Ancient Historiography and Compositional Convention

Summary

The four Gospels tell many of the same stories, and they do not tell them identically.  One says a blind man, another says two.  One puts the crucifixion at the third hour, another has Jesus still before Pilate at the sixth.  Matthew's genealogy of Jesus has 27 generations from David where Luke's has 42.

Skeptics call these contradictions.  A school of scholarship now answers that ancient biographers used "compositional devices" -- transposition, conflation, spotlighting, displacement -- and that the Gospels are doing what Plutarch did.

This article tests that answer against the ancient sources themselves, and finds it partly right and heavily over-applied.  There is a better ordering of explanations, and it starts somewhere much more ordinary.

Four explanations, ranked from most ordinary to least.  Each difference belongs on the lowest rung that accounts for it.

1. Honest witnesses to the same event differ on peripheral detail. This is not an apologetic invention. It is what American courts instruct juries in criminal trials: "People often forget things or make mistakes in what they remember. Also, two people may see the same event but remember it differently."¹ The memory data give the pattern a measurable shape. Over ten years, subjects' memory of what happened on 9/11 held while their memory of their own circumstances decayed to about 60% consistency, and their confidence never dropped.^{2 2}
2. Selection, compression, and arrangement. Every narrator skips, shortens, and groups by topic. Ancient writers said so about themselves. Around AD 130 Papias wrote that Mark recorded Peter's preaching "accurately, though not in order."³ Augustine described Matthew compressing the fig-tree episode across a day boundary in about AD 400, and noticed that you can only see the compression because Mark's fuller version survives.⁴
3. Literary dependence. Where Matthew reproduces 51% of Mark's actual words, and Matthew and Luke share an unbroken run of 24 identical Greek words, nobody is misremembering.^{5 6} One writer had the other's text. A difference inside such a passage is an editorial act. But editorial is not the same as invention -- correcting, supplementing, shortening and clarifying are all changes, and all are still reporting.
4. Deliberate reshaping -- the "compositional devices." Reserved for cases where the first three fail and the author demonstrably knew the other version and changed it for effect.

What the evidence shows:

• The devices are inferred, not attested. Plutarch tells the same story twice in a dozen places and never once calls it a technique. Casca's first blow lands in Caesar's neck in one Life and near his shoulder in another.^{7 8} The single place Plutarch flags his own procedure is Cato Minor 25.5, where he says he has put material out of order because it fitted the topic.⁹ Michael Licona, the leading proponent, concedes the point in his own conclusion: the devices are ones "we inferred from Plutarch's Lives."⁹
• Rung 3 is real but narrow. Robert McIver and Marie Carroll ran the experiment. Volunteers reproducing a text from memory never once produced a run longer than 15 identical words; the copying group routinely did. Their rule: 16 or more identical words, outside poetry and aphorism, means copying.¹⁰ Applied to the Synoptics, that identifies nine passages where copying is near-certain. Their own verdict on the rest is the one apologists and skeptics both skip: "overall, the majority of the parallels in the Synoptic Gospels have characteristics that are more typical of a process of transmission that involves memory rather than straight copying."¹⁰
• Numbers inflate everywhere in antiquity, and the New Testament is the outlier in the honest direction. Herodotus totals Xerxes' expedition at 5,283,220.¹¹ Josephus counts 1,100,000 dead in the siege of Jerusalem.¹² The largest New Testament figure claiming to be a count is 5,000, and nine of twelve crowd figures are marked "about" in the Greek. Matthew, Luke, and John each add "about" to Mark's flat 5,000.¹³
• Every good ancient historian has documented errors, and classicists price them in rather than discarding the author. Ronald Syme catalogued roughly fourteen demonstrable mistakes in Tacitus' Annals.¹⁴ When a bronze inscription of AD 20 turned up and could be checked against Tacitus, he got the names, offices, procedure and penalties right ninety years after the fact -- and compressed a months-long trial into what reads like days.¹⁵
• "Luke is a historian of the first rank" means less than it is usually made to mean. William Ramsay, who wrote it, defined a first-rank historian in the same breath: "historical truth implies not merely truth in each detail, but also truth in the general effect, and that kind of truth cannot be attained without selection, grouping, and idealisation."¹⁶
• A residue remains, and this article does not rescue it. Six differences survive every rung: the hour of the crucifixion, the day of the Passover, Judas' death, Matthew 27:52-53, the Quirinius census, and Acts 5:36 on Theudas. Licona names five of his own that leave him "bewildered," all inside the resurrection narratives, and concludes: "Therefore, the tensions remain."⁹

The cost, stated plainly.  Rung 1 implies that on occasion one evangelist got a peripheral detail wrong.  Hirst's ten-year data show that an inconsistent peripheral memory, once formed, tends to be repeated rather than corrected, and held confidently.²  That costs strict inerrancy about incidentals.  It costs the historical case nothing, because courts and historians alike treat peripheral inconsistency as normal and core agreement as what matters.  This article is not an inerrancy defence and does not pretend to be one.

And a cost on the other side.  Every difference moved up to rung 4 is a difference that can no longer be used as evidence that the accounts are independent.  Lydia McGrew makes the point formally:  "it is not possible simultaneously to theorize that the differences between accounts are due to the authors' embellishment while also arguing persuasively that the accounts have the relevant kind of independence for multiple attestation."¹⁷  The devices argument and the undesigned-coincidences argument draw on the same pool of differences and cannot both spend it.

Implications

The convention argument does real work, but not where it is usually deployed.  It handles compression, arrangement, and loose citation, which is most of what a reader actually trips over.  It does nothing for the six hardest cases, and on two of those its leading proponent's own explanation fails the most basic test -- it invokes a convention no ancient author ever names.

The strongest thing that can be said for the Gospels here is not that ancient genre excuses their differences.  It is that their differences look like the differences between honest accounts, that their numbers are hedged where ancient numbers are not, and that they take less license with their sources than the ancient authors they are compared to.  Licona's own finding, following F. Gerald Downing:  "the extent of editing by the evangelists is minimal by ancient standards...  it is the refusal of their authors to paraphrase more freely that is striking."⁹

Four Explanations, Ranked by How Ordinary They Are

The reason to rank them matters more than the ranking.

When four accounts of the same event differ, there are always several possible explanations, and they are not equally likely.  A researcher who reaches first for the most elaborate one -- a deliberate literary technique known to the author and his audience -- has to carry a much heavier burden of proof than one who reaches first for "two people remembered it differently."  Reaching for the elaborate explanation first is how a field talks itself into seeing technique everywhere.

So this article assigns every difference to the lowest rung that accounts for it, and reserves the word "device" for the top rung alone.

Rung 4 is last for a reason.  It is the only rung that requires knowing what an author intended.

Rung 1: Honest Witnesses Differ on Details

Courts tell juries this in criminal trials

The oldest form of this argument is legal, not theological.  Simon Greenleaf was Royall Professor of Law at Harvard and the author of the standard American treatise on evidence.  In 1847 he applied the rules of evidence to the four evangelists:¹⁸

"The character of their narratives is like that of all other true witnesses, containing, as Dr. Paley observes, substantial truth, under circumstantial variety.  There is enough of discrepancy to show that there could have been no previous concert among them;  and at the same time such substantial agreement as to show that they all were independent narrators of the same great transaction, as the events actually occurred."

His control cases were deliberately non-biblical.  He cites the contradictions among ten eyewitness narratives of the French royal family's flight to Varennes in 1792 -- accounts "by eye-witnesses and personal actors," which "contradict each other, some on trivial and some on more essential points, but in every case in a wonderful and inexplicable manner."  And yet, he notes, "these contradictions do not, in the general public estimation, detract from the integrity of the narrators."¹⁸

Greenleaf's key sentence is borrowed, and it is usually miscredited.  He quotes "an eminent writer" on concurrence, and apologists who repeat it often attribute it to Thomas Starkie.  Greenleaf's own footnote names George Campbell, whose Philosophy of Rhetoric is where the passage originates:¹⁹

"in a number of concurrent testimonies (in cases wherein there could have been no previous concert), there is a probability distinct from that which may be termed the sum of the probabilities resulting from the testimonies of the witnesses...  That such a concurrence should spring from chance, is as one to infinite;  that is, in other words, morally impossible.  If, therefore, concert be excluded, there remains no other cause but the reality of the fact."

This is not a nineteenth-century curiosity.  It is what American courts still instruct juries to do.  The Ninth Circuit's model criminal instruction on witness credibility, revised in March 2024:¹

"Sometimes a witness may say something that is not consistent with something else he or she said.  Sometimes different witnesses will give different versions of what happened.  People often forget things or make mistakes in what they remember.  Also, two people may see the same event but remember it differently.  You may consider these differences, but do not decide that testimony is untrue just because it differs from other testimony."

