Oral Tradition and Memory

"Of one of the most popular his own account remains to show how fact and fancy mingle in such current lore." -- Rena Hogg on a village story about her father, 1914¹

"His testimony had much truth in it, but not at the level of "gist."  It was true at a deeper level." -- Ulric Neisser on John Dean's Watergate testimony, 1981²

Summary

The Gospels were written decades after Jesus died.  Before that, what people knew about him passed through memory, retelling, teaching and perhaps a few written notes.  This article asks what that passage could preserve, what it would change, and how the surviving texts let a reader tell the difference.  Who wrote each Gospel and when are argued in Who Wrote the Gospels? and When Were the Gospels Written?;  which Gospel used which is argued in The Synoptic Problem.  This article takes their conclusions as inputs.

The surviving texts show both preservation and change.  Paul and the Gospels share teachings and central claims about Jesus, while some prayers and formulas retain substantial identical Greek wording.  Narratives also vary in wording, order, participants and interpretation.  These patterns fit a mixture of retelling, repeated teaching and written editing.  They do not identify the transmission route of every passage or prove that every shared claim is accurate.  Historical usefulness must be assessed claim by claim;  memory science alone neither verifies nor disproves a miracle.

The main findings, each argued below:

1. The best test case of a modern "controlled" oral tradition shows both stability and legend. Kenneth Bailey's leading example, stories about the Scottish missionary John Hogg in Egypt, can be checked against a printed source. Hogg's daughter Rena recorded the village version of his "robbers" story before 1914 and printed beside it the account of the two men who were there. In about forty years the village version had acquired a robbery, a gold watch, an all-night Bible lesson and the conversion of the whole band, none of which appears in the participants' account. The men were most likely melon-field watchmen, and the eyewitness who called them robbers preached the story as a sermon within days.¹¹ Bailey's later printed version resembles this village tale, but dependence on Rena's book could explain some of that similarity.³ Stability alone does not establish a story's accuracy.
2. Memory studies distinguish tasks, not one universal reliability rate. Spontaneous recollection often loses exact wording. The particulars of a single episode, such as where one was, who spoke first and what was said, drift and are often confidently wrong: of 44 students asked about the 1986 Challenger news 32 to 34 months later, 11 gave answers inconsistent with their first-day reports on every major detail of their own circumstances.⁴ In one case study, John Dean misreported almost every word of a taped meeting yet, apart from matters affecting his own role, "was right about what had really been going on in the White House."² No experiment yields a percentage chance that a given Gospel story happened.
3. Ancient teaching supplied real tools, and their use by Jesus' followers has to be inferred. Memorisation was normal in ancient schooling, and teachers' sayings were taken down in notes and circulated (Quintilian, Arrian). Paul uses the vocabulary of handing on tradition (1 Cor 11:23; 15:3) and spent fifteen days with Peter about three years after his conversion (Gal 1:18). But the rabbinic rules for repeating a teacher's exact words were compiled around AD 200 and later, and no early Christian source mentions notes taken during Jesus' life. Hezser estimates that at least nine in ten Jews of Roman Palestine could at most write their names; this is a historical reconstruction, not a census.⁵
4. The Greek comparisons constrain possible mechanisms. Matthew and Luke share 62 of Matthew's 77 words in John the Baptist's preaching, including a run of 24 identical words: strong evidence of shared Greek wording, with written dependence a strong explanation. Paul's Last Supper words and Luke's share a 10-word run; Paul's and Mark's share at most 3, and Mark's and Luke's share 5. Repeated worship and written use could both contribute to these patterns. Mark and John tell the anointing at Bethany with no run longer than 5 words, yet share four words found nowhere else in the New Testament: distinctive shared vocabulary, with literary dependence, oral tradition and common subject matter needing assessment.⁶
5. Aramaic traces prove an Aramaic-speaking stage, not an eyewitness. Paul's Greek letters to Greek churches carry abba, maranatha and Cephas, so Aramaic prayer words reached Greek congregations by the 50s. A plausible Aramaic retroversion of a saying shows that it can be put into Aramaic. It does not identify the speaker, show that it was Jesus, or date the saying.⁷⁸
6. The models disagree less on the data than on the default. Form criticism's real strength is that the Gospel units were joined by editors; its "laws" of how tradition grows fail when tested.⁸ Gerhardsson's rabbinic analogy is later than the events and needs independent support; his allowance for alteration prevents variation alone from refuting it. Bailey's model is an analogy, not first-century evidence. Richard Bauckham argues that named eyewitnesses stayed available as authorities; Bart Ehrman argues that most tellers were not witnesses. Both allow preservation and distortion, but the width of the reliable "gist" remains disputed.⁹¹⁰

Verdict.  Early Christian sources attest teaching, repeated formulas and contact between leaders within the first generation.  The Gospel comparisons show shared wording alongside substantial variation, compatible with both repeated performance and written editing.  This supplies reasons to investigate the texts as historical sources, but no general transmission model identifies an error-free core or establishes the history of every episode.  Each claim needs assessment of its sources, independence, chronology and external evidence.  Memory science alone cannot settle a miracle.

