Matthew 2 as a constraint: the interval between birth and death

Matthew 2 yields two data that follow deductively from the text. The first is a sequence: the family stays in Egypt "until the death of Herod",1 so the birth lies before that death. The second is an age limit: Herod has the boys "of two years and under" killed, according to the time he had ascertained from the magi.2 At the massacre the child was thus at most about two years old, and that limit was set by a king who did not want to miss his target, not by a narrator who wanted to state an age: Herod took a margin, and the text does not say how much. Every further figure hung on this axis has been added to these two data.

Before the two data stands a premise that this work has so far made tacitly: that Matthew 2 is a datable report. It deserves the same four questions as every constraint, for run A kills cells on this axis alone (§8.2), and an axis that stands on an undiscussed assumption is no axis. The objection first, at its strongest. Outside Matt 2:16-18 there is no independent account of the massacre at all; Josephus, the only detailed source for Herod's last years, is silent, and there is no Roman source. The critical school reads the episode as a theological composition on the pattern of Moses and the exodus, France defends it as history; neither has been read firsthand for this work, and the positions stand here only as the literature gives them.3 Internal logic: from "no second witness" and "the shape of the exodus" to "did not happen" is an inference; the form in which a report is told does not measure whether it happened, and §4.5 shows that the same form also carries the star. Josephus's silence weighs only where a mention was to be expected, and here it was not: on current estimates Bethlehem had a few hundred to about 1000 inhabitants, so that "all the boys of two years and under" are a handful to a few dozen children, not 14,000 as the Byzantine legend counts; against the backdrop Josephus does record, three of his own sons executed, Mariamne, the burning of the rabbis after the eagle incident, the order to kill the notables in the hippodrome at his death "so that there would be mourning",4 a village drama does not make the chronicle, and his main source for these years was Nicolaus of Damascus, Herod's own court historian. Moreover, §4.1 has already measured that BJ 1 omits a court event of precisely this type from AJ 17.41-45. Correct reading of the one pagan echo: Macrobius has Augustus joke that it is better to be Herod's pig (ὗς) than his son (υἱός), tied to the killing of "the boys under two years in Syria", among them a son of Herod.5 The pun is Greek and the tone Augustan, but Macrobius writes four centuries later, demonstrably mixes the massacre with the execution of Antipater and may depend on the Christian tradition: a late echo, not an independent confirmation, and whoever reads confirmation into it overplays his hand. External coherence: the detail "of two years and under, according to the time he had ascertained from the magi" tallies with the outcome of this chapter, but that is a sum on the hypothesis and not a witness. Falsifier: for historicity a second mention independent of Matthew, and there is none; for the critical reading an identifiable source from which Matthew would have composed the report without an event, and there is none either. The premise is thus a choice, and this work makes it as Casus 31 already made it: it weighs the readings within the testimony, marked INTERPRETED. The price stands beside it: everything that hangs on this axis, the floor on the late side, the ceiling on the early side and the only cause of death in run A, hangs on Matthew 2 as a report. If the premise falls, it is not the window that falls but its late edge, for then only Luke 1:5 remains on that side, and that verse dates the announcement to Zechariah, not the birth.

4.1 What the text withholds

The stay in Egypt is given no duration. The massacre is given no date. The death of Herod is reported, not placed. The τότε with which Matthew ties the massacre to the magi's failure to return,6 is a Matthean stylistic idiosyncrasy and not a time indication; the count of the word in the SBLGNT text confirms that. And the order in which Matthew 2 narrates is demonstrably not the order in which the events lie: the falsifier that had been written down for the strict reading fired, and that reading has been withdrawn. What τότε does do is attach the massacre to the beginning of the Egyptian stay. The child is "two years and under" at the moment the flight has just been completed; what follows afterwards in Egypt does not count towards the age limit. The upper limit of the birth–death interval is thus not "two years" but "two years plus Egypt", and Egypt is undetermined. Davies and Allison put it bluntly: how long the family stayed there is stated nowhere, and the estimates "are just that — guesses".7

Josephus does not close this gap. His densely narrated final stretch, the last seven to 15 months of the reign, is silent on Bethlehem. Whoever wants to use that silence gains from it at most a lower limit for the distance between massacre and death, a support on the side where the constraint is already hard; an upper limit Josephus nowhere supplies. And even that lower limit he does not bear independently: BJ 1 omits AJ 17.41-45, a court event of precisely the type that Matthew 2 narrates, and what AJ 17.1-51 contains is undetermined in time. Three independent routes, τότε, the narrative order and Josephus, come out at the same point: the interval is not measured.

