The legates of Syria, the coins, and what Luke 2:2 does and does not date

Luke 2:2 gives one Roman name, Quirinius, and attaches a census to it. Luke 1:5 dates the announcement to Zechariah under Herod, more than a year before the birth. Between those two data lie the fasti of Syria — the sequence of imperial legates of the province — and the civic coins of Antioch, which anchor some of those names year by year. Two questions, each with its own answer: do the legates permit a birth under Herod, and what does Luke 2:2, read straight, date — the birth or the census? The first answer comes out favorably for the window; the second does not, and that holds for the consensus and for HV alike.

5.1 The fasti in the window: Varus, coin-anchored

Dąbrowa's fasti give these headings around the turn of the era: Sentius Saturninus c. 10–8/7 BC; Varus 7–c. 4 BC; Piso c. 4–1 BC; Volusius Saturninus AD 4–6; Quirinius AD 6/7; Metellus Silanus ?AD 11–17.1 Between 1 BC and AD 4 there is no entry; Gaius Caesar stands in an appendix, outside the fasti. The window 7–4 BC thus falls under two names — Saturninus at the front edge, Varus over the rest — and neither is Quirinius.

For Varus this is more than prosopography. The coins of Antioch carry his name with the formula ἐπὶ Οὐάρου in three consecutive era years: year 25 = 7/6 BC, year 26 = 6/5 and year 27 = 5/4.2 The next legate the coins name is Volusius Saturninus, in year 35 = AD 4/5.3 The year 5 BC, the most robust cell of the window, falls fully within Varus's coin years; the earlier cells 7 and 6 BC fall under his fasti heading and his first coins.

Between year 27 and year 35 the coins are silent about the legate: seven years, 4/3 BC to AD 3/4. Nothing can be read from that silence, for or against any reconstruction. The ἐπί formula is no peculiarity of Varus — after him Antioch names another eight legates in this way4 — and at the base rate 48% of the 21 imperial legates of Syria left no coin; Piso, who according to the fasti sits in the silent slot, carries a dash. Whoever reads a term out of the silence, Varus's or anyone else's, reads noise as signal.

5.2 The Gaius constraint

One scene in Josephus binds the legates to Herod's year of death. There, at the hearing on Herod's will, Augustus reads out the letters Varus has sent from Syria, and seats Gaius Caesar first on the consilium.5 Dio has Gaius as consul in Syria in AD 1.6 Under a death in 4 BC that fits without stretching: Gaius is 16 or 17 and in Rome, Varus's coinage runs to 5/4. Under a death in January 1 BC, Varus must still be legate three to four years past his last coin and right through Piso's slot, and Gaius must still be sitting in Rome while Dio has him in the East.

The constraint prices; it does not close. Gaius's departure date is not measured; an imperium in the East is not the end of a legateship; the fasti reconstructions differ precisely in 4–1 BC; and the coin silence falls within the 48% noise. A price, then, not an exclusion — but a price SY cannot pay in its own coin: its buffer elsewhere is the antedating of the regnal years of Herod's sons, and that cannot reach a Roman whose travels are dated by Romans.

5.3 No Quirinius before AD 6, and what ILS 918 does not say

The harmonizing reading — Quirinius governed Syria once before, or he took the census while Varus governed — must find a place in the fasti, and there is none. Quirinius appears there once, AD 6/7. Josephus calls Varus in the window "ὁ τῆς Συρίας ἡγεμών", and in him Quirinius appears only in AD 6/7, as someone arriving then for the first time. "Quirinius took the census while Varus governed" has no bearer at all: no text, no stone, no coin.

The one stone adduced for an earlier term, ILS 918 from Tivoli, is headless; the name of the honorand is missing. The text says that as legate of Augustus he obtained iterum Syria and Phoenicia. Whether that means "Syria twice" or "legate for the second time, this time of Syria" hangs on the position of iterum — and that predicts nothing: in the comparison series the province name stands 11 times before and 10 times after the word. The editor's brackets, moreover, stand precisely around the disputed word. Sherwin-White, who first writes that the construction "always" means one thing, retracts that four lines further on with "far from stabilized".7 His own note takes Quirinius off the stone: the honorand subdued a regem, and no king is known of the Homonadenses whom Quirinius fought; of the proconsulate of Asia that the stone records, nothing is known for Quirinius. His verdict reads "improbable", not "impossible": "a headless inscription [...] may change sides at any moment". The majority takes the twice-legate to be Piso; Alföldy holds out for Quirinius, Taylor for M. Titius. Whom the stone honors does not decide this axis; that the stone does not put Quirinius in Syria before AD 6 does.

The datum that could overturn this reading lies dormant. The only stone that confirms Quirinius's Syrian census, ILS 2683 (Aemilius Secundus), does not date itself: checked against the epigraphic source it carries no dating at all — the stone does not date Quirinius, Quirinius dates the stone.8 A dated piece from era year 28–34 naming Quirinius as legate of Syria, or a refutation of Sherwin-White's regem and Asia objections, would break the fasti reading. The absence of such a piece does not refute it, since the silence is generic; as long as that stone is undated, the axis stands as the coins set it.

