Typology: patterns tested against controls

The short version

1. The worry is real.

   One entrant in a skeptics' contest found six or more "spooky" coincidences for each of 21 pairs of US presidents. Loose pattern-hunting finds parallels for anyone.

   The evidence ↓
2. So we fixed five rules before scoring anything.

   Old text; structural, unusual matches; a forward-looking reading before Christianity; and facts about Jesus attested apart from the writer who drew the parallel.

   The rules ↓
3. The controls do as well as the Christian types.

   Lincoln as a new Moses and Muhammad in Deuteronomy 18 pass our rules. In a scan of all 929 Old Testament chapters, Jesus sits at about the 30th percentile of random figures.

   The controls ↓
4. One thing survives: Jesus died in Jerusalem at Passover.

   It is early and multiply attested, and Jewish tradition had already tied Passover to the binding of Isaac. But a Galilean prophet was most likely to die in Jerusalem at a pilgrim feast anyway.

   The survivor ↓
5. Typology adds about a factor of 2: no margin for design.

   Thirty items worth about 1 multiply to 1. The survivor is worth about 2, but the range (0.9 to 6) reaches below 1. It is counted once, as a small, discounted line in the prophecy cluster, not multiplied into messianic prophecy, which also finds no margin.

   The analysis ↓

The question

The Old Testament has persons, events and institutions that resemble features of Jesus' life and death: Isaac on Moriah, the Passover lamb, the scapegoat, Joseph, Jonah, the bronze serpent, Melchizedek. Christians call them types. The question is narrow: is the resemblance greater than unconstrained pattern-finding, plus the New Testament writers' deliberate use of the Old Testament, would produce?

Two families of explanation compete.

• Design (H). The God of Israel shaped Israel's history and texts so that they prefigure Jesus.
• The naturalistic account (N), four mechanisms that work together:
  • Selection and chance. In a corpus of 929 chapters and a rich life story, a reader free to choose features and read them loosely will find multi-feature parallels for anyone ("faces in the clouds").
  • Narrative shaping. The Gospel writers told Jesus' story in the patterns of the Old Testament and picked or added details to match. Crossan calls this "prophecy historicized"; Goodacre's middle position is "history scripturalized", real events retold in scriptural colours.¹
  • Enactment. First-century Jews, Jesus among them, acted out scriptural patterns on purpose: festival timing, the donkey entry, choosing twelve, Theudas promising to part the Jordan.²
  • Shared archetype. Israel's literature has recurring motifs, such as the beloved son endangered and exalted, or the righteous sufferer vindicated. Any later Jewish hero story will echo them.³

N is not a rival of all theism: a believer can hold that typology is a human literary practice. What separates H from N is where the best fits fall.

The three bold rows discriminate. The rules below are built to test them.

Rules of evidence

A type is a claim that an Old Testament item corresponds to a cluster of features in Jesus' life and death. Each claimed type is scored 0, 1 or 2 on five criteria, fixed before any scoring and applied in the same way to the controls.

The evidence

The coincidence machine: Lincoln and Kennedy

The list circulated from 1964: elected to Congress in '46 and President in '60; successors named Johnson; assassins killed before trial; shot on a Friday, beside their wives; Lincoln's "secretary Kennedy" and Kennedy's secretary Lincoln.⁴ Held to the rules, nothing survives.

In 1993 the Skeptical Inquirer ran a "Spooky Presidential Coincidences Contest". One winner listed 16 coincidences between Kennedy and the Mexican president Álvaro Obregón. Another found lists for "no fewer than 21 different pairs of U.S. presidents. None involved fewer than six coincidences." The editors' moral: "If the domain from which the entities are selected is not narrowly defined in advance, amazing-sounding coincidences are even easier to find."⁵ That is the null hypothesis for typology, and it is not a straw man.

The controls, scored on the same rules

We took figures nobody thinks the Old Testament was designed to foreshadow, drew their parallels the way their own advocates draw them, and scored them exactly as we score the Christian types.

Four of the six controls pass the same rule that only one Christian type passes, and none of the six, like none of the Christian types, gets full marks. Plato's just man fits independently attested facts about Jesus (scourging, crucifixion) better than most Old Testament types do; it fails only because it was a thought experiment, not a prophecy. The Teacher of Righteousness shows that reading Scripture as fulfilled in one's own leader was ordinary Jewish practice in Jesus' century, so narrative shaping had a ready mechanism.