Part 1: What each model claims, and what it rests on

Five accounts of the pre-Gospel period are in play.  Each is stated here from its own author, with the kind of evidence it actually uses.  The point of the list is the evidence column, not the names.

A sixth approach, "social memory," treats every group's memory of its past as shaped by its present needs.  Bart Ehrman's Jesus Before the Gospels (2016) uses it, together with experimental psychology, to argue that the Gospels are "memories of memories of memories."⁹  Dale Allison's Constructing Jesus (2010) uses the same psychology to argue for trusting "general impressions" over single sayings and scenes.¹¹

Form criticism: what it actually showed

Bultmann defined the task as recovering "the origin and the history of the particular units and thereby to throw some light on the history of the tradition before it took literary form."¹²  His starting point, taken from Julius Wellhausen and K. L. Schmidt, was that "the tradition consists of individual stories or groups of stories joined together in the Gospels by the work of the editors."¹²  That claim has held up because it can be checked.  Matthew and Luke move Mark's units around and supply their own links between them, which is what the observed differences in order show, and the links in Mark ("immediately," "again") are general and replaceable.

The more famous claim is weaker.  Bultmann argued that early Christian prophets spoke "in the name of the ascended Lord" and that the church did not separate such words from remembered ones, because "the dominical sayings in the tradition were not the pronouncements of a past authority, but sayings of the risen Lord."¹²  His examples are the risen Christ's sayings in Revelation 3:20 and 16:15, the "word of the Lord" in 1 Thessalonians 4:15-17, and a saying about miraculous vines that Irenaeus quotes from Papias.¹²  The last supplies a case where a claimed eyewitness chain accompanies a saying outside the four Gospels.  Irenaeus says that "the elders who saw John, the disciple of the Lord, related that they had heard from him" that Jesus taught that each vine would have "ten thousand branches," each grape yielding "five and twenty metretes of wine."¹³  The same picture appears in the Jewish apocalypse 2 Baruch, where each vine has "a thousand branches."¹⁴  The shared imagery places the saying within Jewish apocalyptic tradition.  It does not establish who borrowed from whom, or that Jesus could not have used such imagery.  None of the four Gospels contains it.  A claimed chain alone cannot verify the attribution, and resemblance to another text alone cannot disprove it.

Bultmann did not treat everything as invented.  Of the question about paying tribute to Caesar he wrote, "There is no reason, in my view, for supposing that this is a community product."¹²  Paul's letters also show an early concern to distinguish different grounds of authority.  In 1 Corinthians 7 he gives a ruling on divorce with "not I, but the Lord" (7:10), then a further ruling with "I, not the Lord" (7:12), and on virgins says "I have no command of the Lord" (7:25).  An early teacher who claimed the Spirit (7:40) still marked some advice as his own and declined to invent a command where he had none.  That is evidence of restraint.  It is not a direct test of Bultmann's narrower claim that words attributed to the risen Lord could merge with remembered earthly sayings.

The "laws" of oral growth that form criticism used to sort early from late did not survive testing.  E. P. Sanders checked them against the Synoptics and later Christian texts in 1969 and concluded:  "There are no hard and fast laws of the development of the Synoptic tradition.  On all counts the tradition developed in opposite directions.  It became both longer and shorter, both more and less detailed, and both more and less Semitic."⁸  That verdict binds both sides.  It blocks automatic dating by added names, length or Semitic colouring.  A feature may still contribute to a particular argument when its history and alternatives are independently established.