California's pattern instruction says the same thing:  "Do not automatically reject testimony just because of inconsistencies or conflicts...  People sometimes honestly forget things or make mistakes about what they remember."²⁰

These are not apologetics.  They are the rules a jury is given when a man's liberty depends on the answer.  Rung 1 is the ordinary standard of proof in the highest-stakes fact-finding institution we have.

The memory data: the core holds, the edges drift

The courtroom principle has been measured.

After the September 11 attacks, William Hirst and colleagues surveyed more than 3,000 people in seven American cities, then went back to the same people at 11 months, 25 months, and ten years.  They tracked two different things separately:  memory for the event (what happened) and memory for the witness's own circumstances (where you were, who told you, how you felt).

Event memory held.  Accuracy about the crash sites went from 93% at one week to 88% at 35 months.  Accuracy about the order of the major events went from 88% to 86%.²¹

Personal-circumstance memory did not.  Consistency with the subject's own week-one report fell to 63% at 11 months.  Memory for emotions fell to 42%.  Meanwhile confidence barely moved, from 4.41 to 4.25 on a five-point scale.²¹

The ten-year follow-up is the sharper result, because the researchers did not decide for themselves which facts were central.  They had fifty independent raters score each probed fact for how "critical" it was to any story one might tell about the attack.  Then they looked at what had decayed.  Their finding:  "The forgetting observed during the first year can be traced to the non-critical items."²

That is the shape of rung 1, established by outsiders' ratings rather than by anyone's theory.  The core holds.  The periphery drifts.  Confidence tracks neither.

Map it onto the Gospels.  That Jesus was crucified under Pilate;  that women found the tomb empty is event memory.  Which women, at what hour, one angel or two, who spoke first is exactly the class of detail the data show decaying.

Richard Bauckham, surveying the same literature, notes one finding that matters more than the others for the specific cases in this article:  recollective memories "exclude absolute time information from most events.  While a typical recollection will include information on location, actions, persons, emotions, and thoughts and may include information about the time of day, dates are very uncommon."²²  Clock times are the class of information human memory is worst at carrying.

Where this cuts against us

Three findings in the same literature run the other way, and an honest article states them in the same breath.

Confidence is not evidence of accuracy.  Talarico and Rubin tested Duke students on their 9/11 memories and on an ordinary memory from the same week.  Consistency for both declined at the same rate.  Only belief in accuracy stayed high for the flashbulb memory.  Their conclusion:  "Flashbulb memories are not special in their accuracy, as previously claimed, but only in their perceived accuracy."²³  Any apologetic that leans on "they would surely have remembered that vividly" is leaning on the one variable the data say is uncorrelated with accuracy.

Wrong details stick.  Hirst's ten-year paper found that "inconsistent flashbulb memories were more likely to be repeated rather than corrected over the ten-year period."²  So rung 1 does not predict that all divergent details are reconcilable.  It predicts that some of them are simply wrong, held confidently, and retold.

That is where Greenleaf overreaches, and this article declines to follow him.  He claimed the discrepancies "will not be found sufficient to invalidate their testimony" and that "there are none that will not yield, under fair and just criticism."¹⁸  The first clause is defensible.  The second is a promise the evidence does not support, and it is the same promise this article's residue section refuses to make.

Suggestion can insert details that were never there.  Elizabeth Loftus and John Palmer showed subjects a film of a car crash and asked how fast the cars were going when they smashed, collided, bumped, hit, or contacted each other.  The one changed word moved the mean estimate from 31.8 to 40.5 mph.  A week later, subjects asked with "smashed" were more than twice as likely to report broken glass that was never in the film -- 32% against 14%.^{24 24}

What that experiment shows is real and the article concedes it.  What it does not show is equally important:  no subject denied there had been a crash.  The manipulated variables were a speed estimate -- which the same paper shows people cannot judge accurately even without manipulation -- and one incidental object.  Applied honestly, Loftus and Palmer is evidence about hours, counts, and vivid props.  It is not evidence about whether a crucifixion happened.

Bart Ehrman uses this literature, and concedes more than he is usually quoted as conceding

Ehrman's Jesus Before the Gospels is the standard skeptical deployment of memory science.  His thesis sentence:  the Gospels "are not memories recorded by those who were eyewitnesses...  They are memories of memories of memories."²⁵

What apologists rarely quote is what he grants two chapters in:²⁵

"We tend to remember the 'gist' of an experience pretty well, even if the details get messed up...  there are gist memories of Jesus recorded in the New Testament Gospels that are almost certainly accurate.  At the same time, there are a lot of details -- and in fact entire episodes -- that are almost certainly not accurate...  Much of the gist is correct.  One big question, then, is just how broad does a memory have to be in order to be considered a gist memory?"

Ehrman is not disputing the core/periphery structure.  He asserts it.  His argument turns on two other moves, and both deserve a straight answer.

The chain.  Rung 1 is a claim about witnesses.  If the Gospels are third- and fourth-hand relays, the flashbulb studies are the wrong data, because they measure what a witness retains, not what survives serial retelling.  This is the genuinely load-bearing dispute, and memory science does not settle it.  It is settled, if at all, by the authorship and eyewitness-source question, which is a different article.  This one states the dependency rather than assuming past it.

The width of the gist.  Ehrman's own closing question is the honest one, and it has no experimental answer.  Neither his line nor this article's can appeal to the data to fix it.  What this article stipulates -- as a stipulation, not a finding -- is that a difference is peripheral if changing it does not change what the passage asserts happened.  One blind man or two does not change that a blind man was healed at Jericho.  Whether Jesus rose does not survive the same treatment, and nobody pretends it does.

Rung 2: Selection, Compression, and Arrangement

Nothing on this rung is a technique.  It is narration.  What makes it worth a section is that ancient writers said openly that they were doing it, which removes any suspicion that modern scholars invented the category to rescue the Gospels.

Papias, around AD 130, on Mark

The earliest surviving Christian comment on how a Gospel was composed already separates accuracy from order:³

"Mark, having become the interpreter of Peter, wrote down accurately, though not in order, whatsoever he remembered of the things said or done by Christ.  For he neither heard the Lord nor followed him, but afterward, as I said, he followed Peter, who adapted his teaching to the needs of his hearers, but with no intention of giving a connected account of the Lord's discourses, so that Mark committed no error while he thus wrote some things as he remembered them.  For he was careful of one thing, not to omit any of the things which he had heard, and not to state any of them falsely."

The Greek distinction is exact:  Mark wrote ἀκριβῶς (accurately) but οὐ μέντοι τάξει (not, however, in order), and Papias says this is no ἁμάρτημα -- no fault.  A first- or second-century reader already knew that a true account need not be a chronological one.

Plutarch, once, says what he is doing

Plutarch wrote fifty biographies and openly flagged his own procedure exactly once.  In Cato Minor 25.5, having brought in the women of Cato's family, he adds material about them that belongs later and says so:  "All this happened later, but as I had mentioned the women of Cato's family it seemed sensible to include it here."⁹

That single sentence is the strongest ancient evidence for topical-over-chronological arrangement, and it is worth pausing on how thin the evidence is.  One self-announcement in fifty Lives.

A Jewish writer announced his compression, before Christ

Plutarch is not the only ancient author who tells his readers what he is doing to his material, and the other case is closer to the Gospels than Plutarch is.  The author of 2 Maccabees opens by explaining that he has boiled five volumes of Jason of Cyrene down into one:²⁶

"All these things, I say, being declared by Jason of Cyrene in five books, we will assay to abridge in one volume...  Leaving to the author the exact handling of every particular, and labouring to follow the rules of an abridgement...  To stand upon every point, and go over things at large, and to be curious in particulars, belongeth to the first author of the story:  But to use brevity, and avoid much labouring of the work, is to be granted to him that will make an abridgement."

Three things follow, and each one matters for this article.

Compression was a named, legitimate, self-aware procedure in Jewish Greek writing two centuries before the Gospels.  Rung 2 is not a modern rescue.

An ancient reader of a short book was told to expect delight, ease of memory, and profit -- and explicitly not completeness.  Anyone arguing that a Gospel writer must have reported everything he knew is arguing against 2 Maccabees 2:31.

And the flagging cuts against rung 4.  If a Greek-writing Jew could announce his compositional procedure in a preface, and did, then an evangelist's silence about a supposed device is evidence against the device, not for it.  This is McGrew's demand for the author's flag, and here is an ancient author supplying exactly that flag for a rung-2 procedure.

