Paul has just appealed to ordinary legal practice: once a διαθήκη has been ratified, nobody voids it or adds codicils to it (v. 15). Verse 16 applies that to the Abrahamic promise, and does so by an argument from a single word’s grammatical number. The promises were spoken to Abraham and to his seed — and Scripture wrote seed, singular, not seeds, plural. Therefore, Paul says, the promise was addressed to one recipient, and that one is Christ.

Three things carry the verse, and all three have been contested since antiquity.

1. The divine passive ἐρρέθησαν. No speaker is named. Paul has God in view (he says so at v. 18, κεχάρισται ὁ θεός), but the construction withholds the subject, and the same reticence governs οὐ λέγει in the second clause — whose subject may be Scripture (so 3:8, 22) or God.
2. The grammatical argument on σπέρμα. Hebrew זֶרַע and Greek σπέρμα are both collective singulars: they denote a line of descendants by the singular form, and a plural (σπέρματα, ‘kinds of seed’) would not have been available to say ‘descendants’ at all. Paul’s argument has therefore struck many readers as pressing a form that could not have been otherwise. That difficulty is real and is not to be translated away.
3. The relative ὅς. Masculine, against the neuter σπέρματι it answers. This is regular constructio ad sensum — the pronoun agrees with the person meant rather than the word’s gender — and it is the clearest signal in the verse that Paul’s ἑνός is a person and not merely a numerical unit.

The rendering keeps ‘seed’ rather than ‘offspring’ or ‘descendant’, because an argument conducted on the number of a noun is unreadable in English if the noun itself is replaced by a word with a different number-behaviour.

Base sense is sown seed (Matt 13:24-32; 1 Cor 15:38), extended in both Hebrew and Greek to posterity. In the LXX σπέρμα is the fixed equivalent of זֶרַע, and like the Hebrew it is a collective singular: ‘your seed’ means the line of descendants, construed with singular or plural verbs indifferently (Gen 12:7; 13:15-16; 15:5, 13; 17:7-8; 22:17-18).

The plural σπέρματα does occur in Greek, but of kinds of seed or of discrete seed-grains (1 Cor 15:38 does not use it; cf. 4 Macc 18:1 σπέρματα for descendants is exceptional). This is the nerve of the standard objection: Paul contrasts σπέρματι with a form that was not an available way of saying ‘descendants’ in the first place.

Translation policy follows from this. ‘Offspring’ is sometimes offered as a modern equivalent collective, and it is lexically defensible; but it makes v. 16 read oddly (‘not offsprings’) and, more seriously, it obscures that a concrete noun is being handled as a word. ‘Descendant’ is excluded outright: it is countable and individual, and it imports Paul’s conclusion into the premise. The agricultural root is also live across the letter and the corpus (σπείρω / θερίζω at Gal 6:7-8 is already fixed as live metaphor), so ‘seed’ keeps the family visible.

Proposed — pending reviewJ. B. Lightfoot, Saint Paul's Epistle to the Galatians, London: Macmillan, 1865

Paul’s reasoning is not that σπέρμα cannot mean a plurality. It is that the promise-text, as written, reads as addressed to one, and that the one who answers to it is Christ. Several distinct accounts of what he is doing are current, and they are not equally vulnerable to the grammatical objection.

The translation must be compatible with all three, which means: do not supply ‘a single descendant’, do not supply ‘the offspring collectively’, and do not soften ὡς ἐπὶ πολλῶν into ‘as though many were in view in a general sense’.

Proposed — pending reviewF. F. Bruce, The Epistle to the Galatians: A Commentary on the Greek Text, New International Greek Testament Commentary, Grand Rapids: Eerdmans, 1982 · Richard B. Hays, Echoes of Scripture in the Letters of Paul, New Haven: Yale University Press, 1989

Aorist passive of the defective λέγω / ἐρῶ (stem ῥηθ‑), the standard passive of saying. Three features are load-bearing.

