogue Greek which would be affected by the LXX33. Yet there seems to have been but little careful investigation. Torrey, in Our Translated Gospels, devotes a brief chapter to it, but makes no careful comparison to the LXX usage, nor does he give statistics34. He is more interested in a comparison of a few examples from the Old Syriac or Sinaitic version of the Gospels. When he is working from the Greek of Luke, he gives only eight examples, and no statistics. M. Zerwick is more thorough, and distinguishes three patterns after egeneto: (1) without a kai, (2) with an infinitive, and (3) with a kai plus a finite verb (the apodotic kai)35. Oddly, he says the third one, the apodotic kai, is a "construction of the LXX-which often omits that kai which is so necessary in Hebrew." As we shall soon see from Johannessohn, the opposite is true of the LXX.
It is obvious that more careful work is needed to determine whether or not it is really plausible to say that Luke was consciously imitating the LXX, or translating a document which did so. The purpose of such an imitation would be, of course, to give a Biblical flavor, such as we would give by injecting thee and thou and similar forms.
Fortunately, we have a basis of comparison from the solid study of Martin Johannessohn on the Biblical kai egeneto and its history36. Johannessohn counted only examples of apodotic kai that come after kai egeneto. However, from seeing the attitude of the LXX translators in that context, we can very reasonably suppose they had a similar attitude in examples that lacked kai egeneto.
Johannessohn tells us that the LXX usually keeps the apodotic kai37. In fact, it always occurs in the books of Numbers, Deuteronomy, Ruth and 4 Kings where the Hebrew has apodotic wau. Outside these books, says Johannessohn, there are only occasional omissions. It appears least in Genesis, but there it is still present 45.83% of the time. Exodus uses it 9 out of 13 times (69.23%). Joshua has it 7 out of 11 times (63.63%). First Kings omits it 5 times, but keeps it more than 25 times.
In all, this is a very heavy percentage of cases in which the LXX reproduces the Hebrew apodotic wau by kai. In fact, Johannessohn has found 9 places in which the LXX adds the kai where the Hebrew lacks the usual wau38. Of course, it is quite possible that at those points the LXX was translating from a Hebrew text different from our text.
Obviously, now, if Luke really were imitating the LXX, or using a source which did so, he too should have had the apodotic kai a very high percent of the time. Otherwise it would be as odd as if a person today were to insert thee, thou etc. only a small fraction of the time. It would sound silly.
Finding no study which gave the count for Luke, it was obviously worthwhile to make that count myself. But two counts were necessary: first, one of instances where Luke does use apodotic kai; second, a count of places where normal Hebrew (and so also the LXX) would use it, but Luke did not.
The rules in classical Hebrew for this wau were in general very well observed, though, of course, there were exceptions39. It is specially common in conditional sentences, especially in casuistic statements (though it would not be used before an imperative in the apodosis). It was also frequent in causal sentences, very frequent in temporal sentences, and even after very brief temporal expressions, such as in Ex. 16.6: "Evening, and you will know". It was also quite frequent in relative sentences.
As to the exceptions: Apodotic kai would not be used before an imperative. But it would be found before the jussive (third person), or the cohortative (first person). It was rarely used before a noun, and not ordinarily found before particles, i.e., negatives, adverbs, prepositions, and conjunctions. Finally, when there were two if clauses each with its own apodosis, the wau was specially necessary before the second apodosis, but was not used before the first.
When we make our count of the use of the apodotic kai, we find Luke has the apodotic kai 17 times40, plus two other times in which there is some ambiguity, inasmuch as the kai could have the sense of also (11.34 and 12.8). Twelve of these examples open with (kai) egeneto.
When we count examples of the omission of apodotic kai, we need to distinguish between instances in which classical rules would call for omission, and those in which they would not, Luke omits apodotic kai in 48 instances in which classical rules would not call for the omission, in instances in which no noun or substantive or imperative follows41.
He also omits it in 14 instances before nouns, pronouns, and adjectives where classical rules would omit it42; and another 16 instances before adverbs, prepositions, negatives and interrogatives where classical rules would also omit it43.
