This article is also available in transliteration and text-only formats.

Reuben J. Swanson, ed. New Testament Greek Manuscripts: Variant Readings Arranged in Horizontal Lines against Codex Vaticanus. 4 vols. Foreward by Bruce Metzger. Sheffield: Sheffield Academic Press, 1995; Pasadena: William Carey International University Press, 1995. ISBN (Sheffield hardcover): 1-85075-595-7, 1-85075-596-5, 1-85075-597-3, 1-85075-598-1; (Sheffield softcover): 1-85075-772-0, 1-85075-773-9, 1-85075-774-7, 1-85075-775-5; (Carey softcover): 0-86585-051-8, 0-86585-052-6, 0-86585-053-4, 0-86585-054-2. Pp. xxi+304+xix+271+xx+420+xix+302. Separate volumes from Sheffield (hardcover/softcover): Matthew: UK £43.00/£19.95; US $64.00/$29.95; Mark: £39.00/£18.50; US $58.50/$26.95; Luke: £50.00/£24.95; US $75.00/$39.00; John: £43.00/£19.95; US $64.00/$29.95. 4 volumes from Carey (softcover): US $34.50.

1. Swanson's mammoth new text-critical edition of the Gospels seems to be both innovative in presentation and exhaustive within its compass. Indeed, the detailed thoroughness with which the textual material in the selected Greek witnesses is presented may be unequalled.

2. The 71 Greek witnesses collated for this new textual apparatus include the following, identified by the usual Gregory/Aland notation:

• ^{1 2 5 6 19 22 25 35 36 37 39 45 52 53 59 60 62 63 66 75 76 82}
• ) A B C D E F G H K L M N P Q S T U W Y G D Q L P Y W 0171
• 1 2 13 28 33 69 118 124 157 565 579 700 788 1071 1346 1424 1582
• Clement of Alexandria, TR Oxford 1873, Westcott-Hort, and UBS⁴

3. The main body of this work is organized on each page into two sections and four apparatuses. The first and main section presents the readings found in all the textual witnesses, quoted in full, vertically aligned to match up corresponding words in each reading, something like this (Greek transliterated):

	proseuxesqe	peri	twn	ephreazontwn	umas				B ) L 579 700 u w
	pro........	.eri	tw.	.phreazo....	....				75
kai	proseuxesqe	peri	twn	ephreazontwn	umas				W
	proseuxesqe	uper	twn	ephreazontwn	umas				L
	omit								f13 2*
	proseuxesqe	uper	twn	ephreazontwn	umas	kai	diakontwn	umas	69 124
	proseuxesqe	uper	twn	epereazontwn	umas				33 1071
	proseuxesqe	uper	twn	ephreazontwn	umas	kai	diakontwn	umas	788
kai	proseuxesqe	uper	twn	ephreazontwn	umas				TR
	proseuxesqe	uper	twn	ephreazontwn	umas				rell

4. Each text is quoted in extenso, but identical readings are collected on one line. In the right margin of each line of text are the sigla of the witnesses supporting that form of text. The reading of Vaticanus is always on the first line, and either the Majority Text reading or the reading of the TR generally appears last, if different. As one can see from the example above, the material in one stanza of lines does not represent one unit of textual variation, but often several units combined together. This makes it easier to do continuous reading (or continuous searching) of the text of a single manuscript, but the arrangement in long lines makes it harder to separate out the units of variation for individual analysis. Three group sigla are used to save space and mental energy.

• W
• f¹ = 1 118 1582
• f¹³ = 13 69 124 788 1346

5. Swanson's new finding that the minuscule 1346 belongs to family 13 (and his citing it under that siglum in the apparatus) looked suspicious at first, because von Soden, Geerlings, and Wisse all classify 1346 as a member of family P. After I browsed through sections of Swanson, noting how often 1346 appeared on the same line with P or with f¹³, I decided that 1346 appears to have suffered block mixture, with short blocks of a few chapters in length taken from either family P or 13 alternately.

6. The second section on each page of the text lists mss with lacunae on that page. Unfortunately, the papyri other than ⁴⁵ and ⁷⁵ are not listed here, and the only way to find out where they are extant is to check other reference works, such as the appendices in the Nestle-Aland 27th edition (NA²⁷). Cursory checks suggest that some of the minuscules may also have undocumented lacunae. The extent of the text collated from each fragmentary ms often differs from the limits described by Aland. This is possibly due to the NA²⁷ list of mss combining under one siglum fragments preserved at different sites, if it is judged that the fragments were originally part of one ms. Lacunae are indicated precisely in the main text by ellipses (...), which are necessary to distinguish the lacuna from the blank space needed to align shorter texts vertically.