First, the passive stays passive and the agent stays unexpressed. God is plainly in view from v. 18, but Paul did not write ὁ θεὸς εἶπεν, and ‘God made the promises to Abraham’ supplies a subject the construction withholds. The same reticence runs through οὐ λέγει two clauses later.

Second, the verb is one of speaking, not of giving or making. The promises were spoken, which is what makes the wording of the utterance available for examination. Rendering ‘the promises were made’ quietly relocates the argument from the words to the transaction.

Third, the aorist is complexive — the whole series of utterances taken as one completed speaking — and renders with a plain English past. The Koine spelling with double rho (ἐρρέθησαν) against ἐρρήθησαν is orthographic and carries no weight.

ἐπαγγελία is already fixed for this corpus as ‘promise’ — a formally binding undertaking rather than a prediction, which is what allows Paul to argue it on the analogy of a ratified διαθήκη (vv. 15, 17).

What is marked here is the plural. Galatians has the singular at 3:14, 17, 18, 22, 29 and 4:23, 28, and the plural only here and at Rom 9:4; 15:8; 2 Cor 1:20. The plural reflects the repeated promise-sayings of Genesis (12:7; 13:15; 15:18; 17:8; 22:18; 24:7) rather than a plurality of distinct promises, and it must be retained: a singular English ‘the promise’ levels a deliberate form and loses the fact that Paul is pointing at a body of texts whose wording can be inspected.

The article is anaphoric-definite: the promises, those already under discussion since v. 14.

A citation-formula with no expressed subject, and the candidates are exegetically consequential.

Policy: the literal tier keeps the bare ‘it does not say’; reading tiers may retain ‘it does not say’ as the least resolving option. ‘Scripture does not say’ and ‘God does not say’ both supply a subject the critical text withholds, and the move to supply one is exactly what the later tradition did. Note also that λέγει is present against the aorist ἐρρέθησαν: the text goes on saying what it says, which is what makes present inspection of its wording possible.

ὡς here is not comparative but construal-marking: it introduces how a cited form is to be taken — ‘as [speaking] of many’. The verb is elided in both limbs and nothing is supplied unbracketed; the elliptical compression is the rhetoric, and the versional tradition’s filling of the gap (Latin quasi in multis … quasi in uno) preserves the form better than most English renderings.

ἐπί + genitive after a verb of speaking marks reference: speaking with regard to a party (cf. the reference-sense of ἐπί with verbs of utterance in the papyri and in Hellenistic grammatical discussion of how a term is used ἐπί something). ‘About’ and ‘of’ both carry it; ‘in’ (so the Latin) is too locative for modern English and ‘referring to’ over-formalizes.

The gender of πολλῶν and ἑνός is formally open (genitive plural and singular are identical for masculine and neuter). If neuter, the phrase is about the word’s reference to many things or one thing; if masculine, about many persons or one person. The masculine ὅς at the end of the verse pulls strongly toward the personal reading, and that is adopted — but the ambiguity is a feature of the Greek and the English ‘of many … of one’ reproduces it without deciding.

ὅς is masculine; its antecedent σπέρματι is neuter. This is constructio ad sensum, where the relative takes the gender of the person meant rather than of the grammatical antecedent — the same phenomenon already noted in this corpus at Gal 4:19 (τέκνα … οὕς) and Phlm 10.

The effect in English cannot be reproduced, since English relatives carry no gender, and no compensating adjustment is made in the rendering. But the construction should be recorded, because it is the verse’s internal evidence that Paul’s ἑνός is a person: a bare numerical point about a noun’s reference would have taken the neuter ὅ, which is in fact what a handful of later witnesses supply.

Χριστός is anarthrous as predicate nominative, which is regular and carries no indefiniteness. Per standing policy, Χριστός is rendered ‘Christ’ and the titular force (the Anointed One, and hence the royal-messianic descendant of the σπέρμα promises) is recorded in note rather than built into the rendering.