The haystack: a scan of all 929 Old Testament chapters

To measure how easily loose matching finds "foreshadowing", we gave Jesus and six control figures 20 biographical features each, written as keyword patterns before the run and not tuned afterwards. For each chapter of the Old Testament (KJV¹²) the script counts how many of a figure's features occur. We then built 2,000 random "figures" from the pooled features as a null.¹³

Every figure gets multi-feature "matches". Jesus' best chapter is Ezra 6 (a temple dedication and a Passover), not Genesis 22. The figure most "foreshadowed" is Judas Maccabeus, a Jewish hero whose story is told in the Old Testament's own idiom: that points to shared culture, not design. This is a crude proxy, since words are not structures. Its job is narrow: to show the size of the haystack.

The Christian types, scored

The whole catalogue, about 30 types, scored on the same five rules. Facts about Jesus are graded Established, Inference or Contested in the drill-down below.

Count. Of about 30 types, none gets full marks and one passes the rule: the death at Passover. Four carry pre-Christian anticipation (the prophet like Moses, the Davidic king, Melchizedek, the Day of Atonement in 11Q13), but the Jesus side is either theology or fits several claimants, so they go to the messiah page as expectations, where, scored blind with everything else Jews expected, they give no margin either. The rest fail.

The survivor: Passover and Moriah

Here is what holds up. Jesus was executed in Jerusalem at Passover. All four Gospels say so, John independently of Mark on the chronology, and Paul's letter of about AD 53–55 already links his death with Passover: "Christ, our Passover, has been sacrificed in our place" (1 Corinthians 5:7, WEB). Nobody had a motive to invent the season.

And the season already carried meaning before Christianity. Genesis sends Abraham to "the land of Moriah" with "your son, your only son, Isaac, whom you love" (Genesis 22:2, WEB), and Abraham tells him, "God will provide himself the lamb for a burnt offering, my son" (22:8). Centuries later the Chronicler places the Temple "at Jerusalem on Mount Moriah" (2 Chronicles 3:1, WEB). The Book of Jubilees, a pre-Christian retelling of Genesis, dates the binding of Isaac to the Passover season (Jubilees 17:15–18:19), and a Qumran text (4Q225) develops the story.¹⁷ So a Passover death in Jerusalem landed on a pattern Jews had already drawn: the beloved son, the provided lamb, the mountain of the Temple, the feast of deliverance.

The analysis

A pattern is evidence for design only if it is more probable under design than without it. The likelihood ratio (LR) is that comparison: P(evidence | design) divided by P(evidence | no design). An LR of 1 means the evidence does not discriminate.

1. Parallels that are generic, post-hoc, drawn from details only the parallel-drawer reports, or enacted on purpose are about equally probable either way. The controls get them, chance yields them in numbers, and shaping and enactment explain them. Their LR is about 1.
2. Only one correspondence rests on independently attested facts outside the narrators' control and on a pattern read as forward-looking before Christianity: the death in Jerusalem at Passover.
3. Its LR is modest, because a Passover death is also fairly probable without design.
4. The impression of a "symphony" comes mostly from items with an LR of 1. Counting their number is the error the Lincoln–Kennedy list makes.

E1  death at Passover     P(E1 | design) = 0.7 to 1.0  (central 0.85)
                          P(E1 | no design) = 0.3 to 0.6  (central 0.45)
                          LR(E1) = 1.17 to 3.33, central 1.89
E2  Akedah-Moriah link    mostly redundant with E1 + "in Jerusalem"
                          LR(E2) = 1.0 to 1.5, central 1.2
E3  the other ~28 types   LR(E3) = 0.8 to 1.2, central 1.0

Typology combined         worst 0.93, central 2.27, best 6.00
Monte Carlo (log-uniform within the ranges):
                          5th pct 1.49, median 2.37, 95th pct 3.75

Why not the naive number? The chance of dying in the eight days of Passover and Unleavened Bread on a random day is 8/365, about 0.022, which would give an LR of about 39. That treats the day of death as random. It ignores that a pilgrim prophet's presence and arrest in Jerusalem were concentrated at the festivals. And it ignores the interpreters' freedom: had he died at Tabernacles (water and light, John 7–8), Pentecost (the Law and the Spirit) or the Day of Atonement (the scapegoat, an even better fit), the catalogue would feature that feast instead.

