Method: the gates, the marks, the rubric, and what "constraint" means here
7 min read
In-depth:
This work is a derivation, not a survey of the field. It sets up a variable space — the counting choices on which Herod's year of death, and with it the birth window, depend — encodes each primary datum as a predicate that kills or prices cells in that space, runs the space, and reads off which cells survive and at what price. Every assertion hereafter is traceable to such a predicate, every predicate to a source. What cannot be traced to a source is not in it.
The protocol under which that was done has fifteen gates. None was designed from theory; each gate was added after an error in the research had become visible and could be named. Together they block three failure modes. The first: trading uncertainty for confidence, because confidence sounds more useful — a "possibly" that three paragraphs later has become a "therefore". The second: weighing respectability instead of evidence — a position that stays standing because it is the consensus, not because the data bear it. The third: status inflation of the bearer — something transmitted, a summary, a translation, an OCR layer, the report of a search agent, that carries the label of a firsthand measurement. The marks are built against that third.
The load-bearing gates. Gate 1: no negative claim — "no source", "zero mentions" — without three search passes, the third of which reads: who would have refuted this if it existed, and where is that text? Gate 3: the opponent is quoted, not reconstructed; a reconstructed position may not be refuted. Gate 5: primary or nothing; no transmission carries weight in an argument. Gate 13: one's own shelf first; an audit, a search query, an agent report is a claim, not a measurement. Passing all the gates does not make a claim true. It makes its status honest and testable by a third party; this work promises no more.
Constraint and chain
A constraint is a primary datum that restricts the variable space: a regnal figure in Josephus, a coin, a legate, a priestly course, a census. A constraint is neither an argument nor a reading; it is the datum itself, in the form of a predicate that says yes or no over every cell of the space — or, where the datum can be counted away at a price, how much that price is. Every constraint belongs to a source and to a mark.
A chain is a profile of counting choices: from which anchor year one counts, inclusive or non-inclusive, with year-beginning Nisan or Tishri, with or without antedating. Three chains span the field. C is the consensus in the line of Schürer: investiture 40 BC, capture of Jerusalem 37 BC, inclusive reckoning, year-beginning Nisan; year of death 4/3 BC. SY is the chain of Steinmann and Young: investiture 39 BC, capture 36 BC, non-inclusive reckoning with the exception "when an ordinal number is used", year-beginning Tishri; year of death 1 BC, January. CA is Cameron: investiture 40/39 BC, capture 37 BC, antedating, Nisan; the regnal figures yield 4/3, his antedating rule pushes through to 3/2 BC. This work's own position, hereafter HV, shares the calculator with C — the same four input values on the Herod axis — but not all the outcomes: on the baptism axis (the Tiberius count, the 69 year-weeks) it stands apart, and on the season of death it keeps a reservation that C does not make: spring 4 BC; the winter of 4/3 remains possible but is less likely. A chain is not a camp but a calculation profile; only its outcomes are compared.
The rubric and the cost accounting
How a chain deals with a datum is recorded with one letter. L: it reads it plainly, without operation. T: it counts it by its declared counting convention — inclusive, non-inclusive, antedating, Nisan or Tishri. That is legitimate and no reproach; a chain may have a convention, provided it is declared in advance and applied throughout. P: it postulates something outside that convention to make the datum fit. V: it declares the datum erroneous — Josephus was mistaken. E: it emends against the manuscripts. U: it leaves it as an unresolved anomaly.
The cost accounting tallies these letters across the 20 data of the output. A chain's hard cost is V plus E: what it must reject or change against the tradition. P is softer — a postulate may be true — but is counted in. The profile, in advance of the accounting itself: C carries zero rejections and zero emendations; CA one rejection; SY three rejections and eight postulates. C is the cheapest reading, SY the most expensive. Put that way it is not a verdict but a sum, and that is the intention. This accounting has not been written out anywhere before; it is the first of three places where this work measures something not measured before. The second is the accounting for the season, where until now advocates stood without an account. The third is the difference between the two runs.
The marks
Every entry — every constraint, every reading of a source, every calculation — carries a label for the source type from which it comes. SEEN, PRIMARY (also SEEN FIRSTHAND): seen in the primary text itself, as page image, own count or own calculation. DOCUMENTED: transmission — summary, translation, OCR, editorial punctuation; does not bear weight on its own. CALCULATED: derived mechanically or arithmetically. CONSTRUED: a reading or inference, not a measurement. NOT SEEN: not consulted, and therefore without weight. Gate 15, the bearer rule, puts one more field in front of the mark, SEEN FIRSTHAND or TRANSMITTED, and forbids a paraphrase from being sharper than the author: "not necessarily" is not "not". Hereafter the mark appears only where it matters — where a datum circulates in the literature as a measurement and on inspection turns out to be transmission, or the reverse.