One detail of vocabulary is worth keeping.  The epitomator uses διακριβοῦν -- to handle every particular exactly -- and disclaims it, assigning it to the full-scale author.  That is the same root as Luke's ἀκριβῶς.  When Luke claims ἀκριβῶς he claims it of his investigation, not of an exhaustive narrative.

Augustine described compression in AD 400, and noticed why we can see it

Mark has Jesus curse the fig tree on one day and the disciples notice it withered on the next.  Matthew has them marvel "at once."  Augustine's explanation, four centuries before anyone had a name for it:⁴

"Matthew, after mentioning what was done on the second day...  has omitted certain facts which Mark has inserted, namely, His coming into the city, and His going out of it in the evening, and the astonishment which the disciples expressed at finding the tree dried up as they passed by in the morning;  and then to what had taken place on the second day...  he has attached what really took place on the third day."

And then the sentence that matters most for method:⁴

"These several facts Matthew has connected together in such a manner that, were we not compelled to turn our attention to the matter by Mark's narrative, we should be unable to recognise either at what point or with regard to what circumstances the former writer has left anything unrecorded in his narrative."

Augustine has stated the epistemic condition the whole field depends on.  Compression is invisible unless a second account survives.  We can see the Gospels compressing only because there are four of them, which is the same reason we can see them agreeing.

Chrysostom, around AD 390, already argued rung 1

The idea that small differences are evidence of independence is not a Enlightenment lawyer's invention either.  John Chrysostom made it in his first homily on Matthew, treating divergences "touching times or places" as small, and as proof the evangelists had not colluded.²⁷

Lucian says the same about arrangement and omission

The only surviving ancient handbook on writing history treats arrangement as the historian's craft rather than as a liberty he takes.  The historian's business, Lucian says, is "to superinduce upon events the charm of order, and set them forth in the most lucid fashion he can manage" -- his comparison is with Phidias, who did not make the ivory and gold but arranged them.²⁸  And on selection:  "Brevity is always desirable, and especially where matter is abundant … much, indeed, is better omitted altogether."²⁸

Order and omission are, in the one ancient handbook we have, part of the job description.

The Gospel cases this rung covers

Jairus' daughter.  Mark 5:23 has the father say his daughter ἐσχάτως ἔχει, "is at the point of death," and servants arrive later with the news that she has died.  Matthew 9:18 has him say ἄρτι ἐτελεύτησεν, "has just died."  Matthew has merged the father's plea and the later report into one speech and dropped the second scene.  That is compression, and Matthew does it throughout -- his Gospel is longer than Mark's overall while telling Mark's stories in fewer words.

The centurion.  Matthew 8:5 has a centurion come to Jesus.  Luke 7:3 has him send elders of the Jews, and then friends.  The principle that what a man does through an agent he does himself is an ordinary legal commonplace, and the New Testament uses it without comment elsewhere:  John 19:1 says "Pilate took Jesus and scourged him," which no one reads as Pilate wielding the whip himself.

But here is a fact usually left out, and it weakens the tidy version of this explanation.  The agency direction is not consistent.  In the centurion story Matthew drops the intermediaries.  In Matthew 20:20 he adds one -- the mother of Zebedee's sons makes the request that Mark 10:35 puts in the sons' own mouths.  A convention Matthew applied would run one way.  This runs both ways, which makes it variation rather than technique.

Loose citation.  Mark 1:2 attributes to Isaiah a quotation that opens with Malachi 3:1.  Matthew 27:9 attributes to Jeremiah words found in Zechariah 11:12-13.  Jerome raised both against himself in AD 395 and, notably, did not answer them by appealing to a convention.  On Matthew he appealed to the writer's priorities -- the apostle "made it his care to formulate dogmas rather than to hunt for words and syllables."²⁹  On Mark he simply gave up:  "Let ignorant presumption solve this nice question if it can, and I will ask pardon for being in the wrong."²⁹

And ancient copyists found Mark 1:2 uncomfortable enough to change it.  The great mass of later Greek manuscripts reads ἐν τοῖς προφήταις, "in the prophets," where the earliest text reads ἐν τῷ Ἠσαΐᾳ τῷ προφήτῃ, "in Isaiah the prophet."³⁰  A convention that nobody in antiquity found unremarkable is a weak convention, and this article says so rather than pretending the scribes were untroubled.

The strictest modern inerrancy statement already grants this rung

The 1978 Chicago Statement on Biblical Inerrancy is the high-water mark of conservative Protestant claims about Scripture.  Article XIII denies that inerrancy is negated by "the use of hyperbole and round numbers, the topical arrangement of material, variant selections of material in parallel accounts, or the use of free citations."³¹

Everything on rung 2 was conceded by the strictest party to the dispute, in writing, in 1978.  The live question is whether rung 4 is a further step beyond that line.

Rung 3: Where the Synoptics Had a Text in Front of Them

The overlap is too large for memory

B. H. Streeter counted it in 1924.  Mark has 661 verses;  Matthew reproduces the substance of over 600 of them, compressing heavily -- and still, "Matthew employs 51% of the actual words used by Mark."⁵  Luke reproduces about 350 of Mark's verses and "retains 53% of the actual words of Mark."⁵

A. M. Honoré's word counts make the scale vivid from the other direction.  Of Mark's 11,328 words, only 307 -- under 3% -- occur in Mark alone.³²  Mark is almost entirely inside Matthew or Luke or both.

Counted directly for this article:  Matthew 3:7b-10 and Luke 3:7b-9, John the Baptist's preaching, run 63 and 64 Greek words and share 60 of them, including an unbroken identical run of 24 words.⁶  The whole difference between the two passages is three words.  And this is a translated saying -- Jesus and John preached in Aramaic.  A 24-word identical Greek run is not two men remembering the same sermon.  It is one Greek text copied.

How many identical words rules out memory? Someone ran the experiment

Robert McIver and Marie Carroll had volunteers reproduce jokes and short historical accounts.  One group kept the source text in front of them.  One group handed it back before writing.¹⁰

"long sequences of 16 or more words belong exclusively to the group that retained the source and could copy from it...  The longest sequence of words in the exact order for almost all of them was fewer than 8.  None of them had a sequence of words greater than 15."

Their resulting rule:  "Any sequence of exactly the same 16 or more words that is not an aphorism, poetry, or words to a song is almost certain to have been copied from a written document."¹⁰

That is a real, checkable threshold, derived from an experiment rather than asserted.  It settles that rung 3 exists.

The half of that paper nobody quotes

Then McIver and Carroll applied their own test to the Synoptic Gospels, and the result is not what either side usually reports.

Twenty-three Synoptic parallels clear the 16-word bar.  Seven are short aphorisms or distinctive sayings, which their own experiments show transmit exactly by memory ("Foxes have holes"; "No one can serve two masters").  Six or seven more are longer distinctive sayings they judge ambiguous.  That leaves nine passages where copying is near-certain.¹⁰

And their summary of everything else:¹⁰

"Nevertheless, overall, the majority of the parallels in the Synoptic Gospels have characteristics that are more typical of a process of transmission that involves memory rather than straight copying:  the common vocabulary is found in short sequences of words;  there are changes of mood, tense, and grammatical construction;  synonyms are common;  and the passages are of different length."

So rung 3 is narrower than the confident version of it.  Copying is demonstrated for nine named passages and likely for more.  For the bulk of the parallels, the measured profile is the one their memory-transmission group produced.

One convergence is worth noting:  the passage counted by hand above -- Matthew 3 and Luke 3 on John the Baptist -- is the first of McIver and Carroll's three narrative cases.  Two methods, arrived at separately, pick out the same text.

One caution, which the article states rather than burying:  their threshold is calibrated on English prose reproduced by modern volunteers over short intervals, and they say themselves it holds only for narrative material.  Applying 16 to translated Aramaic sayings in Greek is an extrapolation, not a constant of nature.

What dependence shows, and what it does not

Where the overlap is heavy, "two people remembered it differently" is unavailable, and this article does not reach for it.  Nobody misremembers 51% of a text.

But dependence establishes that a change was made.  It says nothing about why.  Correcting Mark from the writer's own knowledge, supplementing him from another tradition, shortening him, clarifying an ambiguity -- every one of these is an editorial act, and every one is still reporting.  Rung 3 is the ceiling of rung 1.  It is not the doorway to rung 4.

Rung 4: Deliberate Reshaping, and Why It Should Be Last

What Plutarch actually does

Plutarch wrote parallel Lives that overlap heavily, so the same episode is often told twice by the same author.  Comparing them is the closest thing antiquity offers to a controlled experiment, and it produces real variation.