There are two examples also of omission in double conditions, in which kai is left out in both members: in the first member according to classical rule, in the second, in one example, because a noun follows, in the other, when a negative follows44. He also omits it in two instances of genitive absolutes45. Such an omission would have been required in classical Greek, but not in Koine Greek. The LXX sometimes uses kai after a genitive absolute, e.g., 4 Kgs 13.21.
Finally, the kai is omitted in two instances of elliptical protases46; classical Hebrew more commonly omitted the wau with these, but not always.
We can sum up our totals thus:
Apodotic kai used: 17 clear instances, plus 2 ambiguous ones
Apodotic kai omitted:
b) Before nouns, pronouns, adjectives: 14
c) Before adverbs, prepositions, negatives, interrogatives: 16
d) In double conditionals: 2
e) After genitive absolutes: 2
f) After elliptical protases: 2
Since classical Hebrew rules would often call for wau after casus pendens, which resembles a genitive absolute, we may add two instances from Luke to the basic 48 of a) above to reach a total of 50. Also, since the wau is found at times after elliptical protases, we may reasonably take one of the two examples, to give a total of 51 omissions.
As a result, leaving aside the two ambiguous instances where kai does appear, we have a ratio of 17 uses to 51 omissions. That is, we have 17 uses out of a total of 68 possible uses by classical rules. That means that Luke used apodotic kai just 25% of the time. But, as Johannessohn shows, the LXX reproduces apodotic wau most of the time.
Now suppose someone today wanted to use forms like thee and thou to give a Biblical flavor, but used them only one-fourth of the time: he would not only fail to give the flavor, he would get a bizarre result. We conclude: Luke's use of apodotic kai was not due to imitation of the LXX.
Why then did Luke use apodotic kai? Let us say, tentatively, that Luke got his apodotic kai's by translation of Hebrew sources (for that kai was not native to Aramaic).
But now another problem emerges. The use of kai egeneto is also a marker of a Hebrew source (it appears in Aramaic only in the Targums when they are translating Hebrew closely). But Luke uses (kai) egeneto in 20 instances in which he omits apodotic kai47. So, one indication would seem to point to the use of a Hebrew source, while the other would deny it.
However, there is a very plausible solution to the puzzle. We now know that Hebrew was in use in two forms in the first century: in a neoclassical form, and also in a more conservative form in what Fr. Fitzmyer calls "pockets of Palestinian Jews"48.
In these "pockets" people held on to the old language when most of the nation had changed to Aramaic. This shows a conservative bent. Other instances are known in which a language in cut-off areas is very conservative; thus, in some parts of the eastern U.S. there are areas in which a much older form of English is spoken49.
In contrast, it is also well established that the apodotic kai had almost vanished in the revival of the new Hebrew50.
So then, if Luke really did, as he says in his preface, use documents, it easily could have happened that he would have used two kinds of Hebrew documents: one with the apodotic wau, one without it. Yet in both the wa yehi could occur.
Are we straining evidence like Procrustes? Not at all. First, there is no doubt that the two kinds of Hebrew existed. Further, we know definitely that the apodotic wau was almost gone in the new Hebrew51; but we also know that the wa yehi was not entirely gone in the new Hebrew52. And really, this sort of difference is what one would expect: the very odd apodotic wau would more readily disappear than the not-so-strange wa yehi. So we can have two kinds of Hebrew, both with wa yehi, but one lacking the apodotic wau, one still able to have it, out of conservatism.
But we should make our percentages more precise. To do that, we begin by noticing that the nice distinction about omitting the apodotic wau before nouns and particles could fall into disuse in the course of time even in the conservative form of Hebrew. Even in classical times, these fine rules were not always observed, though they were for the most part. Now Luke, though he regularly does seem to keep the old rule of omitting kai before particles, ignores it often before autos: we find that in 9 out of 17 of our clear examples of the use of apodotic kai, he has kai autos 53. In contrast, the Hebrew OT has the equivalent apodotic we hu infrequently54. So we have reason to suspect that, at least in regard to substantives, the old rules had fallen partly into disuse. If this is the situation, we can reasonably add 14 omissions (in which a substantive opens the main clause) to our previous 51 omissions-thinking the 14 omissions are not really due to classical rules. Then we would have a ratio of 65 omissions to 17 uses, which would mean Luke used the apodotic kai only 20.73% of the time.
