Ulrich Luz. Matthew 8-20. Translated by James E. Crouch. Edited by Helmut Koester. Hermeneia. Minneapolis: Fortress, 2001. Pp. xxxvii + 608. ISBN 0-8006-6034-X. US $69.00.

1. This is volume two of three projected volumes containing the author's commentary on Matthew. It was translated from volume two and part of volume three of the four-volume German original. The English volume three will correspond to part of volume three and all of volume four in the German. After the appearance of volume three, a revised edition of volume one is due to appear.

2. We can be glad that this excellent commentary is appearing in English. It is well produced and contains detailed discussion in a wide range of areas. Consequently, it should be regularly used by scholars with many different interests. It seems particularly to excel in its full treatment of the history of interpretation.

3. The actual text-critical comment is in general rather brief, and at several points one may read thirty or so pages without finding a text-critical remark. Reference is made only to the 26th rather than 27th edition of the Nestle-Aland Novum Testamentum Graece, reflecting the date of publication of the original German edition. This does not seem to matter much, however.1

4. Though there is little evidence of detailed work with primary texts, Luz gives thoughtful reasons at several points for going against the reading of the Novum Testamentum Graece, usually preferring the shorter text:

As might be expected, a slightly more extended discussion is given for not treating Matthew 16:2b-3 as authentic (p. 347). The most detailed text-critical discussion is of the patristic variants on the verb "know" in Matthew 11:27 and of patristic variants on the order of clauses in that verse (pp. 155-156). Luz specially thanks coworkers of the Institut für Neutestamentliche Textforschung for help with this discussion, which contains original material. One may be disappointed by the lack of discussion when Luz adopts the reading of o)/xlon in Matthew 8:18 without even mentioning the existence of variants and that all witnesses except B samss read something else (p. 15). Likewise, an excursus on the name "Matthew" covers variant Semitic spellings, but does not mention the spelling Matqai=oj as an alternative to Maqqai=oj (p. 32).

5. We come to consider a few specific examples of Luz's treatment, where it may be possible to improve upon his analysis.

6. In Matthew 13:35 Luz wants to follow )* Q f 1.13 et al. in reading "Isaiah the prophet" rather than just "the prophet" before a quotation from Psalm 78:2 (attributed in MT to Asaph): "It is clearly the lectio difficilior" (p. 265). He appeals to Matthew 27:9 as an analogous case where an original error has been corrected in some manuscripts. As so often, theological concerns are assumed to have been paramount in the transmission of texts when there is actually strong evidence against this. As is known, Matthew's gospel contains a number of fulfillment formulae: 1:22, (2:5,) 2:15, 2:17, 3:3, 4:14, 8:17, 12:17, 13:35, 21:4, 27:9. In four of these "Isaiah" is firmly (and correctly) embedded in the manuscript tradition (3:3, 4:14, 8:17, 12:17), and in another (2:17) the name "Jeremiah" likewise (correctly) occurs. For all of the others there are some witnesses containing the name "Isaiah" alongside the word "prophet":

Clearly at sundry times and in divers manners, for a scribe to use the name "Isaiah" was a lectio facilior. After all, with four of the fulfillment formulae having the name "Isaiah" so firmly fixed in them, assimilatory pressures in the transmission process would have ensured that the name spread. If such an error occurred sufficiently early, it could easily have produced the diversity of witnesses attesting "Isaiah" in Matthew 13:35. Appeal to Matthew 27:9 in fact shows just the opposite of what Luz wants to illustrate. The number of witnesses that have attempted to alter the problematic "Jeremiah the prophet" is remarkably few. So why should we suppose that the same manuscript traditions that left "Jeremiah" in Matthew 27:9 must have felt the need to remove "Isaiah" from Matthew 13:35?