Casca strikes the first blow at Caesar.  In the Caesar it lands "in the neck, not a mortal wound, nor even a deep one."⁷  In the Brutus it lands "not a deep one, near the shoulder."⁸  Caesar grasps "the knife" in one and "the handle of the dagger" in the other.

Caesar bares his throat and invites anyone who pleases to strike.  In the Caesar this happens after a senate slight;  in the Antony it happens at the Lupercalia.³³  That is the cleanest instance of a gesture attached to two different occasions.

Christopher Pelling catalogued this behaviour across the Roman Lives in two articles in the Journal of Hellenic Studies in 1979 and 1980, and gave it names:  chronological compression, displacement, conflation, transfer, fabrication of context.  Michael Licona took Pelling's inventory, tested it against nine Lives and 36 pericopes, and applied it to the Gospels.⁹

The problem: Plutarch never says he is doing it

Here is the fact that should govern how much weight the whole framework carries.  In all of this, Plutarch never once calls any of it a device.  He tells the story differently in two Lives and offers no comment.  The one place he flags his procedure is Cato Minor 25.5, and what he flags there is topical arrangement -- rung 2.

"Deliberate technique rather than two sources or a slip" is an inference drawn by moderns from the discrepancy itself.  Licona concedes this in his own conclusion:  it is "quite plausible that the evangelists were aware of and made use of many of the compositional devices we inferred from Plutarch's Lives."⁹

An inference is not nothing.  But an inference from a discrepancy cannot then be used to explain that discrepancy without arguing in a circle, and the same discrepancy is equally consistent with two sources, or with carelessness.

The comparison case makes the point.  Josephus told the story of his own year commanding in Galilee twice, in the Jewish War and in the Life.  The two accounts contradict each other on about ten numbers, several proper names, whether he had one bodyguard or four, and whether he went as a general to fight or an envoy to keep the peace.  Shaye Cohen's verdict on the leading study of the problem is not "devices" but "inveterate sloppiness" -- while also noting, importantly, that Josephus "does not engage in the free invention of episodes."³⁴

Variation that wide, in one author about his own life, is what the phenomenon looks like without a theory imposed on it.

Licona's own limits

Licona is careful, and the article quotes his limits rather than only his claims.

He names five differences his method does not resolve, all in the resurrection narratives:  where Mary Magdalene first met the risen Jesus, what Jesus said to her, whether angels were present at the women's first visit, how many male disciples were present at the first appearance, and whether that appearance was in Jerusalem or Galilee.  His conclusion:  "Therefore, the tensions remain."⁹

He names a second handful in his chapter summary that "cannot be plausibly understood in light of the specific compositional devices we are considering."⁹  And the fallback he offers for them is striking:⁹

"these sorts of discrepancies could suggest the event itself was remembered while some of the peripheral details were not."

That is rung 1.  Licona reaches it last, as a residual, where this article reaches it first.  That inversion is the whole argument of this piece, stated in one page reference.

He also states the falsifiability problem himself

On whether Matthew's two blind men or Mark's one is nearer the event:⁹

"spotlighting may be occurring more often than is within our ability to recognize it, since Jesus is the main character in all four Gospels.  In a few instances, it is difficult to determine which evangelist may be employing a compositional device.  For example... one is hard-pressed to adjudicate whether the two blind men in Matthew resulted from him doubling up and conflating two healings... or that the single blind man in Mark resulted from Mark's knowledge of two blind men being present while shining a spotlight on a person known to his readers... or because Matthew preferred a version of the story that differed from Mark's."

Three incompatible explanations and no criterion for choosing among them.  And the device in question leaves no trace by construction:  a Gospel that mentions one man is evidence of spotlighting, a Gospel that mentions two is evidence of the underlying event, and no observation distinguishes them.

The same problem appears in his photograph analogy.  A photographer may add a slight haze and deepen the blue of the sky, and "the photo is a 'true representation' in its message, even if not in every detail."⁹  On that rule, no detail can ever count against an account, because any detail may be haze.

And on motive, Licona concedes what rung 4 requires and mostly cannot supply:  "The evangelists' use of these devices most often appear to have no objective other than to follow the literary conventions of their day."⁹  A change with no demonstrable purpose should have been settled at a lower rung.

Lydia McGrew's objection, which this article largely accepts

McGrew is an analytic philosopher working on the same side of the aisle, and her objection is the sharpest in the literature.  She starts by refusing the inflated vocabulary:³⁵

"'Spotlighting,' despite being made to sound like some sort of technical thing, is really just focusing on one thing, person, etc., in a scene rather than another...  it is so banal and so non-literary that it should not be used to inflate the numbers of 'literary devices' in historical works or testimonies."

Her central demand is the author's flag:³⁵

"a characteristic of the 'devices' that Licona and others claim to find in Plutarch and the gospels is that there are no such clues in the text.  This is important to remember.  The authors do not give you a 'heads-up.'  Instead they 'write as if' things are a certain way."

And her proposed alternative is, in substance, the ladder:³⁵

"Why should we think that the author was deliberately introducing data contrary to fact into his narrative?  This is an especially pressing question when the point in question is a matter of detail, timing, the name of a person, etc.  There are so many other possibilities, far more probable:  The author may have made a good-faith error...  may have obtained additional information in the meanwhile...  Two people may have remembered things slightly differently, if they were both present when an event happened."

Her verdict after working through Licona's whole book:  "in no case where harmonization seems implausible between accounts does Licona argue persuasively that the author did it on purpose.  Indeed, he usually just asserts it."³⁵

Where McGrew is also unfalsifiable.  Her first test asks whether accounts "can be harmonized without undue strain by traditional methods," and she concedes the threshold is personal:  "each person's concept of 'undue strain' will be slightly different, so (as they say), your mileage may vary."  Neither pole of this dispute has a public criterion.  That is the gap the next section tries to close.

Where a device argument does meet its burden

McGrew grants one, and it is instructive.  Tacitus gives the British chieftain Calgacus a long speech before the battle of Mons Graupius -- a man who, as Richard Burridge notes, "almost certainly didn't speak in flowing Latin parallelism.  Tacitus presumably knew that as well as we do."³⁵

That is what a demonstrated case looks like:  the author cannot have had the material, the reader can see he cannot have had it, and the purpose is plain.  Note also, as McGrew does, that even here it was not universally accepted -- Polybius attacks historians who invent speeches, which shows the practice existed and that it was contested.

What Ancient Historians Said They Owed the Truth

Rungs 1 through 4 are about explanation.  This section and the next are about the envelope:  what an ancient reader was entitled to expect from a work of history or biography in the first place.

Thucydides 1.22: the most quoted and most abused sentence in the field

Thucydides distinguishes what was said from what was done, and states a different standard for each.  On speeches:³⁶

"as to the speeches...  it was hard for me, and for others who reported them to me, to recollect the exact words.  So my method has been, while keeping as closely as possible to the general sense of the words that were actually used, to make the speakers say what, in my opinion, was called for by each situation."

The Greek phrase carrying the claim is τῆς ξυμπάσης γνώμης τῶν ἀληθῶς λεχθέντων -- the whole sense of what was truly said.  He is conceding that he cannot give the words and claiming that he gives the substance.

On deeds his standard is much higher, and this is the half that gets dropped:³⁶

"And with reference to the narrative of events, far from permitting myself to derive it from the first source that came to hand, I did not even trust my own impressions, but it rests partly on what I saw myself, partly on what others saw for me, the accuracy of the report being always tried by the most severe and detailed tests possible."

And he says plainly why that was hard:  "eye-witnesses of the same occurrences gave different accounts of them, as they remembered or were interested in the actions of one side or the other."³⁶  Thucydides had rung 1's problem and named it, in 400 BC.

A caution about how this passage is used.  Stanley Porter surveyed how New Testament scholars appeal to Thucydides 1.22 to justify the speeches in Acts, and found no stable "Thucydidean view" that can simply be invoked.  The safe inference is narrow:  Thucydides warned his readers that his speeches are not the speakers' exact words.  Anything further has to be argued.³⁷

Polybius prosecutes a colleague for inventing speeches

The clearest proof that ancient historians did not think invention was acceptable is that they attacked each other for it.  Polybius devotes much of his twelfth book to demolishing Timaeus, and his charge is precisely the fabrication of plausible speeches instead of reporting the ones delivered.  He also attacks Phylarchus for writing tragedy instead of history.

That matters for method.  If speech-invention were an accepted convention, it would not have been an accusation.