4.2 Smallwood's measure and the status of her two limits

The measure that the consensus hangs on this axis is Smallwood's: the birth lies "at least a year" before Herod's death, "though not necessarily as much as two years", for "Herod will have put the age-limit well above that of his intended victim", and the stay in Egypt "need not have been lengthy".8 The four tests give one picture. Internal logic: the hinge is "will have", an inference about Herod's caution, not a measurement. Correct reading: Smallwood herself builds a reservation into both limits; "not necessarily" and "need not" are there. External coherence: her sentence collides with nothing, because it measures nothing. Falsifier: a text that gives Egypt a duration or dates the massacre does not exist.

What is decisive is that the two figures hang on the same inference. The generous Herod who sets the limit well above the actual age and thus yields "not two years" is the same generous Herod who yields "at least a year". Whoever books the ceiling as inference books the floor as inference. The text nowhere requires that the child was a year old at Herod's death. It requires that it had been born when the magi came, that the magi had traveled, that the family fled and that Herod died thereafter. That is a handful of months, narrative time for a journey and a flight, on the generous side half a year; not a year.

On this point there have been two corrections. HV's research design initially read "two years and under" as an interval measure; Smallwood's text removed that. The first weighing of this axis then pulled the floor up to a year along with Smallwood and thereby cut away 4 BC; HV rejected that, rightly, for a floor of a year is not in the text. That second correction falls in HV's favor, it restores width on the late side, and therefore receives the same treatment as a correction that would tell against him: no stricter test, no milder one.

The result is a one-sided constraint. On the late side stands a robust floor: birth before the death, with months in between. On the early side stands a ceiling made of three layers of inference: Herod's margin, the duration of Egypt and the question whether the "two years" is a datum prior to the chronology or a product of it. Davies and Allison hold the latter possible, and whoever deploys the figure as a date reads it in. On Finegan's ceiling (35), then, Matt 2:16 cuts nothing that Luke 3:23 did not already cut: if the early edge is cut anywhere, it is by Luke's "about thirty", not by Matthew. In years: 5 BC stands within every reading of the ceiling; 6 BC stands within it as soon as Egypt is given any duration; 7 BC is soft, because it must let go of Smallwood's ceiling and that ceiling is only inference; 8 BC rests on no datum any longer and is a postulate.

Smallwood draws a second inference from her interval argument, and on this axis it weighs more heavily than her floor. Late March or early April 4 BC is the date of death "usually given", not hers; 13 March 4 BC "allows inadequate time"; she lands on around February 4 BC, with Passover 4 BC as the limit.9 Kasher and Witztum, without the eclipse, likewise land a calendar year before the winter of 4/3. The interval argument thus pulls the death earlier, not later. Her verdict on 13 March does not hold on the day count of §6.1: the chain from the eclipse to the Passover of AJ 17.213 needs 22 to 57 days, and 13 March leaves 30. Her own February 4 BC lies before that eclipse and therefore needs the one of 15 September 5 BC, with some four months that Josephus does not narrate.