Controlled tradition: stated, and checked against its dates

Gerhardsson's strongest statement of the model is this:  "In all likelihood the early church arose around the circle of Jesus' closest disciples and it is equally probable that from the beginning they had learned the logia and parables by heart."¹⁵  Its method comes from the rabbis, who repeated a teacher's words until the pupil knew them:  "Among the rabbis we can see how evident it was that a teacher would repeat the texts until his pupils knew them by heart;  four repetitions seems to have been common."¹⁵  The model does not require identical Gospels.  Gerhardsson himself wrote that "Careful oral transmission—with texts learned by heart—thus does not exclude alterations in texts."¹⁵

Two facts limit it.  The rabbinic evidence is written down from about AD 200 onward, and the rule against writing oral teaching (b. Gittin 60a) "probably originated only in the Amoraic period (i.e., from the third century) and should not be projected back onto the first century," as Bauckham notes.¹⁰  The variation rules out identical transmission of the entire text, but Gerhardsson allows alteration and editing.  Bauckham reports the common objection that the model is "too rigid to explain the actual extent of variation in the Jesus traditions as we can observe them in the Gospels," then warns that Gerhardsson's position "is easily exaggerated."¹⁰  Variation alone therefore does not refute Gerhardsson's qualified model;  the question is which parts were memorised, with what controls, and how later authors used them.  Gerhardsson's evidence for Jesus' own practice is inference, in his own words:  "it seems clear to me that Jesus presented such a saying two or more times."¹⁵  The later analogy makes the practice plausible.  It does not document Jesus using that particular method.

Bailey, Dunn and Bauckham: a stable core and who guarded it

Bailey sorted oral material by how much change a community allows.  Proverbs and poems allow none;  parables and stories about people "important to the identity of the community" keep their core;  jokes and news may change freely.¹⁶  Dunn built on this:  the heart of a story, often a short exchange of words, stays fixed while its setting is retold.  His best example is the centurion's servant, where "the heart of the shared memory is the actual encounter between Jesus and the centurion, the dialogue between them, where, noticeably, the words are in close agreement."¹⁷  Part 6 measures that passage.

Bauckham adds a claim about people:  "the period between the "historical" Jesus and the Gospels was actually spanned, not by anonymous community transmission, but by the continuing presence and testimony of the eyewitnesses, who remained the authoritative sources of their traditions until their deaths."¹⁰  The evidence for that claim, the named witnesses and Papias, is weighed in Who Wrote the Gospels?.  For the memory question, what matters is how far Bauckham himself takes it.  He argues that the gist of an eyewitness memory and the gist of an oral tradition "can readily coincide," and that the step from one to the other "need not entail a significant decrease in reliability, though of course this is possible."¹⁰  In his second edition he adds that "it was never my intention to claim that eyewitnesses always get everything right or that the eyewitnesses behind the Gospels always got everything right."¹⁰  The claim is that the gap could be spanned well, not that it was spanned without loss.

Ehrman and Allison: the critics' case

Ehrman's case is not that memory is worthless.  He writes that "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," and asks "just how broad does a memory have to be in order to be considered a gist memory?"⁹  His argument rests on the chain:  stories about Jesus "were circulating even during his lifetime," they "were not being told only by eyewitnesses," and if a wild version reached a witness who corrected it, "there is no guarantee that everyone will hear the correction."⁹

Allison reaches a middle position from the same psychology.  "If general impressions are typically more trustworthy than details, then it makes little sense to reconstruct Jesus by starting with a few of the latter."¹¹  He adds that "the historicity of most—not all—of the events associated with Jesus and the origin of most—not all—of the sayings attributed to him will always fall woefully short of demonstration."¹¹

These authors allow both preservation and distortion, but their claims differ substantially.  Bauckham emphasises eyewitness authority;  Dunn emphasises repeated performance;  Ehrman includes invented episodes among distorted memories;  Allison builds from recurrent impressions.  Their concessions do not establish a shared, precisely bounded reliable core.  The comparisons below test particular claims without assuming that agreement among scholars settles them.

Social memory: what its advocates actually claim

"Social memory" draws on the French sociologist Maurice Halbwachs (1877-1945).¹⁸  Alan Kirk and Tom Thatcher's 2005 collection and Rafael Rodriguez's 2010 monograph develop its application to early Christianity.  It is often cited, by both sides, as if it meant that groups invent their past.  Its New Testament advocates say something narrower.