Cicero's two rules, and his own request to break them

Cicero states the standard in its most quoted form:  "who does not know history's first law, that one must not dare to say anything false?  And its second, that one must dare to say everything that is true?"³⁸

He also, in a private letter, asks the historian Lucceius to write up his consulship and to "neglect the laws of history" in his favour.  And in the Brutus he has Atticus say cheerfully that "orators are allowed to falsify in histories, so that they may say something more pointedly."³⁹

Both sides of Cicero are real, and quoting only one is how this literature gets misused in both directions.  What the pair shows is that a standard existed, that it was known, and that it was under pressure from rhetoric.

Quintilian puts the pressure plainly:  history "is very close to poetry, and is in a manner a prose poem;  it is written to narrate, not to prove."⁴⁰

The rhetorical reading, and its limits

A. J. Woodman argued in 1988 that ancient "truth" in historiography meant something closer to impartiality than to factual accuracy, and that the rhetorical core of the discipline ran far deeper than classicists had admitted.  It is an influential thesis and the article does not pretend it is fringe.

But it does not survive contact with the strongest texts.  Thucydides at 1.20-22 distinguishes report from verification.  Polybius makes truth "the eye of history."  Cicero's first law is ne quid falsi -- not "be fair," but "say nothing false."  The rhetorical pressure is real;  the claim that factual accuracy was not the standard is not what these authors say.

Josephus makes contradiction the test of a false history

One ancient statement deserves special notice because it cuts against the apologetic use of this whole section.  In Against Apion, Josephus argues that the mark of a true history is that the accounts agree, and that Greek historians' contradictions of one another prove they were not writing from records.

An ancient reader, on that standard, was not indifferent to accounts differing.  The Gospels' divergences would have been noticed and would have needed answering -- which is exactly what Origen and Augustine spend their time doing.

Lucian: "tell the thing as it happened"

Lucian's How to Write History, written around AD 166, is the only surviving ancient handbook on the subject, and it is uncompromising:²⁸

"The historian's one task is to tell the thing as it happened."

And:²⁸

"He has to make of his brain a mirror, unclouded, bright, and true of surface;  then he will reflect events as they presented themselves to him, neither distorted, discoloured, nor variable."

Lucian is satirising bad historians, which means the standard he invokes is one he expects his readers to share.  He also insists that history and encomium are different things with different rules, and mocks writers who blur them.

What a Bios Promised

Plutarch's famous disclaimer

The single most quoted sentence in this debate is Plutarch's, opening the Alexander:⁴¹

"For it is not Histories that I am writing, but Lives;  and in the most illustrious deeds there is not always a manifestation of virtue or vice, nay, a slight thing like a phrase or a jest often makes a greater revelation of character than battles where thousands fall."

What he is disclaiming there is completeness, not accuracy.  He asks his readers' pardon for epitomising, and says the anecdote reveals character better than the battle does.  In the Cimon he is explicit that a truthful subject would not want a "false and fabricated narrative," while conceding that he will under-draw his subject's faults.⁴²

That is a precise combination and it is worth stating exactly:  a bios promised truth in what it said, openly did not promise balance, and did not promise completeness or chronological order.

Suetonius says the same about order in one phrase -- he will treat Augustus neque per tempora sed per species, "not chronologically but by topic."⁴³

Are the Gospels bioi? The consensus reversed on the same evidence

This is the section where the article's reason for resting on primary texts rather than on consensus becomes visible.

In 1915 Clyde Votaw called the Gospels popular biographies.  From the 1920s K. L. Schmidt and Rudolf Bultmann established the opposite -- that the Gospels were folk literature with no biographical interest -- and that held for roughly seventy years.  In 1992 Richard Burridge tested the Gospels against Greco-Roman bioi on formal features and reversed the field.  One of his measures:  Jesus is the grammatical subject of 24.4% of Mark's verbs, against 18.7% for Xenophon's Agesilaus and 18% for Tacitus' Agricola.⁴⁴

The consensus moved twice, in opposite directions, on substantially the same evidence.  That is a reason to read the ancient authors directly rather than to cite whatever the field currently holds -- and it is a reason to be careful about how much any genre label can carry.

Because here is what Burridge's case does not do.  It is a formal argument, about shape and features.  It says nothing about accuracy.  His comparison set includes avowed encomia -- Isocrates' Evagoras announces that it is praise, Xenophon's Agesilaus the same.  Identifying the genre therefore does not license any particular liberty.  Each claimed liberty still has to be shown in named authors.

And the identification can be turned around.  Robyn Faith Walsh accepts the literary frame and draws the opposite inference:  that the evangelists were "creative writers... not biographers or historians as we understand the categories today," elite cultural producers whose genre shows craft rather than sources.⁴⁵  The same premise, opposite conclusion.  Genre labels are weaker tools than either side wants them to be.

Worth adding:  the earliest surviving outside label for the Gospels is not "biography" at all.  Around AD 155 Justin Martyr calls them ἀπομνημονεύματα τῶν ἀποστόλων, "memoirs of the apostles" -- the same word that titles Xenophon's Memorabilia of Socrates.⁴⁶  A second-century reader classed them as disciples' memoirs of a teacher.

Luke's Preface Beside Josephus and Polybius

Luke is the only evangelist who tells his reader what he is doing and why, so his preface is the closest thing the Gospels offer to a statement of method.  It is four verses long and it is worth reading against the Greek prefaces of the same period, because both what it matches and what it does not are informative.

"Since many have undertaken to draw up an account of the things fulfilled among us, just as they were handed down to us by those who from the first were eyewitnesses and servants of the word, it seemed good to me also, having followed all things closely from the beginning, to write an orderly account for you, most excellent Theophilus, so that you may know the certainty of the things you have been taught."  (Luke 1:1-4)⁴⁷

What the words claim

αὐτόπται, "eyewitnesses," is the historians' technical term -- Polybius' autopsia, seeing for oneself.  ἀκριβῶς, "carefully" or "accurately," modifies Luke's following, not his narrative.  ἀσφάλεια, the emphatic last word, is not ἀλήθεια, truth, nor a word for proof;  it is the firm ground a reader can rest on.

καθεξῆς is the one that matters most for this article, and it does not mean what English translations suggest.  It is Luke's own word -- all five New Testament occurrences are his -- and in the other four it is plainly not strict chronology.  Acts 3:24 has "all the prophets from Samuel and those who came afterwards."⁴⁷

The best demonstration is one Luke supplies himself, and it is the cleanest evidence for rung 2 anywhere in this article, because the same author wrote both passages.  Acts 10 narrates Cornelius' conversion.  Acts 11:4 has Peter retell it, and Luke labels that retelling with this very word -- Peter "began and explained it to them καθεξῆς."

Compare the two accounts and see what καθεξῆς covers.  Peter compresses a ten-verse sermon into "as I began to speak."  He drops the baptism.  Where Acts 10:23 said "some" of the brothers, Acts 11:12 says "these six brothers."  And the angel's speech is fuller in the retelling than in Acts 10:5-6.⁴⁷

So when Luke calls a narrative καθεξῆς he means an orderly account, not a chronologically exact or verbally identical one, and he shows us what he means by doing it twice himself.  That is rung 2, established from inside Luke rather than from ancient genre.

Where Luke matches Josephus

The tightest verbal parallel between Luke and any first-century writer is his address.  Luke writes κράτιστε Θεόφιλε, "most excellent Theophilus."  Josephus writes κράτιστε ἀνδρῶν Ἐπαφρόδιτε, "most excellent Epaphroditus," in the same slot in his own prefaces.⁴⁸

Josephus also states, as a rule of the craft, the two legitimate routes to an accurate history:  personal presence at the events, or careful inquiry from those who were.  He uses the same words Luke uses -- ἀκριβῶς and παρηκολουθηκότα -- in making the point.⁴⁸

Where Luke falls short of Josephus, and it should be said plainly

This is the part apologetic treatments of Luke's preface tend to skip.

Josephus' eyewitness is himself.  Luke's eyewitnesses are other people, and he does not name them.  Josephus names himself, his father, his office in the war, the fault he finds in his predecessors, the men who can vouch for him -- Vespasian, Titus, Julius Archelaus, Agrippa II -- and even his Greek assistants.⁴⁸  Luke names nobody except his dedicatee, and does not name himself.

So on the specificity that ancient historians used to establish their credit, Luke's preface is markedly weaker than Josephus'.  It claims careful inquiry;  it does not put a single checkable name behind the claim.  That is a real limitation and this article states it rather than leaning on the parallel and stopping.

Josephus is worth one more sentence for a different reason.  In Against Apion he distinguishes the facts of his history from his own rhetorical colouring of them, admits he has supplied the second, and expects his reader to discount for it.  That is direct first-century evidence that ancient writers could tell narration from embellishment and expected their readers to as well -- which is a problem for any argument that ancient audiences simply did not draw the distinction.