Two runs
The space is run twice. Run A is freestanding: only constraints that do not depend on the crucifixion axis — Herod's regnal years, coins, legates, Matthew 2, patristics, priestly courses, census, Luke 1:5, shemitah. No anchor at AD 27, no Luke 3:23. Run B is coupled: run A plus the baptism axis — Tiberius' 15th year, Luke 3:23, the 69-year-weeks anchor at AD 27.
The difference between the two is itself a measurement result. It measures how much the birth window owes to the baptism and crucifixion axis, and thus how much of the window's sharpness disappears when that axis is disputed. That difference has never been measured before; both outputs stand side by side hereafter.
One measure
Gate 11 is the symmetry requirement. One measure, fixed in advance, applied equally to C, SY, CA and to HV's distinguishing positions. No chain an exemption, no chain an extra hurdle. The measure consists of four questions, put to each side. The first is internal logic: does the inference follow from the premises, or is there a "highly likely" in the hinge sentence that has become a certainty in the conclusion? The second is external coherence: does the outcome collide with something already measured? The third is the reading of the source: is it there, or is it read into it — with question marks, square brackets and "possibly" silently counted in? The fourth asks for a falsifier: which datum would break this position, and has that datum been sought?
Two consequences. The consensus is not a null hypothesis the rivals must defeat; it is a claim and is tested as a claim, with the same four questions (Gate 8). And HV's expectation — a birth in 5-4 BC, from the 69 year-weeks — is a hypothesis, not an input; it stands on the side being tested, not on the side doing the testing. Being stricter because a finding helps HV is itself asymmetry, mirrored. The measure is one, or it is no measure.
The missing word
This work proves nothing, and the past participle of that verb does not occur in the text. What it claims is a cheapest reading with the rivals' prices beside it: convergence of independent constraints on one cell, and a figure for what every other cell costs. Where the constraints do not converge — Herod's season of death, the season of the birth — the work says so, and counts it as a result. A question that stays open because the sources do not close it is a measurement. A question that is closed because a closed question reads better is the first failure mode.
This work is a derivation, not a survey of the field. It sets up a variable space — the counting choices on which Herod's year of death, and with it the birth window, depend — encodes each primary datum as a predicate that kills or prices cells in that space, runs the space, and reads off which cells survive and at what price. Every assertion hereafter is traceable to such a predicate, every predicate to a source. What cannot be traced to a source is not in it.
The protocol under which that was done has fifteen gates. None was designed from theory; each gate was added after an error in the research had become visible and could be named. Together they block three failure modes. The first: trading uncertainty for confidence, because confidence sounds more useful — a "possibly" that three paragraphs later has become a "therefore". The second: weighing respectability instead of evidence — a position that stays standing because it is the consensus, not because the data bear it. The third: status inflation of the bearer — something transmitted, a summary, a translation, an OCR layer, the report of a search agent, that carries the label of a firsthand measurement. The marks are built against that third.
The load-bearing gates. Gate 1: no negative claim — "no source", "zero mentions" — without three search passes, the third of which reads: who would have refuted this if it existed, and where is that text? Gate 3: the opponent is quoted, not reconstructed; a reconstructed position may not be refuted. Gate 5: primary or nothing; no transmission carries weight in an argument. Gate 13: one's own shelf first; an audit, a search query, an agent report is a claim, not a measurement. Passing all the gates does not make a claim true. It makes its status honest and testable by a third party; this work promises no more.
Constraint and chain
A constraint is a primary datum that restricts the variable space: a regnal figure in Josephus, a coin, a legate, a priestly course, a census. A constraint is neither an argument nor a reading; it is the datum itself, in the form of a predicate that says yes or no over every cell of the space — or, where the datum can be counted away at a price, how much that price is. Every constraint belongs to a source and to a mark.
