Paul has just recited the bare facts of Gen 16 and 21 (v. 22-23). Now he declares what he is doing with them. Four things carry the verse:

1. ἀλληγορούμενα — a NT hapax, and the single most argued-over word in the chapter. Paul says the narrative is being allegorized, not that it is an allegory; the present passive leaves open whether the figural sense inheres in the text or is being applied to it. Every English rendering leans one way.
2. αὗται is feminine, not neuter. Paul does not say the things are two covenants; he says these women are. The equation is between persons and covenants, which is what makes the whole argument work.
3. μία μέν opens a correlation that never receives its matching δέ in proper form — v. 25b’s δέ and v. 26’s ἡ δὲ ἄνω Ἰερουσαλήμ both compete for the slot. The sentence breaks under its own weight, and that is data about the heat of the argument.
4. εἰς δουλείαν γεννῶσα — the Sinai covenant is depicted as a mother bearing children into slave status. Paul has taken the legal status of Hagar’s son (Gen 21:10, quoted at v. 30) and made it the offspring of a covenant.

NT hapax, absent from the LXX. Built from ἄλλο + ἀγορεύω, ‘to say something other [than what is said],’ the standard term in Hellenistic grammatical and philosophical criticism for reading a text so that its surface narrative yields a further sense. Heraclitus the Allegorist defines it: ‘saying one thing and meaning something other than what one says.’ Philo, Paul’s near-contemporary Alexandrian, uses the cognate vocabulary constantly of exactly this Genesis material (De congressu 11-24 on Hagar and Sarah as encyclopedic learning and virtue).

Two grammatical points constrain the rendering:

• The form is a present passive participle in periphrasis with ἐστιν, not a noun. Paul does not write ἀλληγορία ἐστίν (‘it is an allegory’). He writes that these things ‘are being allegorized.’
• The passive leaves the agent unexpressed. Whether the agent is Paul, Scripture itself, or God is not stated, and English should not state it.

Chrysostom (Hom. in Gal. 4) notices the imprecision on Paul’s own terms and says Paul called it allegory καταχρηστικῶς — using the word loosely — meaning τύπος. That is already a patristic reader defending the historical referent. Augustine, by contrast, took the word at face value and built on it.

Built onbdag · bdf · muraoka-lxx

Neuter plural of the qualitative relative ὅστις, per standing glossary policy rendered ‘whatever things (of such a kind).’ The antecedent is not any single noun in v. 22-23 but the whole preceding narrative — the two sons, the two mothers, the two modes of begetting. The qualitative force is doing work: not merely ‘these things’ but ‘things of such a kind as to be allegorized.’ The literal tier should show the construction; smooth English can carry it with a plain demonstrative.

Singular ἐστιν with a neuter plural subject is ordinary Greek concord (BDF §133) and carries no emphasis.

Built onbdf · bdag

The gender shift is the hinge of the verse and is easily lost in English, which does not mark it. ἅτινα is neuter (the narrative facts); αὗται is feminine. The feminine plural cannot refer back to ἅτινα. Its antecedents are ἡ παιδίσκη and ἡ ἐλευθέρα of vv. 22-23 — the two women.

So the predication is: these two women are two covenants. Not ‘these things represent two covenants,’ which is what most smooth English produces by default. Paul’s equative ἐστιν is the same figural ‘is’ he uses at v. 25 (τὸ Ἁγὰρ Σινᾶ ὄρος ἐστίν) and at 1 Cor 10:4 (ἡ πέτρα δὲ ἦν ὁ Χριστός). Retaining ‘these women’ in the reading text is the minimum needed to keep the antecedent visible.

διαθῆκαι is anarthrous and δύο is predicative: ‘are two covenants,’ not ‘are the two covenants.’ Paul is not cataloguing the whole of covenant history; he is naming two that this narrative yields.

Built onbdf · na28

The LXX’s near-invariant equivalent for בְּרִית (some 270 times), chosen over συνθήκη precisely because συνθήκη denotes a negotiated agreement between equals and בְּרִית does not. διαθήκη in Hellenistic Greek is overwhelmingly a testamentary disposition — a will, unilaterally settled by the disposer, which is why the documentary papyri are full of it. Paul exploits that legal sense explicitly at Gal 3:15-17, where a διαθήκη once ratified cannot be annulled or supplemented.

That context governs here. ‘Covenant’ remains the concordance gloss because the LXX substrate is decisive and because Paul is speaking of the Sinai and Abrahamic dispositions, but the unilateral, disposer-imposed character is part of the word and not an import. Do not render ‘contract’ or ‘agreement.’

Note what Paul does not say: he does not call the Sinai covenant bad. He says it bears children into a slave status — a statement about the standing of those under it, consistent with 3:23-25 and 4:1-3, not a verdict on the covenant’s origin or giver (cf. Rom 7:12; 9:4).

Built onbdag · muraoka-lxx

μέν raises the expectation of a matching δέ, and per §10 the correlation must be marked in English rather than dropped. But the answering member is genuinely hard to locate:

• v. 25a τὸ δὲ Ἁγὰρ Σινᾶ ὄρος ἐστίν — δέ present, but the clause is a parenthetical geographical gloss, not the second covenant.
• v. 26 ἡ δὲ ἄνω Ἰερουσαλήμ — the true logical counterpart (free, over against slavery), but grammatically it answers ‘the present Jerusalem’ of v. 25, not ‘one covenant’ of v. 24.

The result is that the second διαθήκη is never named. Paul’s sentence is broken; the argument arrives anyway. This is not a textual problem to be repaired but a feature of dictated polemical prose (compare the collapse at 2:6). English should mark μέν (‘the one …’) and let the reader feel the missing member rather than manufacturing it.