And where Luke is not writing like Polybius at all

Polybius opens with a justification of the whole enterprise of history, a statement of scope in years, and a thesis about causation.  Luke's preface has none of these.

Loveday Alexander argued from features like these that Luke's preface belongs to the technical or "scientific" tradition -- the prefaces of medical and engineering manuals -- rather than to the tradition of the great historians.  The thesis is often pressed harder than its author presses it.  Alexander later withdrew the word "scientific" as misleading, denied that the argument touches historicity at all, and wrote plainly:  "we shall never resolve the question of Acts' historicity by solving the genre question."⁴⁹

What the preface does not establish

Two things belong here that apologetic treatments of Luke 1:1-4 usually leave out.

The form guarantees nothing.  The Letter of Aristeas opens with the same five features in the same order as Luke -- a single long period, the word διήγησις, a named addressee in the vocative, a verb of attempting, and a clarity adverb governing the writing.⁵⁰  The Letter of Aristeas is a fiction.  Whatever a preface of this shape shows, it is not that the book behind it is true.

And a friendly witness says the preface may not amount to much.  Colin Hemer's The Book of Acts in the Setting of Hellenistic History is one of the most thorough defences of Luke's accuracy ever written.  Hemer nevertheless concedes:  "we may doubt whether the preface amounts to much more than a literary convention," and, more pointedly, "Luke does not claim in his preface to be autoptes, but only to have received material from autoptai."⁵¹

Nor is the key term settled.  Henry Cadbury wrote the fullest English study of the preface ever published and ended it by suspending judgment:  "in the mist which surrounds the exact meaning of παρηκολουθηκότι no certainty can be reached."⁵²

So the honest reading is narrow.  Luke claims careful inquiry from eyewitnesses and an orderly arrangement, for a named reader, in the ordinary furniture of Greek prose.  The preface fixes his claim.  It does not establish his delivery.  What his delivery was worth has to be settled from the checks in the rest of this article, not from four verses of self-description.

Why Ancient Numbers Inflate

Numbers are the place where ancient narrative is least reliable, and it is worth treating them separately, because the reasons have nothing to do with honesty.

The scale of the problem

Herodotus totals Xerxes' expedition at 5,283,220 people.¹¹  He also tells us exactly how the counting was done:  10,000 men were packed into an enclosure, a fence was built round them, and the rest of the army was passed through it in batches.¹¹  He then doubles the fighting total to allow for servants and camp followers.  The number is not a fabrication.  It is an honest chain of assumptions, each of which he states, compounding into an impossible result.

Hans Delbrück made this a testable question by asking what terrain and march rates physically allow.  A column of that size, he calculated, would have been leaving Susa when its head arrived at Thermopylae.  His estimate for the actual Persian force was 20,000 to 25,000.⁵³

Notice what Delbrück does not do.  He does not throw out Herodotus.  He discounts the number and keeps the narrative.  That is standard practice, and it is the model for how any ancient figure should be handled.

There is a small human detail here worth keeping.  Delbrück, in a printed lecture attacking the unreliability of ancient numbers, writes Herodotus' total as "5,100,000."  The real figure is 5,283,220.⁵³  The tendency he is describing does not stop at the ancient world.

Numbers corrupt, and we can watch it happen

Josephus estimates the Passover crowd in Jerusalem by counting sacrificial lambs:  255,600 lambs, ten diners each, therefore 2,700,000 people.¹²  The arithmetic fails by about 144,000 inside a single sentence, and his manuscripts disagree on the lamb count.

This is not a special weakness of Josephus.  Numerals in ancient manuscripts were written as letters, and they corrupt faster than any other kind of text.  The Hebrew Bible shows the same thing when the same event is reported twice:  2 Samuel 10:18 has David killing the men of 700 chariots where 1 Chronicles 19:18 has 7,000;  1 Kings 4:26 gives Solomon 40,000 stalls where 2 Chronicles 9:25 gives 4,000.

Crowd inflation is a human constant

The best-documented modern case is the Million Man March in Washington on 16 October 1995 -- with aerial photography, digital image analysis, and a federal agency counting.

A factor of 2.5 on one afternoon, in living memory, with cameras.^{54 54}  Whatever is going on with ancient crowd figures, it is not a defect peculiar to antiquity.

Even the best ancient historians knew this

Thucydides simply refuses to give the numbers at Mantinea, and says why:⁵⁵

"as to putting down the numbers of either host, or of the contingents composing it, I could not do so with any accuracy.  Owing to the secrecy of their government the number of the Lacedaemonians was not known, and men are so apt to brag about the forces of their country that the estimates of the other side were not trusted."

Catherine Rubincam audited his numbers statistically and found what one would expect of careful estimates rather than counts:  41 of his 45 casualty figures above ten are multiples of ten.⁵⁶

Where the New Testament sits

This is the part of the section that runs in the Gospels' favour, and the reason is worth stating carefully because it is a measurable difference in writing habit, not an interpretive charity.

The largest New Testament figure claiming to be a count is 5,000 -- roughly one Roman legion.  Nothing in the New Testament approaches Herodotus' 5,283,220, Josephus' 1,100,000 dead, or Pliny's 1,192,000 Gauls killed by Caesar.

Nine of twelve crowd figures are marked as estimates in the Greek, carrying ὡς or ὡσεί, "about."  Two more use μυριάδες, "myriads," the way English uses "tens of thousands" -- a quantifier of impression, not a count.¹³

And the hedging increases as the tradition develops, which is the opposite of what legendary growth predicts.  Mark 6:44 gives a flat πεντακισχίλιοι ἄνδρες, five thousand men.  Matthew, Luke, and John each add "about."⁵⁷  Matthew also adds χωρὶς γυναικῶν καὶ παιδίων -- "besides women and children" -- a clause that tells the reader his figure is a partial count.  A writer inflating a crowd does not itemise what he left out of it.

The two places where New Testament numbers strain

Honesty requires naming these rather than filing them under "ancient numbers are loose."

Acts 21:38 has a Roman officer refer to the Egyptian false prophet who led out 4,000 sicarii.  Josephus says "about 30,000."¹²  This is the one New Testament figure an outside source flatly contradicts.  It should be conceded.  Two things are worth adding without pretending they dissolve it:  Josephus contradicts himself far worse on the same episode, reporting 400 killed and 200 captured in the Antiquities;  and Luke's number is of sicarii specifically, not of the whole following.

Acts 4:4 is the harder one, and it is rarely raised.  It reports about 5,000 male believers in Jerusalem.  Joachim Jeremias' estimate of the city's resident population is 25,000 to 30,000.⁵⁸  Five thousand adult men is then something like a third to a half of all the adult males in the city.  Pilgrims at festival swelled the city enormously, and the Greek hedges the figure -- but the difficulty is real and this article does not resolve it.

One more finding cuts against a common apologetic move.  Josephus' 2,700,000 Passover pilgrims and Tacitus' 600,000 besieged are often cited as two independent witnesses to a crowded Jerusalem.  Jeremias argues that Tacitus took his 600,000 from Josephus' own account of corpses carried out of the gates and mistook it for the population.⁵⁸  If he is right, that is one bad number, not two.

Genealogies as an Ancient Genre

The two genealogies of Jesus are the most-cited contradiction in the Gospels, so they are a fair test of how much genre knowledge actually buys.

What a genealogy was for

Robert R. Wilson's study of genealogies in the ancient Near East states the finding plainly:  "genealogies are not normally created for the purpose of conveying historical information...  the makers of genealogies are not historians."⁵⁹  They argue for status, legitimacy, or the shape of history.

Three conventions follow, and all three are demonstrable inside the Hebrew Bible, not imported from outside it.

Telescoping.  Ezra 7:1-5 traces Ezra's descent and drops six names that 1 Chronicles 6:6-10 supplies.  Ruth 4:18-22 covers roughly 850 years in nine begettings.  "Son of" regularly means descendant.

Schematic numbers.  Matthew announces three sets of fourteen (Matthew 1:17) and gets there by omitting Ahaziah, Joash, and Amaziah between Joram and Uzziah -- four generations that 1 Chronicles 3:11-12 records.  This is not a slip a reader is meant to miss;  Matthew tells you the scheme.

Generation counts that do not match.  Exodus 6:16-20 gives Moses two generations from Kohath;  1 Chronicles 7:23-27 gives Joshua ten from Ephraim, over the same span, for two men who stood at Sinai together.  The phenomenon the Gospels are accused of is already in the Torah.

Greeks and Romans said pedigrees were faked

The conventions are real, but "it was conventional" is not a blanket excuse, because the ancients themselves distinguished honest lists from padded ones.