They agree with the form critics on one point.  Kirk and Thatcher write:  "Consistent with the form-critical model, social memory theory views present social realities as decisive factors in the constant rearticulation of a community's salient past".  They break with the form critics on the next:  "But social memory theory departs significantly from the form-critical legacy by refusing to authorize any sharp distinction between memory and tradition."¹⁸  Their objection is to a clean separation of untouched memories from later invention, rather than to all historical judgement.  The idea that Gospel material can be "analytically and cleanly separated into two piles" is, in their words, "predicated upon questionable views of the operations of memory and tradition."  The same paragraph states the other half:  social memory analysis "remains convinced that the commemorated past bore upon the traditioning activities of the early communities in a far more thoroughgoing manner than the form critics envisioned."¹⁸  In his introduction Kirk adds that "commemorative projects are dependent upon the core realities they take up," and that remembered persons and events are "to a significant extent resistant to whimsical make-overs into the image of shifting ideological forces."¹⁸

So the claim is that every telling of the past is shaped by present needs, and that the past also limits what those needs can make of it.  It is not a claim that memory is false or that the past is out of reach.  Barry Schwartz, the sociologist whose work on how Americans remembered Abraham Lincoln Kirk cites throughout his introduction, puts it bluntly:  "Stable images of the past are often, but not always, demonstrably true images.  Sometimes false ideas are transferred across generations and accepted as if they were true; sometimes true ideas are rejected as if they were false."¹⁹  Eric Eve, reviewing the competing approaches, calls the strong constructionist version, in which the past is a total fabrication, "implausible" because "it ignores both the extent to which the actual past remains immanent in the present ... and almost certainly underestimates the resilience of individual memory."²⁰

Schwartz offers one test case.  After Lincoln's assassination in 1865, his law partner William Herndon collected recollections from people who had known Lincoln in New Salem, Illinois, in 1831-1837:  "Herndon captured 30 years of Lincoln oral tradition, just as Mark captured approximately the same number of years of Jesus oral tradition."  James G. Randall dismissed the story of Lincoln's romance with Ann Rutledge as legend, because it reached Herndon at third hand and the papers are full of errors elsewhere.  Douglas Wilson later went back to the original documents and "found all respondents agreeing that Lincoln courted and planned to marry Ann and grieved with unusual intensity after her death."  Schwartz concedes that shared belief need not be true, but argues that a bar set at proof beyond doubt would leave Lincoln's young adulthood blank.¹⁹  The case is one disputed episode, and agreement among informants is also what a shared legend would produce.  It illustrates a dispute about how to weigh delayed testimony.  Schwartz's report of agreement, without an independent check of the underlying documents, cannot establish either the romance or a general reliability rate.

Rodriguez applies the approach to a Gospel passage:  Luke 4:16-30, where Jesus reads Isaiah in the synagogue at Nazareth.  His observation is about Luke, not about Jesus.  Even here, "where we can be especially confident that Luke should be looking upon a written text (an Isaiah scroll), he appears unconcerned to replicate that text, either verbally or structurally."  The quotation follows Isaiah 61:1-2 closely but leaves out phrases and inserts a line from Isaiah 58:6.²¹  So a Gospel writer could retell even a written source freely, the way an oral performer would.  The limitation is stated by Rodriguez himself.  He asks how Jesus' remembered programme constrained Luke, "irrespective of whether Jesus ever said anything like Luke 4.16-30 in any context like that of a rural Galilean synagogue."²¹  The method sets the historical question aside.  It cannot by itself say whether the Nazareth scene happened.

Social-memory analysis asks how communities use their past;  it is not a numerical test of a report's truth.  Rodriguez explicitly says that "the stability of social memory does not ensure its accuracy," and that stability "inheres at the level of meaning rather than on a 'textual' level."²¹  Kirk and Thatcher say the relation of past to present "can only be assessed empirically, on a case by case basis."¹⁸  An application must therefore identify the evidence for continuity and change in that particular case.  Early attestation, demonstrable dependence, independent corroboration and external records can help.  A story's usefulness to a later community does not by itself make it invented, while agreement among related sources does not independently verify its content.  Part 6 examines shared wording without treating it as proof of historical independence.

Part 2: What ancient practice can and cannot establish

The question here is narrow.  It is not whether the ancient world had good memories, notes and books.  It did.  It is which of those practices can be shown among Jesus' followers between about AD 30 and the first Gospel, and which are only possible.  Each witness below carries its date, because a rule written down in AD 200 is not a description of Galilee in AD 30.