4.3 The axis per chain

The interval hangs not on the year alone but on the day within that year, and in this work the season of birth is a free variable (§6.3, §8.7). This axis is therefore reckoned per part of the year, on the same measure for every part.10 Under a death in 4/3 BC (chain C), the margin of a birth in 5 BC hangs on the season of death. Under a death in the winter of 4/3 it lies amply before the death, in every part of the year: reckoned to both ends of that window of death (Kislev 4 BC to Passover 3 BC), 15 Nisan 5 BC gives 19.5 to 23.8 months and the autumn feasts of 5 BC give 13.3 to 17.6, both within Smallwood's inferential "at least a year". Under a death in the spring of 4 BC (about 3 April) only 15 Nisan 5 BC is ample, at 11.4 months; the autumn feasts of 5 BC lie 5.5 months before the death, and under Smallwood's February 4 BC 4.0: tight, but within the hard floor. 5 BC is thus the only cell that survives on this axis under both seasons of death and in every part of the year; ample in every part of the year it is only under the winter. A birth in 4 BC does hang on the season of death and on the part of the year. If Herod dies in the spring of 4 BC, every birth day of that year from 15 Nisan onward lies after the death and the cell falls on Matt 2:15; that is a seasonal argument, not an interval argument. If he dies in the winter of 4/3, a birth on 15 Nisan 4 BC lies 7.5 months before the death and a birth at the autumn feasts 1.7 to 2.1 months; at Kushnir-Stein's outer limit (Passover 3 BC) that is 11.7 against 5.8 to 6.3. Tight in both cases, excluded in neither — but the difference between the parts of the year is larger than the difference between the dates of death, and on the late edge the part of the year makes more than half a year of difference. The early side, 6 and 7 BC, remains open at the price of the inferential ceiling. C pays almost nothing here. The death in 4/3 itself stands not on this axis but on the elimination elsewhere,11 and Smallwood's February 4 BC is a consensus authority that arrives at an early death on the same ground as HV.

For SY (death January 1 BC) the interval is non-discriminating. Every birth in 4-2 BC amply satisfies both floor and ceiling; the axis yields SY nothing and cuts nothing away. SY gains no datum here at all and leans entirely on the counting technique. To keep the death in 1 BC, SY must drop Smallwood's February 4 BC and reject the entire elimination of the death in 4/3 BC, prices outside this axis. That death hangs on the eclipse of 10 January 1 BC plus antedating; without that eclipse SY has no positive anchor left at all, and the two-year difference with HV consists of two links of one year each with a starting point at zero, counting conventions that the chain does not bear independently elsewhere.

For CA (death 3/2 BC) the axis runs against the grain. First a source flag: the fixed frames give Cameron "death 4/3 BC", his own pieces give 3/2 BC; here the weighing is on 3/2, and under 4/3 CA coincides with C on this axis. Against a death in 3/2 BC, a birth in 5 BC puts the interval at 2.5 to 3.5 years. That fits only with a long Egypt: the same stretch in the inferential ceiling that C needs for 7 BC, with this difference, that CA itself deploys the age-limit reading and with the ceiling gives up its own interval reading. What further costs CA is Smallwood's second inference: the interval argument pulls the death towards February 4 BC, not towards 3/2 BC. If CA wants to keep the birth within Smallwood's measure, it must move to 4-3 BC, and that collides with Luke's "about thirty" and with the standing result. Cameron's 3/2 hangs on the regnal years (37 from 40/39, 34 from 37/36, with antedating), not on the eclipse, which for him is defensive; the axis adds nothing to that and sets one objection against it.

HV's distinguishing positions, birth "at the earliest 5-4", the interval reckoned "generously", winter 4/3 with a tight "thirty", come out adjusted. "Herod reckoned generously" is, symmetrically with Smallwood's ceiling, an inference and not a measurement; it may not make the ceiling stricter or looser than the opposing side. The 5 BC leg stands ample under winter 4/3, and under spring 4 only on 15 Nisan. The 4 BC leg stands tight and only under winter 4/3. The further narrowing to "only 5", with 6 and 7 BC out, would hang on the ceiling; that ceiling is inference, and HV does not bear that cut independently. On the season of death the adjusted position reads: spring 4 BC; the winter of 4/3 remains possible but is less likely (§6.2).

4.4 The cheapest reading and the prices beside it

Cheapest surviving cell: birth in 5 BC under a death in 4/3 BC, and it is so in every part of the year. It survives both seasons of death, lies within the hard floor and within every reading of the ceiling, and stretches no convention. Price on this axis: zero under the winter of 4/3. Under the spring of 4, 15 Nisan 5 BC is ample (11.4 months before the death); the autumn feasts of 5 BC lie 5.5 months before it, under Smallwood's February 4 BC 4.0: tight for magi, flight and massacre.