But since we cannot be certain about the small rules, it is best for us to use a bracket, and say that Luke uses apodotic kai from 20 to 25% of the time (for, as we saw, without the extra 14 omissions, we would have 25%).
To omit the kai from 75% to 80% of the time surely means he was not using the kai to imitate the LXX.
Further reinforcement of this conclusion comes from some added facts. The typical sentence with kai egeneto consists of three members: (1) kai egeneto, (2) a time expression, (3) the apodotic kai with following clause.
We turn to the second of these elements. Johannessohn tells us that the LXX prefers to translate such time expressions (which in Hebrew are commonly be with an infinitive) by a dependent clause with a finite verb, introduced by hos, kathos, henika or hote 55. But Luke shows the reverse preference. He uses a preposition with the infinitive (commonly en to, corresponding to Hebrew be with infinitive-but with no equivalent in normal Aramaic56) 21 times57, while he has the dependent clause only 4 times, with hos58.
Now the fact that this sort of time structure is not striking means it would be a poor means of giving a Biblical flavor. For certain, Luke's pattern in it is not the way to imitate the LXX, which has the reverse preference.
What is our conclusion? I suggest that we take seriously what Luke says in his prologue, namely, that he did use written sources. The sparse distribution of apodotic kai shows that he was not just imitating the LXX but was translating, and translating slavishly59. He must have been translating Hebrew sources at certain points, not Aramaic-for kai egeneto, en to with the infinitive, and apodotic kai are not Aramaic but Hebrew. Further, because he often uses kai egeneto without apodotic kai, he must have had two types of Hebrew sources: some in new Hebrew, some in conservative old type speech, from the "pockets" of Hebrew.
Such slavish translation is, of course, known elsewhere, e.g., in the Old Latin versions of Scripture. And the reason for its appearance is obvious and well known: the translator had such great respect for his text, and used such extreme care for accuracy that he went to the extreme of introducing foreign structures into his translation. Therefore we may assume that Luke must have been a meticulous author.
The careful analysis of Bruce Chilton, mentioned above, by showing Luke's great care to follow tradition, gives us at least some corroboration for our conclusion that Luke was meticulous. For if Luke so carefully followed tradition (as Chilton finds), then that tradition is apt to be more substantive than just imitating the LXX-it would be following traditional Semitic sources, written or oral.
What of the fact that Luke's style is so very different in different passages, i.e., that he does at times write a good quality of normal Greek? Could it be that Luke just chose to use different styles at different times? We must say no. For if it were just a free choice by Luke, we would have to ask: Why the choice? Writers do not normally shift style that way. Luke could hardly have chosen to imitate the LXX closely at times to provide a Biblical favor, and then without reason have dropped that imitation. We could not imagine what such a reason would be. But we can, on the contrary, see a very plausible reason for the variation if we take Luke at his word and affirm that he did use documents.