7. Commenting on Matthew 15:14 (noted above), Luz says:

Tuflw~n (thus Nestle26) is missing in the best MSS. The shorter original text is represented by )* B D. It may be that a scribal error first changed tufloi/ to tuflw~n (K pc sys, c), then tufloi/ was added again. (p. 325)
This analysis is problematic. Nestle-Aland26/27 reads tufloi/ ei0sin o9dhgoi\ [tuflw~n] and continues tuflo\j de\ tuflo\n ktl. The three witnesses for the shorter text read as follows: The witnesses then continue, tuflo\j de\ tuflo\n ktl. What Luz does not mention is the possibility that tuflw~n was omitted by homoeoarcton as a scribe read TUFLWNTUFLOS. This is surely a much more widely attested scribal error than the error he suggests, namely that tufloi/ was changed to tuflw~n. Moreover, whereas homoeoarcton explains the shorter readings by means of one error, Luz has to suppose that two errors occurred in most witnesses (the change of tufloi/ to tuflw~n and the reintroduction of tufloi/). Furthermore, whereas Luz talks of the "shorter original text," he nowhere tells us whether this is the reading of )* (odhgoi eisin tufloi) or B (tufloi eisin odhgoi). If we suppose he means the reading of B, then Luz is positing that a scribe changed tufloi eisin odhgoi to tuflwn eisin odhgoi--a reading which supposedly generated all other versions of the long reading, but which is itself no longer extant. Moreover, to add tufloi to this in the only place possible would create the order tuflwn eisin odhgoi tufloi, of which I can find no trace. We can only conclude that by the "shorter original text," Luz did not mean the reading of B or its graphic variant in D. His "shorter original text" must be the reading of )* alone. Now let us follow his mechanism through: odhgoi eisin tufloi ex hypothesi was altered to odhgoi eisin tuflwn, a reading found in K. tufloi was then added to this reading to give odhgoi eisin tufloi tuflwn. As noted earlier, this involves two steps, while the hypothesis of homoeoarcton involves but one. It also involves the oddity (but not impossibility) that the supposedly second generation reading (odhgoi eisin tuflwn) is less well supported than either the first generation (odhgoi eisin tufloi) or third generation (odhgoi eisin tufloi tuflwn). What makes Luz's explanation less satisfactory is that it does not actually explain the distribution of readings, especially that of )1 L Q f 1.13 et al., which read tufloi eisin odhgoi tuflwn. On his hypothesis, the only way to get this would be to make it a fourth generation reading, derived from the third generation odhgoi eisin tufloi tuflwn, but with a change in order. Thus odhgoi eisin tufloi tuflwn and odhgoi eisin tufloi, though only one step away from each other, would be defined as fourth and second generations respectively in relation to the original. Rather than dividing, as Luz does, between "long" readings (including both tufloi and tuflwn) and "short" readings (with only tufloi), it is useful to distinguish between those readings that place odhgoi before eisin and those that place it after. Once readings are judged to be close (i.e., only one step apart) on the basis of where they place odhgoi in relation to eisin and we reject Luz's theory that tuflwn was ever substituted for tufloi, we are able to derive any of the extant readings (spelling variants aside) by a two-step process from any of the readings tufloi eisin odhgoi, odhgoi eisin tufloi, tufloi eisin odhgoi tuflwn, or odhgoi eisin tufloi tuflwn. When it is recognised that haplography (by homoeoarcton) can more readily get rid of two similar but not identical forms than dittography can create two different forms, it is seen that the longer readings, including tuflwn, are preferable. Of these the reading of C W et al., by presenting the triple sequence TUFLOITUFLWNTUFLOS, would almost certainly invite haplography, which would be most likely to affect the middle item of the sequence. Thus Luz's proposal leads directly away from a solution.

8. On p. 345 Luz accepts the reading Magadan in Matthew 15:39 and comments: "Many manuscripts surmise that it means Magdala on the western shore of the lake. In addition, there is an analogy in Josh 15:37 that suggests that Magadan could have been a popular form of the name Magdala." A footnote tells us: "The LXX reproduces Migdal Gad with Magadagad." Here the LXX is defined as B. Joshua 15:37 may well help us with Matthew 15:39, but, given the similarity in some hands between alpha, delta and lambda, MAGADA and MAGDALA are not a particularly good basis for positing new popular forms of names. We must equally allow for inner-Greek corruption to have occurred.

9. On p. 349 there is a brief discussion of Matthew 16:12, where Nestle-Aland26/27 read prose/xein a0po\ th=j zu/mhj tw~n a!rtwn, but Luz inclines to a reading without tw~n a!rtwn. He cites the Peshitta along with other witnesses as reading tou= a!rtou rather than tw~n a!rtwn. This is to conclude too much from a translation. The Peshitta could support either the singular a!rtoj or plural a!rtoi, since the singular would be demanded by Syriac idiom regardless of the number in its Vorlage (Williams 2001).

10. The decision of the commentator to focus on matters of interpretation should not be criticised, even though text critics might have wished for more frequent and detailed discussion of their interests. Commentators are, after all, free to write the style of commentary they wish in accordance with the guidelines of the series. What perhaps gives some justification to the sense that the treatment is not properly balanced is that considerable attention is given to the relationship between Matthew and reconstructed or hypothetical sources (Q, QMt, Aramaic sayings, pre-Markan oral tradition, etc.). The study of sources is not to be disparaged, but it is rather ironic if we study texts that we do not have more carefully than texts (i.e., manuscripts) that we do.

11. Readers may undertake a certain amount of textual criticism of their own as they read this commentary. Typographically, it is produced at least on a par with, if not above, average levels of accuracy. One can see, however, that certain minor glitches in Greek or Hebrew fonts have arrived in the transfer of files. Though readers can easily make allowance for these, they ought also to know that the siglum l in this commentary clearly stands for the Gothic M of the Byzantine text (pp. 186, 349, 405, 509). Moreover the translator does not always choose correctly in deciding whether German syrisch should be rendered "Syriac" or "Syrian" (e.g., pp. 23 and 486).

© TC: A Journal of Biblical Textual Criticism, 2003.


1Unless one notes that Luz cites Origen as supporting an addition in Matt 10:23 that is given as "Or" in NA26 but Orpt in NA27. image


Williams, P. J. 2001. "Bread and the Peshitta in Matthew 16:11-12 and 12:4." Novum Testamentum 48: 331-333.
P. J. Williams
Divinity and Religious Studies
University of Aberdeen