How typology sits beside messianic prophecy

In the unified argument, typology is a small line of its own. Its one factor, about 2 (range 0.9 to 6), sits in the same prophecy cluster as messianic prophecy and the prophecies about nations, where overlapping lines are discounted rather than multiplied in full. Three reasons keep it small and separate.

• Its strongest items are prophecy. Everything here with real pre-Christian anticipation (the prophet like Moses in 4Q175, the Davidic king, Melchizedek in 11Q13) is an expectation, and expectations are what the messiah page tests. It has now scored them blind, with everything else Jews before AD 70 expected and with every miss, and the whole set gives no margin: a likelihood ratio of about 0.3, at most about 2. Scoring them here too would count the same texts twice.
• Its survivor rests on facts other pages carry. The execution in Jerusalem belongs to the Gospels and resurrection pages; "the third day" and "firstfruits" belong to the resurrection page. The messiah page leaves out every feature about the death by rule. What is left for typology is only the extra fit of the Passover timing, and it is counted here and nowhere else.
• It cannot be multiplied type by type, or into messianic prophecy. Thirty items with an LR of about 1 multiply to about 1. Treated as thirty independent "hits", they would inflate the case in exactly the way the controls show is wrong. Nor should the factor of 2 be multiplied into the messiah page's result: neither line, scored honestly, carries the case.

So the two pages agree. Scored with their misses and against controls, the texts do not show foreknowledge, and the patterns do not show design. What they show is how Jesus' first followers read a death they did not expect. The weight of the case lies in the events, weighed on the resurrection page.

Objections

What this shows, and what it doesn't

Typology remains a rich way to read the Bible and shows how the first Christians made sense of a death they did not expect; that is real devotional and literary value, just not evidence of a divine Author.

What we don't claim

• "Dozens of independent authors wove the pattern by accident." They were not independent: later writers knew the earlier texts. And the pattern was drawn by the New Testament readers, which is not an accident.
• "The odds of these types fitting by chance are astronomically small." Most types have an LR of about 1; lenient matching produces hundreds of parallels; and every post-hoc conjunction looks "unique".
• "Moriah is Golgotha" / "on the same mountain". 2 Chronicles 3:1 puts the Temple on Mount Moriah; Golgotha was outside the walls. Say "in Jerusalem".
• "Jesus died at the very hour the Passover lambs were slain." That is John's chronology; the Synoptics differ, and astronomy does not settle it.
• "Even the Talmud confirms he died on Passover eve." b. Sanhedrin 43a was redacted late, is polemical, has textual variants and probably depends on Christian tradition. It is hostile tradition agreeing, not independent evidence.
• "No bone broken" as a fulfilled sign. Only the writer who cites the verse reports it.
• "Joseph was sold for silver, Jesus for thirty pieces of silver." Joseph was sold for twenty; the thirty is Matthew's alone and comes from Zechariah.
• "Jesus fulfilled the sign of Jonah: three days and three nights." Friday to Sunday is not three nights, and Luke's version has no time element.
• "Joshua's name proves the type." It was one of the commonest Jewish names.
• "Firstfruits fell on Resurrection Sunday." Only on the Sadducean reckoning, which the Pharisees disputed.
• "Typology is unique to the Bible; no one else's story fits." Lincoln-as-Moses, Muhammad in Deuteronomy 18, Plato's just man and Bar Kosiba's star show otherwise.
• "The Jews always read Isaac, Joseph or the scapegoat as the Messiah." The expiatory Isaac and the Messiah son of Joseph are mostly rabbinic and post-Christian. What is pre-Christian (Jubilees, 4Q175, 11Q13) we cite precisely.
• "The types are too numerous, too precise and too systematic to be anything but the signature of one Mind." No measure of chance comes with that claim. Measured, the number of threads adds nothing: the controls get comparable lists, and one correspondence passes.
• "Typology was a necessary preparation, without which Jesus could not have been recognized." A purpose claim with no test; the Gospels say the recognition came after the event.
• "The three temptations are a complete set: there is nothing else to test." Complete only because the three categories (world, self, God) are chosen to cover everything; Luke's order differs, and the scene has no witness.
• "Jesus scores high on the hero scale, so he is a myth." Moses, Lincoln and many historical figures fit too. The scale cannot discriminate either way.