What this axis does not decide is the choice between 5 and 4 BC within chain C. The interval leaves "5 or 4" standing, with 4 BC as the tightest cell; what would remove 4 BC is a death before the birth day itself, and which day that is depends on the part of the year: before 15 Nisan (early April) or before the autumn feasts (early October). That is a season-of-death question and a part-of-the-year question, and both remain open on the sources. The constraint of Matthew 2 is thus exactly what the text permits and no more: a floor on the late side that stands, and a ceiling on the early side that one reads into it.

4.5 The star as a report: what the text describes, and what can satisfy it

Up to this point Matthew 2 has served only as a constraint on the interval. The same chapter also contains a description, and that description can be tested as an observational report, on one measure for every candidate.

What the text says. The magi come "from the eastern regions" (ἀπὸ ἀνατολῶν, plural) and have seen "his star ἐν τῇ ἀνατολῇ" (singular, with the article); the difference stands in the Greek and carries the reading "at its rising".12 Herod ascertains "the time of the appearing star" (τὸν χρόνον τοῦ φαινομένου ἀστέρος).13 Then: "and behold, the star which they had seen at its rising went before them (προῆγεν), until it came and stood over (ἐστάθη ἐπάνω) where the child was. When they saw the star, they rejoiced with exceedingly great joy."14 Five properties follow. The star (A) appears; (B) is for a time out of view; (C) is recognisably the same object when it returns (ὃν εἶδον); (D) goes before travellers on a road of a few hours southward; (E) comes to stand over a place.

Property B is inferred, and the inference is strong. Matthew does not narrate the journey. He nowhere says that the star led the magi to Jerusalem, and he nowhere says that it vanished. But two features of 2:9–10 point that way. ἰδού marks what comes into view unexpectedly. And the joy of 2:10, fourfold intensified (ἐχάρησαν χαρὰν μεγάλην σφόδρα), is tied to the seeing itself (ἰδόντες). A sign that had stood in view without interruption since the East explains neither the "behold" nor a joy that attaches to seeing it. The most economical reading is that the magi saw the star again after a period in which they had not seen it. The alternative, joy at a confirmation of direction after Herod's secret audience, is possible, but it places the cause of the joy where the text does not put it. This work adopts the renewed sighting, marked INTERPRETED.

One physical fact applies to every celestial candidate. Whatever is fixed to the celestial sphere, whether star, planet, nova or comet, turns with it from east to west in the course of a night. On a walk of two hours southward it "goes before" the walker only if it happens to stand high in the south. And it can never stand still over one house or one village: the point beneath it lies thousands of kilometres away and does not shift when the observer steps aside.15 Property E therefore admits only two ways out: ἐστάθη ἐπάνω as a direction, as Martin reads it, or an object that is not far away.

The elimination.

On the celestial side only a comet satisfies all five, with E as an optical impression: a coma with its tail pointing upward appears to hang over what lies beneath it, the image Dio uses for the comet over Rome in 12 BC.16 The cheapest physical mechanism for B and C is a comet seen before perihelion, lost in the Sun's glare, and seen again after it. A conjunction fails on C, a nova on B.

Ho no. 63 fits only half. At its first sighting the lodge Qianniu (β Cap) stood at the start of civil twilight 35° to 39° high in the south-south-east to south, and thus in front of a walker from Jerusalem to Bethlehem.17 But B and C must then fall within its "more than 70 days". That requires a departure immediately after the first sighting and a journey of one to two months at most, where Ezra 7:9 gives four; and the Chinese entry records neither motion nor a switch from morning to evening sky.18 The one contemporary interpretation of the object is Chinese and dynastic. The Han Shu finds it fulfilled in the failed renewal of the Han mandate in the summer of 5 BC and, later, in Wang Mang's usurpation; and in the Chinese field allocation Qianniu belongs to Wu-Yue in the south-east of China, just as Capricorn in Ptolemy belongs to India and Macedonia and not to Judaea.19

The other registers are silent, and the silence weighs against nothing. The latest dated Babylonian observation is a solar eclipse of 30 June 10 BC. After it only almanacs survive, for 7/6 BC and then for AD 16/17, and an almanac is a computed forecast that cannot contain a comet.20 The Greco-Roman record has no comet between Halley in 12 BC and AD 9.21 No Parthian astronomical archive survives. The Chinese record is the only one from these years that survives and noted comets systematically.