Transitional, not adversative. Verse 15 laid down the human analogy; v. 16 turns to the case it is to illuminate. Rendering ‘but’ manufactures a contrast between the analogy and its application, which is the opposite of the argument’s shape. Per §10 the particle is rendered, never dropped: ‘now’ carries the move to a new stage of one argument. Compare the standing policy on continuative δέ (Jas 1:15) and on the discourse-formula Λέγω δέ (Gal 4:1), which recurs four words into v. 17 (τοῦτο δὲ λέγω).

Paul’s quoted words καὶ τῷ σπέρματί σου reproduce LXX wording found at Gen 13:15, 17:8, and 24:7 (τῷ σπέρματί σου), with καί appearing in the collocation σοὶ … καὶ τῷ σπέρματί σου at Gen 13:15 and 17:8. No single verse is identified, and the plural αἱ ἐπαγγελίαι in the first half of the verse indicates that Paul is pointing at the promise-texts as a group rather than citing one.

The Hebrew behind all of these is זַרְעֲךָ, likewise a collective singular, so the LXX translators had no option to pluralize either. This matters for assessing Paul’s argument: the singular is not a choice made in either language at the level of morphology, which is why accounts (b) and (c) above have been developed.

Proposed — pending reviewHans Dieter Betz, Galatians: A Commentary on Paul's Letter to the Churches in Galatia, Hermeneia, Philadelphia: Fortress, 1979

The critical text is stable in substance. Two points are worth recording.

ἐρρέθησαν. Spelled with double rho in NA28; the single-rho ἐρρήθησαν / ἐρέθησαν appears in parts of the tradition. Purely orthographic.

ὅς ἐστιν Χριστός. A small number of later witnesses read the neuter ὅ instead of masculine ὅς, regularizing the relative to the gender of σπέρματι. This is the easier reading — it repairs an irregular concord — and is therefore secondary; the direction of scribal activity confirms the masculine as original and shows that ancient copyists felt the same roughness modern readers do.

οὐ λέγει. Some witnesses expand with a subject (λέγει ἡ γραφή / ὁ θεός). This is the ancient instance of exactly the move the rendering is forbidden to make, and it is evidence that the subjectless text is original.

The verse was discussed early, and the discussion already divided along lines still live.

Chrysostom, commenting on Galatians, takes the singular as pointing to Christ and reads the promise as fulfilled in him and in those who are his, connecting v. 16 forward to v. 29 rather than treating the two as in tension.

Jerome, in his commentary on Galatians, registers the grammatical difficulty frankly — that Hebrew zeraʿ has no ordinary plural for descendants — and defends Paul by appealing to the sense rather than the form.

Augustine reads the one seed representatively: Christ is the seed, and the church is in him, which is substantially account (b) above.

These are recorded as reception, not as lexical evidence. That the Fathers felt the need to explain the argument is itself a datum about the verse, and it should not be smoothed over in translation.

Proposed — pending reviewJohn Chrysostom, Commentary on the Epistle to the Galatians, in Patrologia Graeca 61 · Jerome, Commentarii in epistulam Pauli ad Galatas, in Patrologia Latina 26

• ‘seed’, not ‘offspring’ or ‘descendant(s)’. An argument on the number of a noun requires that the noun survive into English with its number intact. ‘Offspring’ is permitted in a declared modern register with the compression recorded; ‘descendant’ is excluded.
• ‘were spoken’, not ‘were made’. The verb is one of speaking, and the whole argument depends on the wording of the utterance.
• no supplied subject for οὐ λέγει, and no supplied agent for ἐρρέθησαν.
• ‘as of many … as of one’, with no verb supplied unbracketed and no gender specified.
• plural ‘promises’ retained against the singulars elsewhere in the chapter.
• the relative left as ‘who’, which in English neither reproduces nor conceals the Greek concord.