| 1 | A. Deissmann, Bible Studies, tr. A. Grieve, Edinburgh: T. & T. Clark, 1902. |
| 2 | E.g., J. Wellhausen, Einleitung in die drei ers ten Evangelien, 2nd ed. Berlin, 1911; E. Nestle, Philologica Sacra, Berlin, 1896; C.F. Gurney, The Aramaic Origin of the Fourth Gospel, Oxford, 1922; C.C. Torrey, Our Translated Gospels, New York, 1936; Documents of the Primitive Church, New York, 1941, The Four Gospels, 2nd ed. NY. 1947. |
| 3 | E.g. G. Dalman, The Words of Jesus, tr. D.M. Kay, Edinburgh, 1902; M. Black, An Aramaic Approach to the Gospels and Acts, 3rd ed. Oxford, 1967. |
| 4 | Especially C.C. Torrey (note 2 above). |
| 5 | Op. cit. p. 5. |
| 6 | E.g., E.G. Turner, "The Unique Character of Biblical Greek," VT, V (1955) 478; H.S. Gehman, "The Hebraic Character of LXX Greek", VT 1 (1951) 81-90; Fred L. Horton, "Reflections on the Semitisms of Luke-Acts" in: C.H. Talbert, ed., Perspectives on Luke-Acts, Danville, 1978, esp. pp. 13-14,23. |
| 7 | H.F.D. Sparks, "The Semitisms of St. Luke's Gospel" in JTS 44 (1943) 129-38; "The Semitisms of the Acts" in JTS 1 (new series) 1950) 16-28; "Some Observations on the Semitic Background of the New Testament" in SNTS Bulletin 2 (1951) 33-42. |
| 8 | Art cit. (1943) 132,134. |
| 9 | Ibid. 129. |
| 10 | Art cit. (1951) 39. |
| 11 | Cf. W.R. Farmer, The Synoptic Problem, Dillsboro, NC, 1976; Bernard Orchard, Matthew Luke and Mark, Manchester, 1977; E.P. Sanders, The Tendencies of the Synoptic Tradition, Cambridge, 1969; T.R. Rosche, "The Words of Jesus and the Future of the 'Q' Hypothesis" in JBL 79 (1960) 210-20; Sanders, "The Argument from Order and Relationship between Matthew and Luke" in NTS 15 (1968-69) 249-61; O.L. Cope, Matthew, A Scribe Trained for the Kingdom of Heaven in CBQ Monograph Series 5, 1976, esp. p. 12: John M. Rist, On the Independence of Matthew and Mark, Cambridge, 1978; Hans-Herbert Stoldt, History and Criticism of the Marcan Hypothesis, tr. D.L. Niewyk, Macon and Edinburgh, 1980. |
| 12 | Note 8 above. |
| 13 | Art cit. (1943) 131-32. The two are: "the verb 'to be' with the participle in place of the finite verb ... and the phrase 'to begin to do something"'. |
| 14 | Art cit. (1951) 38. |
| 15 | Bruce D. Chilton, God in Strength. Jesus' Announcement of the Kingdom, in Studien zum Neuen Testament und seiner Umwelt, Freistadt, 1979, esp. pp. 123-77. |
| 16 | Art. cit. (1943) 133. |
| 17 | W. Bauer, W. Arndt, F.W. Gingrich, and F.W. Danker, A Greek-English Lexicon of the New-Testament, Chicago, 2nd ed. 1979. 270. |
| 18 | Often with a subject and other words too. |
| 19 | H.W. Smyth, A Greek Grammar for Colleges, New York, 1920, §§ 2033, 2034-37. |
| 20 | Dalman, 33. |
| 21 | Cf. M. Wilcox, The Semitisms of Acts, Oxford, 1965, 180. |
| 22 | J.D. Denniston, The Greek Particles, Oxford, 1959, 308-09. |
| 23 | F. Blass and A. Debrunner, A Greek Grammar of the New Testament, tr. R.W. Funk, Chicago, 1961, 227, § 442.7. |
| 24 | Beyer, 68. |
| 25 | It is not native to Aramaic: Klaus Beyer, Semitische Syntax im Neuen Testament, Band I. Göttingen, 1962. 30. |
| 26 | Ibid. 66. |
| 27 | Ibid. 67. |
| 28 | Ibid. 30. |
| 29 | Ibid. 30. |
| 30 | Art. cit. 4. |
| 31 | Ibid. 6. |