5.4 The account on the legates axis

The cheapest reading is the measured fasti reading itself: Varus 7–c. 4 BC, coin-anchored; Piso c. 4–1; Quirinius only AD 6/7; no Quirinius legation before AD 6; the iterum legate of ILS 918 is not Quirinius. It permits the window under Herod. What it does not deliver is the reconciliation with Luke 2:2 — that shifts to the grammar of the sentence. The four chains pay unequally.

C fits fasti, coins and Gaius constraint without a single extra assumption. Its price is one constraint it drops: Luke 2:2 as a datum contemporaneous with a birth under Herod. C books that as an error of Luke's, or takes Sherwin-White's third way out — the birth deliberately placed in AD 6 — which costs Matthew 2 and Luke 1:5 at once and is the dearest of the three that Levick lists.9 The price does not hang on a convention undefended elsewhere: the census of AD 6 is independently anchored by Josephus, the 37th Actian year as Dąbrowa counts it, and ILS 2683.

SY must pull Varus through Piso's slot and thereby drops the fasti heading "Piso c. 4–1" and the coin terminus 5/4 alike; it must still put Gaius in Rome at a death in 1 BC, against Dio.10 Several constraints at once, and the antedating pays none of those bills. SY survives the axis at a price, without exclusion.

CA shares the fasti cell with C — Varus 7–4 — and pays nothing there. Its numismatic underpinning does fall: the idea that from year 28 Varus's name gives way to that of Augustus breaks on the coins themselves, since the Augustus series already begins in year 27, concurrently with Varus's last coin.11 Molnar's Aries hinge begins 17 years too late.12 CA inherits the bare C cell; its own props are struck out.

HV has two routes on this axis, unequally priced. Route A, an early Syrian legation of Quirinius, postulates a term that the fasti exclude, that ILS 918 does not deliver and that Josephus contradicts — three constraints at once, the dearest reading on the axis; that it would help HV does not make the measure milder. Route B reads πρώτη as "before": this census preceded the one under Quirinius. It asks for no Quirinius in the fasti and costs nothing here; its account lies on the grammar axis (§5.6). The legates axis thus itself pushes HV from Route A to Route B, closes Route A for everyone, and is otherwise neutral between "Luke was mistaken" and "Luke's πρό": both are compatible with the fasti. To be stricter against the second because it helps HV would here be the mirror asymmetry.

5.5 What Luke 2:2 dates

The sentence says that "this census" (αὕτη ἀπογραφή) was "the first" (πρώτη), with a genitive absolute: "when Quirinius was governing Syria". Ordinal plus genitive absolute is, in Luke and in the Septuagint, a temporal dating, not a comparative; the traditional reading is the grammatically covered one. Quirinius's Syrian governorship is AD 6, and the census under Quirinius is the known provincial census of that year: Acts 5:37 names it with the article, as the census pure and simple. In the window the governor is called Saturninus and then Varus, with just under 50 coins as bearer; no witness at all places a census of Quirinius under Herod. Luke 1:5 puts the birth under Herod, Luke 2:2 ties the birth census to Quirinius: an internal tension of about 10 years.

A documentary datum is added to this. A genuine provincial census document, P.Yadin 16 from AD 127, dates itself by imperial year and consuls, not by the name of the governor, and Luke precisely does not use the two Greek dating conventions such a document would use. His Quirinius is not a documentary dating operator but a narrative placement. That deprives the sentence of its sharpness as a constraint, and the mainstream does not use it as one either. Keener, the most extensive modern defender of Luke's reliability, leaves 2:2 undefended and books it as the "most [...] persuasive" case of error; Braunert reads it as deliberate construction; Smallwood and Schürer-Vermes-Millar (SVM) book it as a Lukan error.13

The core is symmetric. The census is datable — but it dates the ἀπογραφή to Quirinius, AD 6, and not the birth. As a constraint on the birth year it points about 10 years away from the window.

5.6 The three ways out

Whoever still wants to bring Luke 2:2 to the window has three exegetical ways out. All three break, by the same measure HV applies on its own side.

(a) πρώτη as "before Quirinius".14 Luke says "before" with πρό where he means it. Ordinal plus genitive absolute is temporal in three other cases in him and in the Septuagint. The word order carries nothing, and the counterexample with which Fitzmyer wanted to rescue the reading is spoken for. Nicolet rules out that one sentence admits both readings at once. Externally, Braunert rejects the reading on Acts 5:37, where Luke calls the census "des Zensus schlechthin" — a third rejection alongside Nolland and Llewelyn. Keener calls it "not decisive grammatically but nevertheless plausible" and thereby himself undercuts that "plausible": he does not choose it.

(b) An earlier census or an earlier office of Quirinius.15 This way out stands on zero data. The fasti have Saturninus and Varus; Josephus has Varus as "ὁ τῆς Συρίας ἡγεμών". Syme does not rule out a first governorship, but constructs one himself, and his rejection of rival candidates rests on one restored syllable from which his own candidate is exempted.16 Keener himself puts on the brakes here: many find it difficult to date an earlier office of Quirinius in that way. HV has considered this route and closed it on the same grounds.