Built onbdf · na28

ἀπό + genitive of source: the covenant originating from Sinai. Anarthrous ὄρους Σινᾶ is normal after a preposition (BDF §255) and English supplies the article. Σινᾶ is an indeclinable Semitic transliteration, as throughout LXX Exodus.

The mountain, not the law, is named — a place, and a place outside the land. That geographical point is what v. 25 then presses: Sinai is ἐν τῇ Ἀραβίᾳ, in Hagar’s own territory (Gen 25:12-18 locates the Ishmaelites there), and therefore ‘lines up in the same column with’ (συστοιχεῖ) the present Jerusalem. The Judaizing opponents’ authority is thus placed, geographically and genealogically, outside the promise.

Built onbdag · na28 · muraoka-lxx

The participle is feminine nominative singular, agreeing with μία (διαθήκη), and continues the maternal image: the covenant is a mother. γεννάω with a female subject is ‘bear, give birth to’; the object (τέκνα) is unexpressed and must be supplied in brackets.

εἰς δουλείαν is result/goal, not sphere: the children are born into a slave status, not merely in slavery. This is legal status language, and it is generated from the narrative itself — Gen 21:10 LXX, quoted at Gal 4:30, calls Ishmael ὁ υἱὸς τῆς παιδίσκης, and it is precisely his mother’s servile status that disqualifies him from inheritance. Paul’s word for it, δουλεία, ties back to δεδουλωμένοι in v. 3 and δουλεύειν in vv. 8-9: the same servitude the Galatians were rescued from is what they are re-entering.

The present tense is characteristic and, coupled with δουλεύει in v. 25, ongoing — the covenant goes on bearing such children, which is why the situation in Galatia is urgent and not merely historically instructive.

Built onbdag · bdf · muraoka-lxx

The abstract noun to δοῦλος, already fixed in the glossary as ‘slave’ rather than ‘servant.’ δουλεία is the condition of being owned — legal non-freedom, the exact opposite of ἐλευθερία, which is the term Paul sets against it here (ἡ ἐλευθέρα, vv. 22, 23, 26, 30, 31; cf. 5:1).

Per §5 and §9 the ancient institution is not softened. Paul is arguing inside a legal world where a slave woman’s son had no claim on the estate, and the force of his argument depends on the reader knowing that. ‘Bondage’ is archaic; ‘servitude’ blunts the ownership; ‘slavery’ is the word.

Built onbdag · muraoka-lxx

Second qualitative relative in the verse. Its gender is feminine singular, so it agrees with μία (διαθήκη) — the covenant is Hagar — rather than reaching back to αὗται. The direction of the equation therefore reverses within the sentence: v. 24a said the women are covenants; v. 24c says the covenant is Hagar. Figural identification runs both ways, which is characteristic of the mode and part of why the ‘allegory or typology’ question resists resolution.

The qualitative force (‘which is of such a kind as to be Hagar’) is thin here and reduces in practice to a defining relative; keep ‘which is’ in the reading text and note the class-describing sense.

Built onbdf · bdag

The verse is stable in the critical text. Three points worth recording:

• ἅτινα is read universally; the smoother ἃ appears in a few late minuscules and is plainly a simplification of the qualitative pronoun.
• Σινᾶ / Σινά is an accentual variation only, with no exegetical weight.
• The genuine variation lies immediately downstream in v. 25, where witnesses divide over τὸ δὲ Ἁγὰρ Σινᾶ ὄρος ἐστίν (with Ἁγάρ) against τὸ δὲ Σινᾶ ὄρος ἐστίν (without). The variant matters for how v. 24’s ἥτις ἐστὶν Ἁγάρ is heard: if v. 25 lacks Ἁγάρ, v. 24 carries the whole identification. Since the decision is made at v. 25 and not here, it is recorded rather than argued.

Metzger’s discussion of the v. 25 problem is the standard entry point; the committee’s judgment there does not disturb anything in v. 24 itself.

Built onna28 · metzger-tc

The Fathers divide, and the division is instructive rather than embarrassing.

Chrysostom (Hom. in Gal. 4) insists Paul used ἀλληγορία imprecisely for τύπος and stresses that the history is not dissolved: Hagar was a real woman, Ishmael a real son, and the figure rests on what actually happened. His concern is plainly polemical — to keep Antiochene historical exegesis from being outflanked by Alexandrian method using Paul’s own vocabulary.

Jerome (Comm. in Gal.) renders the participle per allegoriam dicta and takes it at face value, noting Paul’s own rabbinic training as the source of the procedure rather than Greek grammatical theory.

Augustine treats the verse as apostolic warrant for figural reading generally, and on his account Paul’s practice licenses the Church’s.

The common ground is worth naming: none of them reads Paul as saying the Sinai covenant is evil, and all of them read δουλεία as a statement about the standing of those under law before Christ, not about the law’s giver.

Built onobrien-nigtc · silva-becnt

The Vulgate’s quae sunt per allegoriam dicta is a periphrasis that keeps the passive and keeps the verb verbal — closer to the Greek than English ‘which is an allegory,’ and a useful check on the temptation to nominalize. Jerome’s ancilla for παιδίσκη in the surrounding verses preserves the slave status precisely.

The Peshitta renders with a form of the root ܦܠܐ/‌ܐܪܙ vocabulary of hidden or figured speech, again verbal rather than nominal, and likewise leaves the agent unexpressed.

Both ancient versions thus resist the classificatory noun that modern English reaches for first. That convergence is a reason to keep the verb verbal in the reading text.

Built onmuraoka-lxx · na28