Cicero on Roman funeral eulogies:  they made "the history of our affairs more faulty," because "many things were written in them which never happened -- false triumphs, extra consulships, false genealogies."³⁹  Livy says the same about family records.⁶⁰  Plutarch opens the Numa by noting that the genealogies had been tampered with.

And Suetonius reports that when someone tried to trace Vespasian's family to a companion of Hercules, the emperor laughed at it.⁶¹

So an ancient reader knew pedigree inflation existed and could tell it was happening.  Genre does not confer immunity.

What the conventions explain, and what they do not

They explain Matthew's omissions cleanly.  They explain "son of" spanning generations.  They explain the 3×14 scheme -- though it is worth noting that the list contains 41 distinct names, not 42, and Marshall Johnson argues the three kings may have dropped out first, with the fourteens noticed afterwards:  "accident was turned into virtue."⁶²

They do not explain the two things a skeptic actually presses.

The fork.  Matthew makes Joseph's father Jacob;  Luke makes him Heli.  Matthew descends from David through Solomon;  Luke through Nathan.  The lines run apart for twenty generations.

The counts.  Matthew has 27 generations from David to Jesus.  Luke has 42.  Between Zerubbabel (about 520 BC) and Jesus, Matthew has 11 and Luke has 20.

The earliest surviving harmonisation is Julius Africanus, about AD 220, preserved by Eusebius:  both lists are Joseph's, joined by a levirate marriage.⁶³  Augustine took both as Joseph's and used adoption instead.  Notably, neither ancient reader entertains the idea that Luke gives Mary's line.  That solution first appears explicitly with Annius of Viterbo around 1490 and with Luther,⁶² and the Greek of Luke 3:23 names Joseph, not Mary.

The objection is not modern either.  Origen concedes in writing that in the genealogies "there are certain points which occasion some difficulty even to Christians."⁶⁴

Honest verdict.  Ancient convention clears away every standard objection to Matthew 1 except the fork and the counts.  The most economical remaining account is that Matthew gives a schematised royal-succession list and Luke a fuller line of descent, joined by a levirate or an adoption at Joseph.  That is coherent, it is ancient in direction, and it has no evidence behind it -- which is exactly what Africanus admitted about his own solution.  Nothing here belongs above rung 2.

How Good Are the "Good" Historians?

Apologists routinely say the Gospels should be treated like Tacitus or Thucydides.  That is the right instinct.  It is worth knowing what it actually buys, because Tacitus is not what a modern reader imagines.

Tacitus against a bronze inscription

In 1996 a bronze tablet was published carrying the Roman Senate's own decree about the trial of Gnaeus Calpurnius Piso, dated 10 December AD 20 -- the very episode Tacitus narrates in Annals 3.10-18, ninety years later.

Tacitus comes out well.  Names, offices, the procedure, the halved estate, the change of praenomen, Plancina's remission, the consul -- all correct.¹⁵

He also compresses a trial that ran for months into what reads like days, which Richard Talbert calls "to some degree historian's license," and he leaves out material that cuts against his reading of Piso's motives.

The response of classicists to a demonstrated Tacitean error is the model this article recommends.  Talbert:  the comparison "should be taken to confirm his high reputation as a historian rather than lower it."  Ronald Syme, whose appendix catalogues roughly fourteen demonstrable errors in the Annals -- wrong legion, wrong rank, wrong war, wrong ages, a name given three ways -- rebukes scholars who try to harmonise them away as "generously transferring to the historian their own inerrancy."¹⁴

Every one of Syme's fourteen is peripheral.  None has cost Tacitus his standing.

Three good authors, one fire, three levels of certainty

On the Great Fire of Rome in AD 64, Tacitus writes that whether it happened by accident or by the emperor's treachery is uncertain -- forte an dolo principis incertum.  Suetonius says flatly that Nero set the city alight.  Dio describes a secret arson squad.

Confidence grew as the distance from the event grew.  That is a pattern worth remembering whenever a later, more certain source is preferred to an earlier, more hesitant one.

Josephus against himself

The single best control we have is an author telling the same year of his own life twice.  Josephus' Jewish War and his Life both cover his command in Galilee, and they contradict each other on about ten numbers, several proper names, whether he had one bodyguard or four, and whether he went to fight a war or keep the peace.

Shaye Cohen's verdict is "inveterate sloppiness."  But he adds a qualification that matters just as much:  Josephus "does not engage in the free invention of episodes."³⁴

That combination -- wide variation in detail, no invented events -- is a useful benchmark.  The variation between the Synoptics is markedly narrower than the variation between Josephus and himself.

Where Luke fails a check

Acts 5:36 has Gamaliel, speaking around AD 33, refer to a rising by Theudas as already past, and place it before the census of AD 6.  Josephus dates Theudas' rising under the procurator Fadus, about AD 45.⁶⁵

The standard defences are that Luke means a different Theudas, or that Josephus erred.  Both are possible.  Neither has evidence.  This article concedes the case, in the same way it concedes Syme's fourteen for Tacitus, and for the same reason:  an author with a documented error is still a usable historian, and pretending otherwise is what makes apologetics untrustworthy.

What "historian of the first rank" actually meant

The most-quoted sentence in popular apologetics on this subject is William Ramsay's, that Luke should be placed "among the historians of the first rank."  It is real, and it is on page 4 of St. Paul the Traveller.  The next sentences are almost never quoted:¹⁶

"The first and the essential quality of the great historian is truth.  What he says must be trustworthy.  Now historical truth implies not merely truth in each detail, but also truth in the general effect, and that kind of truth cannot be attained without selection, grouping, and idealisation."

Ramsay's compliment is a rung-2 compliment.  He is praising Luke for the same selection and arrangement this article puts on rung 2, and he explicitly says detail-level accuracy is not what "first rank" means.

A similar trimming happens to A. N. Sherwin-White, whose remarks on the historicity of Acts are a staple of apologetic quotation.  In the same pages he calls Acts "no less of a propaganda narrative than the Gospels, liable to similar distortions," and notes that the Gospels have less Greco-Roman confirmation than Acts does.⁶⁶

Quoting these men accurately still leaves a real argument.  It leaves a smaller one than the popular version.

The honest expectation

Putting the checks together, here is what a careful first-century author can be expected to get right and wrong:

• Right: events, persons, public offices and titles, the political framework, geography.
• Silently compressed: chronology and order.
• Sometimes wrong: a minority of secondary names and connections.
• Unreliable: numbers, often by a factor of two or more.
• Shaped: motive and characterisation.

Synoptic variation sits comfortably inside that band.  That is a real result, and it is the strongest general claim this article makes.

When Does "It Is a Convention" Stop Being Testable?

The complaint that runs through this whole debate, from both directions, is that the explanations are unconstrained.  Ehrman says harmonisation "requires so much speculation and fancy interpretive footwork" that it eventually collapses.⁶⁷  McGrew says device attribution is asserted rather than argued.  They are making the same complaint about opposite parties.

So the useful question is not who is right.  It is what a disciplined version of either move would look like.

Four conditions, each closing an escape route that someone in this literature actually took.

1. The convention is named by an ancient author. Described as a thing writers do, in a non-Christian source of the period -- not inferred by a modern from a discrepancy. This closes Licona's own "devices we inferred from Plutarch's Lives."
2. It is applied symmetrically. The same explanation would be granted to Josephus, Plutarch, or Tacitus in the same case, and the arguer says so. This cuts both ways: a skeptic who calls a Gospel difference a contradiction and a Plutarch difference a variation fails it too. McGrew's version: "Even if you shy away from saying that John made some minor error... why shy away from saying that Tacitus did so?"³⁵
3. A residue is named in advance. The framework must leave some differences counting as errors, and the arguer must say which. Peter Williams' closing sentence -- "I do not know of any that cannot possibly be resolved"⁶⁸ -- sets a bar nothing can clear, since a possible harmonisation always exists. Licona passes this condition: he names five, then five more.
4. A falsification condition is stated before the case is examined. What observation would make the arguer say "this one is an error"? This closes the photograph analogy, on which no detail can ever count against an account.

Two notes belong with the rule.  First, it constrains skeptics as much as apologists:  Ehrman does supply a criterion -- a harmonisation must not require asserting something no account states⁶⁷ -- which is real and statable, and which ordinary witness practice arguably fails, since "he mentioned one of the two men present" is exactly how testimony works.  Second, the rule does not settle truth, only method.  A case can pass it and be wrong, or fail it and be right.

The cases, sorted

Five pass, seven undecided, six fail.

