“Overall this is a very difficult problem.” Thus wrote Wieland Willker, in his Textual Commentary on the Greek Gospels, about the variation-units in Matthew 21:29-31, to which he devoted eight pages of analysis before concluding that “a fully convincing solution is currently not available.”
“Verses 29-31 involve a rather complex and difficult textual problem.” Thus begins the NET’s note on the same subject – one of the longest notes in the NET.
But you wouldn’t know that there is a very tough textual variant here from most of our English Bibles. The CSB, ESV, NASB, NIV, NLT, EHV, MEV, NRSV, and NKJV all say that the answer to Jesus’ question was “The first,” and they have no footnote here. (The EOB-NT is a rare exception; the EOB-NT has “The first” in the text and a footnote says, “A few manuscripts, notably D, read “the second” which is unlikely but presents the Jews as spoiling the parable by giving (seemingly deliberately) the wrong answer.” The EOB’s note could be improved, though, by replacing “the second” with “the last.”)
I think one has to go back to J.B. Phillips’ version to find, in English, anything other than “the first” (or, in the CEV, “the older one”) as the answer that was given to Jesus’ question in Matthew 21:31. (In Phillips’ version, the answer is given as “The second one,” echoing the variant δευτερω. A different reading, ὁ ὕστερος, was in the text of the 25th edition of the Nestle-Aland compilation, and was also adopted by Westcott & Hort and Tregelles. (The 1881 Revised Version reads “The first.” Perhaps the Revision Committee was unimpressed with Westcott and Hort’s divided opinion.) (The Tyndale House GNT deviates from Tregelles’ compilation, adopting the usual reading ὁ πρῶτος, with an exceptionally thorough apparatus-entry.)
Hort, in 1881, devoted over two full pages (in Notes on Select Readings) to Mt. 21:28-31, and mentioned the view of Lachmann that the Jews’ answer to Jesus in Mt. 21:31 is “an early interpolation” (along with the four words which follow it). Westcott inserted a note of his own into Hort’s analysis, stating, “Considering the difficulty of the Western combination of readings it seems not unlikely that Lachmann is substantially right.”
The array of readings in this passage is interesting: first comes a contest in verse 29: οὐ θέλω, ὔστερον δέ μεταμεληθεὶς ἀπῆλθεν is read by the Byzantine Text, and by C L M W Π 157 565 579; À* has almost the same reading but without the δέ). Codex Vaticanus has, instead, ἐγώ κε καὶ οὐκ ἀπῆλθεν (“I go, lord, and did not go”). Third, f13 and 700 (Hoskier’s 604) read ὕπαγω κύριε καὶ οὐκ ἀπῆλθεν. Fourth, Θ (038) has ὕπαγω καὶ οὐκ ἀπῆλθεν. Fifth, Codex Bezae (05) reads οὐ θέλω ὔστερον δέ μεταμεληθεὶς ἀπῆλθεν but adds εἰς τὸν ἀμπελωνα (repeating the words from v. 28).
There are other textual contests (Byzantine witnesses have και before προσελθὼν; Alexandrian witnesses tend to have προσελθὼν δε, and the scribe of À skipped ὁ δὲ ἀποκριθεὶς ειπεν in v. 30), but the main contest near the beginning of v. 30 is between ἑτέρω (another) and δευτέρω (second).
Δευτέρω (second) is supported by B L M S Ω and by most Greek manuscripts, and by 28 33 700 892 and 1424. The Byzantine Textform is somewhat split here, though: δευτέρω is in the text and ἑτέρω is in the margin.
Ἡτέρω (the other) is supported by À*, D, Q, K, Π, W, Y, Δ, 157, 565, 579, et al. In Codex À, someone creatively changed ἑτέρω by putting δ in the left margin and υ between the first ε and τ, thus producing the variant δευτέρω.
Those two textual contests must be kept in mind as we approach the main textual contest in verse 31, where we see (after another variation-unit in v. 31: after λεγουσιν, most MSS have αυτω but not B À L D Θ 33 f13 788) the answer that Jesus’ listeners gave:
Codex B and some Ethiopic copies support ὁ ὕστερος (“the later [one]”).
Almost all manuscripts support ὁ πρῶτος (“the first”).