Taken literally, the five properties point to something near. All five elements of Matt 2:9 already stand in the Greek Pentateuch, around the guiding light of the exodus: ὁ δὲ θεὸς ἡγεῖτο αὐτῶν [...] δεῖξαι αὐτοῖς τὴν ὁδόν [...] ἐν στύλῳ πυρός; ὃς προπορεύεται πρότερος ὑμῶν ἐν τῇ ὁδῷ ἐκλέγεσθαι ὑμῖν τόπον [...] ἐν πυρὶ νυκτός; κατέβαινεν ὁ στῦλος τῆς νεφέλης καὶ ἵστατο ἐπὶ τὴν θύραν τῆς σκηνῆς; ᾗ ἐστάθη ἡ σκηνή [...] ὡς εἶδος πυρός.22 Going before, showing the way, choosing a place, coming to stand over a dwelling, light by night. Num 24:17 supplies the verb for a star out of Jacob, ἀνατελεῖ ἄστρον ἐξ Ιακωβ, spoken by a seer from the East.23 And John Chrysostom already asked the same questions: the star moved "from north to south", "hid itself" once the magi were in Jerusalem and "shows itself" again after they left Herod, and "came down, and stood over the very head of the young child". His conclusion: "some invisible power transformed into this appearance".24 He reads more than Matthew says, namely that the star led the magi to Palestine and shone by day; but his core is the text's core.

Three positions remain, each with its price.

1. A celestial trigger, told in the form of the exodus. The magi saw something real "at its rising", no. 63 or a planetary phase, and Matthew tells the last stretch in the language of Exodus and Num 24. Price: E is no longer an observation, and which object it was remains undecided. 2. A comet, taken literally. New, gone, back, going before, and "over" as an optical impression. Price: it rests on one Chinese object with a tight interval and no recorded return, and the one contemporary reading of that object points to China. 3. A non-natural light in the appearance of a star. It satisfies all five properties. Price: it cannot be measured and yields no year.

Position 3 has a textual support that the other two lack. Herod has to ascertain from the magi when the star appeared (2:7); Jerusalem is troubled by their report, not by a sign in its own sky (2:3); and no surviving register records the object. Scripture knows the pattern of a sign seen by those for whom it is meant and not by others who stand beside them. Elisha's servant sees the hills full of horses and chariots of fire only when his eyes are opened (καὶ διήνοιξεν κύριος τοὺς ὀφθαλμοὺς αὐτοῦ, καὶ εἶδεν); Balaam's donkey sees the angel standing in the road before Balaam does, the same Balaam whose oracle supplies the rising star out of Jacob.25 The Damascus road shows the pattern in a finer form: the companions "heard the voice but saw no one", "saw the light but did not hear the voice of the one speaking to me", while the light shone around them all.26 Selective sight explains the silence in Jerusalem and in every register. It also makes position 3 immune to measurement, since it can absorb any record and any absence. That is its strength as exegesis and its limit as evidence: whoever holds it gains the most coherent reading of Matthew 2 and gives up the star as a dating instrument. The window of this work does not need the star (§8.2).

What this does to the window. Nothing, and that is the result. The star stands in the model as an empty predicate (§8.2); this section shows why it has to stay empty. Position 3 yields no date. Position 1 dates at most the trigger, not the birth. Position 2 dates the arrival of the magi to within some 70 days of the first sighting of no. 63, with the birth before that arrival at an undetermined distance (ὁ τεχθείς, Matt 2:2). Whoever dates by the star must choose between positions 1 and 2 and say what he does with Matt 2:9b. Falsifier: a dated first sighting of no. 63 independent of Matthew, or a recorded switch of no. 63 from the morning to the evening sky, would move position 2 from possible to measured.