The convention argument is not empty.  It does real work on compression, arrangement, and citation, which between them account for most of the differences an ordinary reader meets.  It does no work at all on the six hardest cases, and on two of those -- the hour of the crucifixion and the day of the Passover -- its leading proponent's own explanation invokes a convention no ancient author names.

The case that shows what a real vindication looks like

One entry in that table deserves expansion, because it is the shape of an argument that could have falsified Luke and did not.

Luke 3:1 dates John the Baptist's ministry by listing rulers, including "Lysanias tetrarch of Abilene."  The only Lysanias in Josephus before that date is a king of Chalcis executed in the 30s BC -- sixty years too early, and titled king, not tetrarch.  This was for a century a standard example of Lukan error.

Then an inscription from Abila turned up:  a dedication by "Nymphaios son of Abimmes, freedman of Lysanias the tetrarch," made "for the safety of the lords Augusti."⁶⁹  The plural "Augusti" dates it, because the first pair of joint Augusti were Tiberius and his mother Livia, who took the title at Augustus' death in AD 14 and died in AD 29.

The decisive endorsement comes from a hostile witness.  Emil Schürer -- whose census excursus is the standard critical case against Luke 2:2 -- concluded:⁷⁰

"The evangelist Luke is thoroughly correct when he assumes (Luke iii. 1) that in the fifteenth year of Tiberius there was a Lysanias tetrarch of Abilene."

Two honest qualifications.  Schürer records and rejects a dissent from Wieseler, who read the two Augusti as Augustus and Tiberius, which would push the stone earlier and dissolve the argument.  And the inscription proves only that a tetrarch named Lysanias existed at Abila in that window;  it does not date his rule to precisely AD 28.  Stated at that strength, it stands.

The residue, named and not rescued

Six cases survive every rung.  This article does not solve them, and lists what each costs.

The hour of the crucifixion.  Mark 15:25 has Jesus crucified at the third hour;  John 19:14 has him still before Pilate at the sixth.  The popular harmonisation -- that John uses Roman civil time from midnight -- is rejected by Licona himself, on the ground that the trial would then have had to begin before 6 a.m., and Roman court business typically started at the third hour.⁹  Bauckham's memory finding puts this in the right class -- clock times are what recollection carries worst -- but that explains how an error could arise, not that there is none.

The day of the Passover.  The Synoptics make the Last Supper a Passover meal;  John has the leaders avoiding defilement so they can still eat the Passover after Jesus is condemned.  Licona's own preference is that John moved the day for theological reasons.  McGrew's counter is the strongest thing on this case:  John's trial scene interlocks with Luke's in small unplanned ways that argue he is reporting rather than composing, which makes deliberate relocation in the middle of that same scene hard to credit.⁷¹

Judas' death.  Matthew has him hang himself and the priests buy the field;  Acts has him fall headlong and burst, and buy the field himself.  By the early second century Papias had a third version involving a chariot.³  Three accounts by AD 130 is what an unstable tradition about a lurid peripheral detail looks like.  The rope-broke harmonisation is possible and unsupported.

Matthew 27:52-53, the tombs opening and the risen saints appearing in Jerusalem.  Licona favours reading these as apocalyptic imagery rather than reportage, and says so;  the resulting dispute was inside conservative Christianity, not between it and its critics.⁹

The Quirinius census.  Luke 2:2 dates the nativity registration to Quirinius' governorship;  Josephus dates Quirinius' census to AD 6/7 and says it happened because Judaea had just been annexed.  The structural problem is not only the date:  a Roman census of an independent client kingdom under its own king is not something Roman practice provides for, and none is attested.  Both main defences -- an earlier governorship, or reading πρώτη as "before" -- are contested on the evidence and the grammar.  This is not a case for the convention argument at all.  Nobody invokes a literary convention;  both sides argue about facts and Greek.  Bringing genre into it is a category error.

Theudas, above.

Common Objections

Objection 1: "This is just how memory works. Why call any of it a device?"

Largely right, and this article is built on the concession.  Three of the four rungs describe ordinary human behaviour, and the fourth is reserved for cases where the first three fail.

But the objection has one real limit, and stating it is what keeps the article from being special pleading.  Where Matthew reproduces 51% of Mark's actual words, and where two Gospels share 24 identical Greek words in a row in a translated saying, memory is not the explanation.  One writer had the other's text.  In that documented handful of passages -- McIver and Carroll identify nine as near-certain -- a difference is an editorial act.

What still does not follow is that the change was a literary device.  Correction, supplementation, compression, and clarification are all editorial and all reporting.  Dependence shows that a change was made, never why.

Objection 2: "Memory science cuts both ways"

It does, and this article quotes it doing so.  Talarico and Rubin show that confidence is uncorrelated with accuracy on peripheral detail.  Hirst's ten-year data show that wrong details tend to be repeated rather than corrected.  Loftus and Palmer show that suggestion can insert a detail that was never there.

Two limits on the counter-argument.  First, in every one of those studies the event survives:  no subject in Loftus and Palmer denied the crash, and Hirst's subjects remembered the crash sites at 88% accuracy after three years while their memory of their own feelings fell to 42%.  Second, Ehrman himself grants the structure -- "much of the gist is correct" -- and rests his case on the transmission chain instead.  That is a different argument, about whether the Gospels rest on witnesses at all, and it is not settled by memory data in either direction.

Objection 3: "The genre defence proves too much. Any difference can be called a convention"

This is the strongest objection in the set, and it is why the article proposes a rule and then applies it to eighteen cases with a scoreboard rather than a verdict.  Six of the eighteen fail.  Naming the failures in advance is the only answer to this objection that is worth anything.

Objection 4: "Ancient writers had a looser idea of truth, so none of this matters"

Partly true and often overstated.  The looseness is real for speeches, order, selection, and numbers.  It is not what the ancient theorists say about events.  Cicero's first law is ne quid falsi -- do not say anything false.  Lucian's rule is to "tell the thing as it happened."  Polybius attacks a colleague specifically for inventing plausible speeches instead of reporting delivered ones, which is only an accusation if the practice was disreputable.

A. J. Woodman's stronger thesis, that "truth" in ancient historiography meant impartiality rather than accuracy, does not survive those texts, though the rhetorical pressure he documents is real.

Objection 5: "If one evangelist got a detail wrong, the Bible is not inerrant"

This article does not answer that question, and it should not pretend to.  What it can say is what follows and what does not.

Rung 1 does imply that on occasion an evangelist reported a peripheral detail as he or his source remembered it, and was wrong.  That is a cost to strict inerrancy about incidentals.  The 1978 Chicago Statement already conceded round numbers, topical arrangement, variant selection, and free citation;  whether it can absorb "one blind man or two" is a theological argument, and readers should notice that the dispute over Licona's work was conducted mainly inside conservative Christianity.

What it costs the historical case is nothing.  Courts and historians both treat peripheral inconsistency as normal and core agreement as what matters, and neither institution is being generous to Christianity when it does so.

Objection 6: "You are just picking whichever explanation is convenient"

The ladder is an ordering rule, and an ordering rule is only honest if it can be broken.  So here is what would break it:  show that a difference is a deliberate change, made by an author who knew the other version, for a purpose he states elsewhere.  That is what McGrew grants for Tacitus and Calgacus, and it is what nobody has yet shown for a Gospel case.  Matthew 27:52-53 and John's Passover are the two candidates.  Neither currently meets it.

Verdict

On the ancient evidence:  ancient historians and biographers did claim to report events truthfully, did not claim to report speeches verbatim, openly did not promise chronological order or completeness, and were routinely unreliable about numbers.  That envelope is well attested from named authors in their own words.

On the compositional-devices framework:  partly justified and heavily over-applied.  The phenomena Pelling identified in Plutarch are real.  The inference that they are deliberate techniques is an inference, conceded as such by its leading defender, and Plutarch flags his own procedure exactly once in fifty Lives.

On the Gospels' differences:  most sit on the first two rungs, which need no special pleading.  A documented handful fall inside passages of demonstrable literary dependence, where the difference is editorial and the motive unknown.  A residue of six resists every rung, and this article names them rather than resolving them.

On the comparison with ancient historians:  it holds, and it buys less than apologists usually claim and more than skeptics usually allow.  Tacitus has fourteen catalogued errors and remains a first-rank source.  Josephus contradicts himself about his own life more widely than the Synoptics contradict each other.  The Gospels' numbers are hedged where ancient numbers are not, and Licona's own finding is that the evangelists edit their sources less freely than the ancient authors they are compared with.

The honest summary is not that ancient genre excuses the Gospels' differences.  It is that the differences look like what honest, partly dependent accounts of the same events look like -- and that six of them do not, and are still waiting for an answer.

Sources