Friday, May 28, 2010

More on the Flaw in McIver and Carroll's article on detecting copying in the Gospels

In a recent post, I attempted to explain A Flaw in McIver and Carroll's Article about detecting the existence of written sources in the Gospels. The difficulty is that they did experiments with students in English and then applied the resulting criterion (16 or more words suggests copying) to texts in Greek, a problem given the fact that Greek invariably users fewer words to convey similar material. The difference is usually substantial and sometimes overwhelming. In one case, a 24 word conjoined sequence in Greek is represented in standard English translations by a 42 word conjoined sequence (Matt. 7.7 // Luke 11.9).  In other words, the 16 (or 18) word criterion is far too restrictive as a criterion of copying, and should have been greatly reduced to adjust for differences in language.

The difficulties with illegitimate transfer between English experiments and Greek texts does not end there. In this post, I would like to draw attention to a further, related issue that is not considered by McIver and Carroll: they are comparing results from experiments using a non-inflected language (English) with an inflected language (Greek). This is important because their experiments focus specially on conjoined sequences of words. A non-inflected language like English is far more likely to retain word order across two parallels because word order is often intrinstic to the meaning. In Greek, by contrast, word order is less important in the determination of meaning. This has the effect of diminishing the likely extent of parallel conjoined sequences in Greek.

The point can be illustrated with a simple look at the Synopsis. Matt. 9.5-6 // Mark 2.9-10 // Luke 5.23-4 features an eighteen word triple agreement in English translation, the kind of agreement that would satisfy even the stricter McIver and Carroll criterion. Here's the passage in the RSV of Mark, with the triple sequential agreement of eighteen words underlined:

. . . 'Your sins are forgiven,' or to say, 'Rise, take up your pallet and walk'? But that you may know that the Son of man has authority on earth to forgive sins" --he said to the paralytic . . .

In Greek, however, there is no such lengthy agreement because of differences in the ordering of the elements translated above as "has authority" and "Son of Man" and "upon the earth". Take a look at the texts:

Matt. 9.5-6: . . . Ἀφίενταί σου αἱ ἁμαρτίαι, ἢ εἰπεῖν· Ἔγειρε καὶ περιπάτει; ἵνα δὲ εἰδῆτε ὅτι   ἐξουσίαν ἔχει   ὁ υἱὸς τοῦ ἀνθρώπου   ἐπὶ τῆς γῆς   ἀφιέναι ἁμαρτίας, τότε λέγει τῷ παραλυτικῷ . . . .

Mark 2.9-10: . . . . Ἀφέωνταί σου αἱ ἁμαρτίαι, ἢ εἰπεῖν· Ἔγειρε καὶ ἆρον τὸν κράβαττόν σου καὶ περιπάτει; ἵνα δὲ εἰδῆτε ὅτι   ἐξουσίαν ἔχει   ὁ υἱὸς τοῦ ἀνθρώπου   ἀφιέναι ἁμαρτίας   ἐπὶ τῆς γῆς, λέγει τῷ παραλυτικῷ . . .

Luke 5.23-4: . . . . Ἀφέωνταί σοι αἱ ἁμαρτίαι σου, ἢ εἰπεῖν· Ἔγειρε καὶ περιπάτει; 24 ἵνα δὲ εἰδῆτε ὅτι   ὁ υἱὸς τοῦ ἀνθρώπου   ἐξουσίαν ἔχει   ἐπὶ τῆς γῆς   ἀφιέναι ἁμαρτίας, εἶπεν τῷ παραλελυμένῳ . . .

What this illustrates is that the chances for conjoined sequence in parallel passages is greater in contemporary English than it is in first century Greek.  Where English translations of this passage have eighteen words in conjoined sequence across all three Synoptics, the Greek has just eleven (Matthew // Mark) or five (Mark // Luke and Matthew // Luke).

The same kind of thing is easy to illustrate from elsewhere.  Take the following parallel from Matthew and Mark, for example, in which there is conjoined sequence for nineteen English words of the NRSV translation:

Matt. 26.13 (NRSV): Truly I tell you, wherever this good news is proclaimed in the whole world, what she has done will be told in remembrance of her.”