7. The first apparatus below these two sections lists spelling problems in the mss. The spelling of names is fully documented in the main text section, but otherwise minor itacisms and transparent blunders are regularized in the main text and the original spelling noted here. This helps a lot in uncluttering the presentation of the evidence, and it also helps the reader identify individual scribal weaknesses and habits.

8. The second apparatus documents the nomina sacra originally found in the mss, before such contractions were expanded as they appear in the main text. The third apparatus documents the kefalaia, titloi, lection arxh and teloj marks, and incipit texts found in the mss. The fourth apparatus lists the chapter and section numbers found in the margins and the Eusebian canon tables.

9. Pericopes are labeled with English section headings derived from those in UBS⁴, with some Greek titloi, and with synoptic cross references. Septuagint quotations are printed in bold, poetry is indented, and textual variations are highlighted by underlining to indicate which words differ from Vaticanus. The Clement quotations are given with some context, and the portion matching the scripture is underlined.

10. A preliminary printing of the Matthew volume appeared in 1994, which I will not discard, because it contains an extra appendix of textual commentary for chapters 1-19 not found in the final printing. There does not seem to be any other material difference between the preliminary and final printings. I did substantial work with this preliminary printing, transcribing the Freer Gospels text in Matthew and searching for singular readings in several uncials. I found the format clear and easy to work with.

11. I am accumulating an errata list, which is available on the Web and by FTP, and which at present lists 66 potential errors, including both minor typographical problems and significant textual errors. I found these errors by referring only to secondary sources, not to microfilm images of the mss. Contributions of error reports to broman@nosc.mil are welcome.

12. Every book contains some errors, but the degree of reliability expected of a textual apparatus is a critical factor in its being useful and its staying in demand. Tischendorf's reputation for accuracy has kept his eighth edition popular for more than a century, while von Soden's and Legg's major works have not been widely used, partly because of the perception that their error rate was too high.

13. I performed a simple test of Swanson's accuracy in collation by comparing the apparatus in NA²⁷ for the 22nd and 23rd chapters of Luke against the corresponding section of Swanson. The NA²⁷ apparatus for these two chapters reports on 215 units of variation, and there were 25 manuscripts, some fragmentary, which are consistently cited by both NA²⁷ and Swanson. Discrepencies between Nestle-Aland and Swanson were found at 33 places in these two chapters. I attempted to adjudicate these discrepencies by referring to the IGNTP apparatus as an independent tie-breaker. Whichever source was supported by the IGNTP was considered to be correct in that place, and the other source to have erred. In Lk 23:45, the reading of the first corrector of C could not be determined from these secondary sources. In Lk 22:18 it appears that all three sources erred in different ways in representing the reading of W. Without access to an image of the Freer Gospels ms, I conjectured that the most likely reading was piw apo tou nun genhmatoj, which is the reading reported in Goodspeed's collation of W.

14. The final tally indicates that in these 33 disagreements, NA²⁷ was at fault 12 times, and that Swanson erred 21 times, One could accordingly estimate that his error rate is probably less than twice that of NA²⁷, and would amount to around .01 errors per manuscript per unit of variation. This is a respectable showing when measuring oneself against the latest Nestle-Aland edition, a respected production for which the collations have been checked and rechecked by a substantial staff over many years. Besides, the NA²⁷ data had to be given the benefit of the doubt in several cases during this side-by-side test because of the selective nature of its apparatus, while the Swanson edition aspires to be exhaustive and offers no excuses.

15. One other concern arose in my mind during these tests: a doubt whether defections from group readings of f¹, f¹³, and were invariably reported. To test Swanson on this point, I searched the IGNTP apparatus in Lk 22:1-10 and found 41 instances where the mss assigned by Swanson to the groups f¹, f¹³, or did not read unanimously as a group. In 29 of these 41 instances a defection was documented by Swanson, and in 12 instances the defection was not noted (or was due to IGNTP error). This result suggests that a failure to note group defections could be of significant concern. But to be fair, this kind of detailed information is completely absent from sources like NA²⁷, which cites f¹ and f¹³ by group siglum only.

16. Swanson's publication of Greek manuscript texts of the Gospels seems to be reasonably accurate, and it is valuable both for its approachable full text format and its no-sparrow-may-fall coverage of the material. The appearance of the book of Acts in the near future is eagerly awaited.

© TC: A Journal of Biblical Textual Criticism, 1996.

Vincent Broman
San Diego, California, USA