Codex D and Θ 700 f13 and several Old Latin copies (a, b, d, e, ff2, h, l) and the Sinaitic Syriac and the Armenian version support ὁ ἔσχατος (“the last”).
Let’s remember one of the canons of textual criticism: prefer the more difficult reading. The reading of Codex Bezae is the more difficult reading here – but it also makes Jesus’ hearers appear idiotic; it is obvious that the son who told his father that he would go, but did not go, did not do what the father wanted him to do.
Jerome’s Vulgate supports ὁ πρῶτος. But in his Commentary on Matthew, Jerome said something about the variant in verse 31: “one should know that with respect to what follows: ‘Which of the two did the father’s will? And they said, ‘the last,’ the authentic copies do not have ‘the last’ but ‘the first.’” Jerome proposed that “If we want to read ‘the last,’ the interpretation is plain. We would say that the Jews indeed understood the truth, but they are evasive and do not want to say what they think.” In other words, to Jerome, the reading ὁ ἔσχατος (“the last”) makes the Jews seem not stupid, but duplicitous.
Bruce Metzger, in his Textual Commentary, rejected the very difficult reading of Codex Bezae, stating that “it is not only difficult, it is nonsensical” – and explained that the UBS committee judged that D’s reading originated “due to copyists who either committed a transcriptional blunder or who were motivated by anti-Pharisaic bias.” Whatever the mechanism was, the range of its effect must have extended not only to Codex D but to several Latin copies, to the Sinaitic Syriac, and to Jerome’s “authentic” Latin manuscripts – most of which are major representatives of the Western Text.
[But not the Curetonian Syriac. As Willker notes, the Curetonian Syriac was erroneously cited in NA27 as if it supports ὁ ἔσχατος. Willker supplies Peter Williams’ rendering of the Curetonian Syriac, which concludes with “The first/former”.]
[Another oddity in the apparatus of NA27 is that Θ is assigned two readings in v. 31: ἔσχατος and ὕστερος.]
[A mistake might have been made by Kurt and Barbara Aland in their Text of the New Testament where they took a close look at this passage (beginning on p. 233), and they zoom in on Mt. 21:31 (beginning on page 235). The Alands stated that ὁ ὕστερος is supported by Codex Vaticanus “and other Greek manuscripts as well as in some Sahidic manuscripts and the whole Bohairic tradition.” What are the “other Greek manuscripts” here? The apparatus of NA27 lists B Θ f13 700 al; however, this seems to refer to the reading ὕστερον near the end of v. 30; Swanson gives ὁ ἔσχατος as the reading of Θ f13 700 in v. 31 before λέγει. If there are any other Greek manuscripts that read ὁ ὕστερος other than Codex Vaticanus, I do not know what they are. If anyone knows of any, please mention them in the comments.]
Now let’s apply another canon: prefer the variant which accounts for its rivals better than they account for it. The reading in Mt. 21:31 adopted in the UBS compilation – ὁ πρῶτος – does not explain ὁ ὕστερος. And ὁ ὕστερος can account for ὁ πρῶτος and ὁ ἔσχατος.
As as answer to Jesus’ question, ὁ ὕστερος is a somewhat fluid answer; to someone whose first language was Latin, ὁ ὕστερος might be misunderstood as if it means “the latter.” And such a misunderstanding explains the origin of the Western reading; there is no need to suppose that an “anti-Pharisaic bias” was involved here.
Ὁ ὕστερος also explains ὁ πρῶτος: The answer “The later one” refers to the first son, not initially (when he said that he would not go), but later (after he changed his mind). An early scribe who perceived that ὁ ὕστερος could be misunderstood as a reference to the second son could easily avoid the misunderstanding that mars the Western Text by making the wording clearer, and he did so, creating the reading ὁ πρῶτος.
This internal evidence – (1) ὁ ὕστερος is difficult but not nonsensical, and (2) ὁ ὕστερος accounts for ὁ πρῶτος better than ὁ πρῶτος accounts for ὁ ὕστερος, and (3) ὁ ὕστερος accounts for ὁ ἔσχατος – compels the adoption of ὁ ὕστερος.
By the way, I adopted ὁ ὕστερος over a decade ago in my Equitable Eclectic English Edition of the Gospel of Matthew, rendered as “The later one.” EEEE Matthew, as I call it, is available as a Kindle e-book on Amazon for $1.99.