A chain is a profile of counting choices: from which anchor year one counts, inclusive or non-inclusive, with year-beginning Nisan or Tishri, with or without antedating. Three chains span the field. C is the consensus in the line of Schürer: investiture 40 BC, capture of Jerusalem 37 BC, inclusive reckoning, year-beginning Nisan; year of death 4/3 BC. SY is the chain of Steinmann and Young: investiture 39 BC, capture 36 BC, non-inclusive reckoning with the exception "when an ordinal number is used", year-beginning Tishri; year of death 1 BC, January. CA is Cameron: investiture 40/39 BC, capture 37 BC, antedating, Nisan; the regnal figures yield 4/3, his antedating rule pushes through to 3/2 BC. This work's own position, hereafter HV, shares the calculator with C — the same four input values on the Herod axis — but not all the outcomes: on the baptism axis (the Tiberius count, the 69 year-weeks) it stands apart, and on the season of death it keeps a reservation that C does not make: spring 4 BC; the winter of 4/3 remains possible but is less likely. A chain is not a camp but a calculation profile; only its outcomes are compared.
The rubric and the cost accounting
How a chain deals with a datum is recorded with one letter. L: it reads it plainly, without operation. T: it counts it by its declared counting convention — inclusive, non-inclusive, antedating, Nisan or Tishri. That is legitimate and no reproach; a chain may have a convention, provided it is declared in advance and applied throughout. P: it postulates something outside that convention to make the datum fit. V: it declares the datum erroneous — Josephus was mistaken. E: it emends against the manuscripts. U: it leaves it as an unresolved anomaly.
The cost accounting tallies these letters across the 20 data of the output. A chain's hard cost is V plus E: what it must reject or change against the tradition. P is softer — a postulate may be true — but is counted in. The profile, in advance of the accounting itself: C carries zero rejections and zero emendations; CA one rejection; SY three rejections and eight postulates. C is the cheapest reading, SY the most expensive. Put that way it is not a verdict but a sum, and that is the intention. This accounting has not been written out anywhere before; it is the first of three places where this work measures something not measured before. The second is the accounting for the season, where until now advocates stood without an account. The third is the difference between the two runs.
The marks
Every entry — every constraint, every reading of a source, every calculation — carries a label for the source type from which it comes. SEEN, PRIMARY (also SEEN FIRSTHAND): seen in the primary text itself, as page image, own count or own calculation. DOCUMENTED: transmission — summary, translation, OCR, editorial punctuation; does not bear weight on its own. CALCULATED: derived mechanically or arithmetically. CONSTRUED: a reading or inference, not a measurement. NOT SEEN: not consulted, and therefore without weight. Gate 15, the bearer rule, puts one more field in front of the mark, SEEN FIRSTHAND or TRANSMITTED, and forbids a paraphrase from being sharper than the author: "not necessarily" is not "not". Hereafter the mark appears only where it matters — where a datum circulates in the literature as a measurement and on inspection turns out to be transmission, or the reverse.
Two runs
The space is run twice. Run A is freestanding: only constraints that do not depend on the crucifixion axis — Herod's regnal years, coins, legates, Matthew 2, patristics, priestly courses, census, Luke 1:5, shemitah. No anchor at AD 27, no Luke 3:23. Run B is coupled: run A plus the baptism axis — Tiberius' 15th year, Luke 3:23, the 69-year-weeks anchor at AD 27.
The difference between the two is itself a measurement result. It measures how much the birth window owes to the baptism and crucifixion axis, and thus how much of the window's sharpness disappears when that axis is disputed. That difference has never been measured before; both outputs stand side by side hereafter.
One measure
Gate 11 is the symmetry requirement. One measure, fixed in advance, applied equally to C, SY, CA and to HV's distinguishing positions. No chain an exemption, no chain an extra hurdle. The measure consists of four questions, put to each side. The first is internal logic: does the inference follow from the premises, or is there a "highly likely" in the hinge sentence that has become a certainty in the conclusion? The second is external coherence: does the outcome collide with something already measured? The third is the reading of the source: is it there, or is it read into it — with question marks, square brackets and "possibly" silently counted in? The fourth asks for a falsifier: which datum would break this position, and has that datum been sought?
Two consequences. The consensus is not a null hypothesis the rivals must defeat; it is a claim and is tested as a claim, with the same four questions (Gate 8). And HV's expectation — a birth in 5-4 BC, from the 69 year-weeks — is a hypothesis, not an input; it stands on the side being tested, not on the side doing the testing. Being stricter because a finding helps HV is itself asymmetry, mirrored. The measure is one, or it is no measure.
The missing word
This work proves nothing, and the past participle of that verb does not occur in the text. What it claims is a cheapest reading with the rivals' prices beside it: convergence of independent constraints on one cell, and a figure for what every other cell costs. Where the constraints do not converge — Herod's season of death, the season of the birth — the work says so, and counts it as a result. A question that stays open because the sources do not close it is a measurement. A question that is closed because a closed question reads better is the first failure mode.