Hand to Hand Combat: Codex D vs 2397

GA 2397 - The end of John,
with a colophon.

            In the year 1303, a copyist named Hyacinthus finished the text of a Greek manuscript of the Gospels.  Below the closing verses of the Gospel of John, he wrote a little note which went something like this:  first, in red:  “The work is completed, glory to our God, in the year 6811 [from the beginning of the world].”  And then:  “Written by the hand of Hyacinthus, sinner and writing-specialist.   Reader, pray, and curse not by the Lord that the writer has finished.  And may the Lord save you all, brothers!  Amen and amen and amen.”   

            This volume, now known as minuscule 2397, is catalogued as manuscript 135 in the Goodspeed Manuscript Collection at the University of Chicago.  It is also known as the Hyacinthus Gospels, in honor of the copyist who wrote its text.  

            Hyacinthus was, it seems, devoutly dedicated to his work.  But how good was the accuracy of the text he wrote?  Let’s take a sample of the contents of minuscule 2397 and compare it to the contents of a much older manuscript – Codex Bezae.  Researcher Dr. David Parker has assigned Codex Bezae to “about 400.”  (Other researchers have given it a production-date a century or so younger.)

            Using John 15:1-10 as printed in the Nestle-Aland Novum Testamentum Graece (27th edition) as the standard of comparison, let’s list all the differences between it and the text of 2397.  Then we will set aside differences that are usually considered minor – differences that involve vowel-exchanges, spelling, and word-order – and see how materially different the text of 2397 is from the base-text of English versions such as the ESV and CSB.  Then we will do the same kind of comparison, using the text of John 15:1-10 made by the copyist of Codex Bezae. 

2397:  Comparison to NA27

(Black diamonds signify readings that diverge from the Byzantine Textform)

1 – 2397 reads αληθηνη instead of αληθινη (+1, -1) ♦

1 – 2397 reads εστι instead of εστιν (-1)  ♦

2 – 2397 reads φερων instead of φερον (+1, -1) ♦

2 – 2397 transposes, reading πλειονα καρπον instead of καρπον πλειονα

2 – 2397 reads φερει instead of φερη (+2, -1) ♦

3 – no variants

4 – 2397 reads μεινη instead of μένη (+2, -1)

4 – 2397 reads μεινητε instead of μένητε (+2, -1)

5 – no variants

6 – 2397 reads reads μεινη instead of μένη (+2, -1)

6 – 2397 does not have το before πυρ (-2) ♦

6 – 2397 reads βαλλουσι instead of βαλλουσιν (-1) ♦

7 – 2397 reads αιτήσασθε instead of αιτήσεσθε (+1, -1) ♦

8 – 2397 reads γενησεσθε instead of γένησθε (+2)

9 – 2397 reads ηγάπησε instead of ηγάπησεν (-1) ♦

9 – 2397 transposes, reading ηγαπησα υμας instead of υμας ηγαπησα.

10 – no variants

            Thus, in John 15:1-10, 2397 contains 13 non-original letters, and is missing 12 original letters, for a total of 25 letters’ worth of corruption.  When all movable-nu variants and normal vowel-exchanges are removed from the equation, there are just two substantial variants in 2397:  the non-inclusion of το before πυρ in verse 6, and the reading  γενησεσθε in verse 8 (a reading supported by Codex Sinaiticus and Codex Alexandrinus), yielding four letters’ worth of corruption in these 10 verses. 

            Now let’s make the same kind of comparison, using Codex Bezae’s text of John 15:1-10:

Codex D:  Comparison to NA-27
(This takes into account the text made by the scribe, not corrections that were introduced later.)