(c) An error of textual transmission.17 A variant exists at precisely the decisive point, with a name and a history. But the apparatus decides against it: SVM give a different manuscript picture from Fitzmyer's, and NA28 makes Fitzmyer's "some MSS." too little.18 None of the four chains has measured this way out itself.

Keener lists all the ways out in a single footnote and chooses none; the only direction he leaves standing unchallenged is "association with a striking event from roughly the same period" — association, not synchronism. Tertullian read 2:2 comparatively, but the one who writes that out is the author who rejects the reading, and Tertullian contradicts himself over four to six years; the strongest patristic witness this axis had thereby drops away.

5.7 Braunert p. 210 and the impossibility read in

The constraint the consensus deploys against any Herodian census also gets this measure. Sherwin-White writes that a provincial census in Judea at the time of the kingdom is "an impossibility", with reference to Braunert p. 210.19 That does not cover what is there. Braunert poses the question on p. 210 and on p. 211 expressly leaves it undecided ("nicht mehr relevant"); the word "impossibility" is Sherwin-White's. The mirror image: Nicolet's "impossible" is likewise not in SVM, to which he refers.20 Braunert's own denial is prosopographical, not constitutional: his "Damit fällt die Konstruktion" rests on Groag, Syme and Roos — three who remove a necessity, not establish an impossibility — and Sherwin-White himself, three pages earlier, rejects Groag's rule, the first of those three.21

The absolute modality read in thus drops away, and with it a constraint against the consensus. But Luke gains nothing by it. Braunert grants Luke good sources for the modalities of the census and explains the wrong dating as "bewußte Aufrichtung eines Geschichtsbildes" by a "christlich-zelotische Gruppe" — a group of which no text, no name and no witness exists.22 And way out (b) remains after Braunert exactly where it stood: on zero data.

5.8 The account on the census axis

The cheapest surviving reading: Luke 2:2, in its grammatically and textually covered form, dates the ἀπογραφή to Quirinius's governorship of Syria — the census of AD 6 — and not the birth. It is not a constraint on the birth year but a Lukan datum that, read straight, points about 10 years later. Its value for the window is neutral to adverse: at most association with an event from roughly the same period, at its heaviest the strongest single case against Luke's accuracy. The window rests on other axes; the census does not carry it.

C pays the least, because it never leaned on the census: it dates the birth by Herod's death and the priestly courses, drops Luke 2:2 as a dater and books it as a Lukan error. Way out (a) looks cheaper — Luke is then right — but is dearer, since it breaks the grammar and Acts 5:37 alike.

SY survives only if the census can be pulled to 3/2 BC. For that it must bend the grammar (way out a or b), or postulate an unattested census of 3/2 with only Orosius (c. AD 417) as bearer, against RG 8 — lustra in 28 BC, 8 BC and AD 14, no 3/2 — and with RG 35 read as a registration where an acclamation stands. It drops P.Yadin 16. And the construction is circular: the census of 3/2 hangs on the antedating scaffold that the census was supposed to carry.

CA has a census argument of its own (πληρόω, Agrippa I, Syllaeus), but in this dossier it stands only as a label; the elaboration has not been seen and carries nothing here. As far as measurable, CA too must bring the census to 3/2, and then collides with the Varus coins, which cut off two-thirds of the window, and with the grammar.

HV survives precisely by not using the census as a positive constraint. Route B from §5.4 — free on the legates axis — gets its account here: the "before" reading breaks on πρό, on the ordinal and on Acts 5:37, and HV, whose measurements those are, does not take it. So census = Quirinius = AD 6, and HV books the datum as "datum that does not carry the dating". It adds P.Yadin 16 and Nicolet's layer: the οἰκουμένη of Luke 2:1 is the orbis Romanus, client kingdoms are "a different case", and Luke 2:3 is documentarily correct. It does not rescue the census; the grammatical measure it applies against the "before" reading, it applies on its own side. The census thus stands in the same slot for C and for HV — not the favorable outcome, but the honest one.

5.9 What the two axes carry together

The two outcomes are connected. Because the fasti keep Quirinius out of the window — Saturninus and then Varus, coin-anchored to 5/4 — every attempt to rescue Luke 2:2 as a dater of the birth is forced onto the grammar, and there it breaks. C and HV therefore read the sentence on the same grammar and book it the same; SY and CA pay the most on both axes, and the antedating that carries them elsewhere reaches neither a Roman nor a Roman census.

Two falsifiers remain standing. Against the fasti reading: a dated piece from era year 28–34 with Quirinius's name as legate of Syria, or a reversal of ILS 918 — both dormant in the undated ILS 2683. Against "the census does not carry": an attested census of Quirinius under Herod, or a grammatically watertight "before" reading — both not found. P.Yadin 16 belongs with them: it cuts against Luke's Quirinius as a precise dating operator, and HV books that as support for "does not carry"; it does not book it away.