Mark 14.9 (NRSV): Truly I tell you, wherever the good news is proclaimed in the whole world, what she has done will be told in remembrance of her.”

Now contrast this with the Greek texts (NA27) of Matthew and Mark:

Matt. 26.13: ἀμὴν, λέγω ὑμῖν, ὅπου ἐὰν κηρυχθῇ τὸ εὐαγγέλιον  τοῦτο ἐν  ὅλτκόσμῳ,   λαληθήσεται   καὶ ὃ ἐποίησεν αὕτη   εἰς μνημόσυνον αὐτῆς.

Mark 14.9: ἀμὴν δὲ, λέγω ὑμῖν, ὅπου ἐὰν κηρυχθῇ τὸ εὐαγγέλιον  εἰς ὅλον τὸν κόσμον,   καὶ ὃ ἐποίησεν αὕτη   λαληθήσεται   εἰς μνημόσυνον αὐτῆς.

The differing position of λαληθήσεται ("will be told") in Matthew and Mark generates a shorter conjoined sequence of words in Greek than one sees in English translation, only four words in the comparison with the nineteen in the English of the NRSV.

In other words, English is more likely to feature a lengthy conjoined sequences of words given that it is a non-inflected language with fewer options in relation to the ordering of words.  This should have been factored into McIver and Carroll's attempt to transfer their contemporary English experimental results to ancient texts written in Greek.


James said...

Is it relevant that Jesus spoke Aramaic, so that his sayings were first remembered in this language and then translated--by the rememberer, by the hearer, by the transcriber?--into Greek?

Mark Goodacre said...

I wouldn't have thought so, at least in the context of these experiments, but I could be wrong.

judyredman said...

So what we need is someone teaching in a Greek university to replicate the experiments and see what happens? I know that modern Greek is not the same as Koine, but it comes much closer than English.

I have a rampaging head cold and am not thinking as clearly as I might, but it would seem to me that the fact that word order is not as important in Greek is less relevant when working with a hypothesis that the original text was written than it is when working with an oral original text, though. ISTM that if you have a written text in front of you, you would copy it faithfully unless you felt that it needed altering, so alterations in word order etc would be deliberate. If you were working from an oral text, then the kind of differences between Matt 26 and Mark 14 could be considered a totally faithful representation of the oral text.

Mark Goodacre said...

The way I see it, Judy, is that McIver and Carroll are specifically trying to design experiments that will "develop criteria for determining the existence of written sources". The fact that they introduce this criterion of 16 (18) words in conjoined sequence simply does not take seriously the way ancient Greek works. What I am suggesting is that the differences between languages should have been factored in. I am worried about going beyond that to any further speculations about what their results could have shown if they had run the experiments differently.

Jim Deardorff said...

In any future test resembling that of McIver & Carroll, the motivating theme ought to be to determine the degree of sequential verbal agreement along the lines indicated by Judy Redman stressing redaction criticism, and not that due to imperfect memory. Four basic categories of editing may be suggested for exploration:
(a) Replication of the existing text (faithfully);
(b) Same as (a) except for purposeful improvements in grammar and factuality,
(c) Same as (b) except for additional alterations expressing the author's different theology and mental outlook; and
(d) Same as (b) or (c) except in addition the purposeful replication of a significant number of lengthy word strings (identical words in unaltered sequence as per Poirer, JBL 123 (2004) 315).

One does not need student experimentation for this, but needs to look at the frequency distribution of the length of unaltered word strings between parallel passages. This can be done not only for the Gospels, but for other cases such as comparing the Greek text of 2 Chr 35-36 and all of Ezra and Neh 7:73-8:12, with 1 Esdras. In such a case one finds an exponential frequency distribution as in (b) above. But for Matthew/Mark and Matthew/Luke, one finds an anomalous tail to the respective frequency distribution, as in category (d).

With Mk/Luke category (c) seems to apply, as the two anomalies indicated by Poirer (p. 320) are too small in number for (d) to apply.