Sources

1. Goodacre, M. "Scripturalization in Mark's Crucifixion Narrative." In van Oyen & Shepherd (eds.), The Trial and Death of Jesus (Peeters, 2006), 33–47. DukeSpace. Crossan, J. D. Who Killed Jesus? (1995). Cited from the scholarly literature; not re-checked by us.
2. Josephus, Antiquities 20.97–99 (Theudas), tr. Whiston. lexundria.com. Read. Ant. 20.169–172 (the Egyptian) not re-checked.
3. Levenson, J. D. The Death and Resurrection of the Beloved Son (Yale, 1993). Not re-checked by us.
4. "Lincoln–Kennedy coincidences urban legend." Wikipedia. en.wikipedia.org. Read.
5. "Our 'Spooky Presidential Coincidences Contest' Winners." Skeptical Inquirer 17 (Winter 1993): 212–213. PDF. Read.
6. Rankin, J. E. "Moses and Joshua," sermon, Winthrop Church, Charlestown, 19 April 1865. Remembering Lincoln catalogue (catalogue entry read). On Moses in the Lincoln memorial sermons: Lincoln Lore (not re-checked).
7. Deedat, A. What the Bible Says about Muhammad (1976). Not re-checked by us. Deuteronomy 18:15, 18 quoted from the World English Bible, ebible.org.
8. Jerusalem Talmud, Ta'anit 4:5 (68d), tr. H. Guggenheimer, Sefaria. Quotation checked. Schäfer, P. The Bar Kokhba War Reconsidered (2003). Not re-checked by us.
9. Plato, Republic 361e–362a, tr. Paul Shorey. Perseus. Read (the closing clause).
10. 1QpHab (the Habakkuk Commentary) 7:1–5. Cited by reference; not re-checked by us.
11. "Rank–Raglan hero archetype." Wikipedia, summarizing Raglan, The Hero (1936), Dundes (1977/1990), Sienkewicz and Utley (1965). en.wikipedia.org. Summary read. Carrier, R. On the Historicity of Jesus (2014), not re-checked.
12. King James Version (public domain), machine-readable text: thiagobodruk/bible, en_kjv.json. Read by our script.
13. Our scripts, 10-typology-controls.py (base rates, lexical scan with controls, random-figure null, chance model; seeded, but the random null still varies slightly between runs) and 10-typology-lr.py (likelihood ratio and Monte Carlo). All numbers on this page are from their saved output.
14. Tacitus, Annals 15.44. Cited by reference.
15. Josephus, Jewish War 2.223–227, tr. Whiston. lexundria.com. Read. Antiquities 17.213–218 (Archelaus) not re-checked.
16. Ilan, T. Lexicon of Jewish Names in Late Antiquity I (2002), via Bauckham, R. Jesus and the Eyewitnesses (2006), ch. 4. Not re-checked by us.
17. Fitzmyer, J. A. "The Sacrifice of Isaac in Qumran Literature." Biblica 83 (2002): 211–229. bsw.org. Abstract and final page read. VanderKam, J. C. "The Aqedah, Jubilees, and PseudoJubilees" (1997), not re-checked. Genesis 22 and 2 Chronicles 3 quoted from the World English Bible, ebible.org.
18. Humphreys, C. J. & Waddington, W. G. "Dating the Crucifixion." Nature 306 (1983): 743–746. Brown, R. E. The Death of the Messiah (1994). Not re-checked by us.
19. Babylonian Talmud, Sanhedrin 43a, Soncino translation with notes. halakhah.com. Read.
20. Schäfer, P. Jesus in the Talmud (Princeton, 2007). Not re-checked by us.
21. Sandmel, S. "Parallelomania." Journal of Biblical Literature 81 (1962): 1–13. Not re-checked by us.
22. 4Q175 (4QTestimonia), content summary with Vermes' translation linked: virtualreligion.net (summary read). 11Q13 (11QMelchizedek): Wikipedia and translations by García Martínez–Tigchelaar and VanderKam; 1QS 9:11; 4Q174; 4Q252; Psalms of Solomon 17; Wisdom 16:6–7; m. Menaḥot 10:3; b. Sukkah 52a (cited by reference; not re-checked). Mitchell, D. C. "The Fourth Deliverer: A Josephite Messiah in 4QTestimonia." Biblica 86 (2005): 545–553 (not re-checked).
23. Hays, R. B. Echoes of Scripture in the Letters of Paul (1989) and Echoes of Scripture in the Gospels (2016); Allison, D. C. The New Moses: A Matthean Typology (1993). Not re-checked by us. New Testament verses quoted from the World English Bible.