1 – no variants

2 – D reads καρπο instead of καρπον (-1)

2 – D reads φορον instead of φερον (+1, -1)

2 – D reads καθαριει instead of καθαίρει (+1, -1)

2 – D transposes, reading πλειονα καρπον instead of καρπον πλειονα

3 – D omits all of verse 3, and the first part of verse 4 up to (and including) the word φέρειν.  That is, D omits ηδη υμεις καθαροί εστε δια τον λόγον ον λελάληκα υμιν· μείνατε εν εμοι, καγω εν υμιν. Καθως το κλημα ου δύναται καρπον φέρειν (-100)

4 – D reads μεινη instead of μένη (+1)

4 – D reads μεινητε instead of μένητε (+1)

5 – D reads γαρ after εγω (+3)

5 – D does not have εν before εμοι (-2)

6 – D reads δυνασθα instead of δυνασθε (+1, -1)

6 – D does not have ουδεν at the end of the verse (-5)

6 – D reads αυτο instead of αυτα (+1, -1)

6 – D does not have το before πυρ (-2)

6 – D reads καιετε instead of καιεται (+1, -2)

7 – D reads δε after εαν (+2)

7 – D reads υμειν instead of υμιν (+1)

8 – D reads αιτησασθαι instead of αιτησασθε (+2, -1)

8 – D does not have υμιν at the end of the verse (-4)

8 – D transposes, reading πολυν καρπον instead of καρπον πολυν

8 – D reads γένησθαι instead of γένησθε (+2, -1)

8 – D reads μου instead of εμοι (+1, -2)

9 – no variants

10 – D reads καγω instead of εγω (+2, -1)

            Codex Bezae thus has 18 non-original letters in John 15:1-10, and is missing 124 original letters, for a total of 142 letters’ worth of corruption.  When all movable-nu variants and normal vowel-exchanges are removed from the equation, Codex D has 7 non-original letters in this passage, and is missing 113 letters, for a total of 120 letters’ worth of corruption. 

            So, setting the results side by side:  if we didn’t filter the variants at all, 2397 would have a total of 25 letters’ worth of corruption, compared to 142 letters’ worth of corruption in the text written by the copyist of Codex Bezae.  If normal vowel-exchanges, transpositions, and minor spelling differences in both manuscripts are set aside, then 2397 detours from the Nestle-Aland compilation’s text of John 15:1-10 at just two points, yielding four letters’ worth of corruption.  Using the same approach, Codex Bezae has 120 letters’ worth of corruption in these ten verses.  

            Now, let’s consider the reliability of the transmission-streams of these two manuscripts.  Minuscule 2397 was produced in 1303, according to its colophon.  That means that, reckoning that the Gospel of John was written in A.D. 90, 1,213 years separate the autograph from this copy.  Using raw data, the copyists in the transmission-line of the Hyacinthus Gospels contributed one letter of corruption (either adding a non-original letter, or removing an original letter) per 10 verses once every 48.5 years.  Using filtered data (i.e., without considering benign, normal vowel-exchanges, movable-nu, and similar variants), the copyists in the transmission-line of the Hyacinthus Gospels contributed one letter of corruption (either adding a non-original letter, or removing an original letter) per 10 verses once every 303 years.

            Meanwhile, if we assign a production-date of A.D. 500 to Codex Bezae, this would mean that 410 years separate the autograph from this copy.  Using raw data, the copyists in the transmission-line of Codex Bezae contributed one letter of corruption (either adding a non-original letter, or removing an original letter) per 10 verses about once every three years.  Using filtered data, the copyists in the transmission-line of Codex Bezae contributed one letter of corruption (either adding a non-original letter, or removing an original letter) per 10 verses once every 3.4 years. 

            These results show that no matter how one filters the data, the text of John 15:1-10 in Codex Hyacinthus is over six times more accurate than the text that came from the hand of the copyist of Codex Bezae.  These results (based, admittedly, on one sample passage) also indicate that the copyists in the transmission-stream of the Hyacinthus Gospels – a Byzantine transmission-stream – were about twenty times better at copying the Greek text than the copyists in Codex Bezae’s Western transmission-stream.  I note in closing that if the Robinson-Pierpont Byzantine Text had been used as the standard of comparison instead of the Nestle-Aland compilation, the results would favor the Hyacinthus Gospels even more heavily.

[Readers are welcome to double-check the data and calculations in this post.]