| 32 | M. Zerwick, Graecitas Biblica, Romae, E Pontificio Instituto Biblico, 4a ed. 1960. § 389; Sparks, art. cit. (1943) 132. |
| 33 | Horton, art. cit. 4,6,13-14,23. |
| 34 | Torrey, 64-73. |
| 35 | Zerwick § 389 and note 1. |
| 36 | Martin Johannessohn, "Das biblische kai egenéto und seine Geschichte" in Zeitschrift für Vergleichenden Sprachforschung, 1926, 161-212. |
| 37 | Ibid. 184. |
| 38 | Ibid. 190. 3 Kgs 8,54; 14,28; 15,29; 16,11; 17,17; Judges 2,19; 2 Chr 24,4; Ezech. 1,1. |
| 39 | Paul Joüon, Grammaire de l'Hebreu Biblique, 10th ed. Rome: Institut Biblique Pontifical, 1947. § 176. |
| 40 | In Lk 2,21; 2,27; 5,1; 5,12; 5,17; 7,12; 8,1; 8,22; 9,28; 9,51; 13,25; 14,1; 17,11; 19,1; 19,15; 24,4; 24,15. |
| 41 | In Lk 1,8; 1,23; 1,41; 1,44; 1,59; 2,1; 2,6; 2,22; 2,39; 2,42; 2,46; 4,8; 5,4; 5,13; 6,13; 7,1; 7,11; 8,40; 9,18; 9,24; 9,33; 9,36; 9,37; 10,6; 11,1; 11,24; 11,27; 11,36; 12,9; 12,10; 12,45; 15,25; 15,30; 16,9; 16,30; 17,6; 17,15; 17,33; 17,34; 19,5; 19,29; 19,41; 20,1; 21,31; 22,14; 22,66; 24,30; 24,51. |
| 42 | In Lk 2,15; 8,42; 9,29; 9,48; 10,38; 11,19; 11,22; 12,38; 13,3; 13,5; 16,11; 16,12; 18,35; 19,40. |
| 43 | In Lk 6,32; 6,33; 6,34; 10,13; 11,8; 11,13; 11,18; 11,20; 11,21; 12,39; 12,54; 16,31; 18,4; 19,31; 21,20; 23,31. |
| 44 | In Lk 20,5-6; 22,68-69. |
| 45 | In Lk 4,42; 22,10. |
| 46 | In Lk 5,36 and 37. |
| 47 | In Lk 1,8; 1,23; 1,41; 1,59; 2,1; 2,6; 2,46; 7,11; 9,18; 9,33; 9,36; 11,1; 11,27; 17,15; 19,29; 24,30; 24,51; 2,15; 9,29; 18,35. |
| 48 | J.A. Fitzmyer, "The Languages of Palestine in the First Century A.D.: in CBQ 32 (1970) 501-31, at 531. |
| 49 | Cf. H.L. Mencken, The American Language, 3rd ed. New York, 1923, 71-72: "In remote parts of the United States there are still direct and almost pure-blooded descendants of those seventeenth century colonists. Go among them and you will hear more words from the Elizabethan vocabulary, still alive and in common service, than anywhere else in the world, and more of the loose and brilliant syntax of that time and more of its gypsy phrases". Cf. also T. Pyles, The Origin and Development of the English Language, 2nd ed. New York, 1971, 273. |
| 50 | Beyer, op. cit., 67. |
| 51 | Beyer, 67; Horton, art. cit. 4. |
| 52 | Beyer, 30, says that it retreats, "geht ... züruck", but that it does not completely disappear until the Mishna. |
| 53 | In Lk 2,27; 5,1; 5,17; 8,1; 8,22; 9,51; 14,1; 17,11; 19,1; 24,4; 24,15. |
| 54 | Johannessohn, art. cit. 190. It occurs mostly only in 1 Kgs and 3 Kgs. |
| 55 | Art. cit. 199 and 201. |
| 56 | Ibid. 199. |
| 57 | Johannessohn found only 19 instances. I found 21 as follows: in Lk 5,1; 5,12; 9,51; 14,1; 17,11; 19,15; 24,4; 24,15; 1,8; 2,6; 8,40; 8,42; 9,18; 9,29; 10,38; 11,1; 11,27; 17,15; 18,35; 24,30; 24,51. The first 8 examples have apodotic kai, the latter 13 do not. |
| 58 | In Lk 1,23; 1,41; 19,29; 2,15. No apodotic kai in these examples. |
| 59 | For an interesting discussion of the influences of bilingualism, see M. Silva, "Bilingualism and the Character of Palestinian Greek", in Biblica 61 (1980) 198-219. Of course, bilingualism could not account for using a structure only 20-25% of the time. |