Literacy: graded, and rare at the top

Reading, writing one's name and composing a literary narrative are different skills.  Estimates of low literacy must also distinguish these levels.  Catherine Hezser's study of Jewish literacy in Roman Palestine concludes that "at least ninety percent of the Jewish population of Roman Palestine could merely write their own name or not write and read at all."⁵  Chris Keith adds that literacy came in grades, not as a yes or no:  "first-century Palestinian Jews typically gained and used what literate skills their position in life required of them."²⁴  A village clerk could fill in a form he could not compose.  Keith's summary is blunt:  "illiteracy was the rule of the day."²⁴

Two things follow.  The tradition passed mainly by mouth.  And the question whether one of Jesus' followers, such as the tax collector Matthew, could write notes is a question about one or two people, not about the movement.  That question is argued in Who Wrote the Gospels?, which finds that a tax collector's trade put him on a low rung of the ladder rather than at the top.

Notes: available, plausible, and unattested

Note-taking from a teacher is real and documented.  Quintilian complained that "two books on the art of rhetoric are at present circulating under my name," one "taken down by the boys who were my audience" and the other made of "such notes as my good pupils succeeded in taking down."²⁵  Arrian wrote of his teacher Epictetus, "whatever I heard him say I used to write down, word for word, as best I could, endeavouring to preserve it as a memorial, for my own future use," and that the notes had "fallen, I know not how, without my will or knowledge," into public hands.²⁶  Both cases concern educated teachers and pupils, one in the late first century and the other in the early second.

Bauckham argues that the same thing probably happened in the Jesus movement:  "It seems more probable than not that early Christians used them," and "it does seem unlikely that no one would have even noted down Jesus traditions in notebooks for the private use of Christian teachers."¹⁰  But he also concedes that "their closeness to orality must make it virtually impossible for us to distinguish them from oral sources."¹⁰  No early Christian writer mentions notes taken during Jesus' ministry.  Papias, an early writer describing the collection of Gospel material, expressed a preference for personal testimony:  "I did not think that what was to be gotten from the books would profit me as much as what came from the living and abiding voice."²²  A preference for living testimony is compatible with using notes or books;  it is not evidence that notes were absent.  Notes could preserve wording, but this passage does not establish that Jesus' hearers took them.  Long agreements between Matthew and Luke could also come from later written sources or repeated Greek formulations.

Texts could circulate before their authors intended

Quintilian and Arrian illustrate one route of circulation:  pupils' notes reached readers without the teacher's approval.  They do not show that all ancient books lacked an intended publication or an authorised form.  Luke's remark that "many" had already "undertaken to compile a narrative" (Luke 1:1) directly attests earlier accounts;  these examples show how additional written collections could also have circulated without surviving.  It also means notes and the teacher's own account could differ a good deal.  Quintilian says that, compared with his pupils' notes, his own books would contain "eadem aliqua, multa mutata, plurima adiecta":  some of the same material, much changed, and a great deal added.²⁷

Paul: the earliest datable handling of tradition

Paul gives the earliest surviving first-person evidence considered here for handling Jesus traditions.  In 1 Corinthians 11:23 he writes, "I received from the Lord what I also delivered to you," and in 15:3, "I delivered to you as of first importance what I also received."  The Greek pair paralambanō and paradidōmi is the ordinary vocabulary of receiving and handing on.  He had delivered these formulas in person when he founded the church, about AD 50, and he quotes them in a letter about AD 53-55.  Galatians 1:18 says he went to Jerusalem "to visit Cephas" and stayed "fifteen days," about three years after his conversion, conventionally placed in the mid 30s.  This establishes an opportunity to learn from Peter, not what they discussed.  Paul does not name the person from whom he received these formulations.  "From the Lord" can identify their ultimate authority without specifying whether they reached him through people or revelation.  Gerhardsson read this as a pupil passing on a text learned by heart:  "To "hand over" a text is not the same as to recite it once."¹⁵

What Paul shows is a fixed formula, quoted about twenty-five years after the events, that he says he had received earlier.  What Paul does not show is a single controlled body of tradition.  His list of resurrection appearances in 1 Corinthians 15:5-7 names Cephas, the twelve, "more than five hundred brothers," James and "all the apostles."  The Gospels' resurrection stories name women first and never mention the five hundred or a separate appearance to James.  A brief list and a narrative need not include the same appearances, and omission is not contradiction.  These differences rule out treating the surviving passages as one identical script.  They do not establish whether the communities shared additional information or who, if anyone, controlled its wording.

The rabbis: a real model, written down late

The Mishnah preserves the ideal Gerhardsson described.  Rabban Yohanan ben Zakkai praised his pupil Eliezer ben Hyrcanus as "a plastered cistern which loses not a drop."²⁸  And Mishnah Eduyot 1:3 explains why Hillel used an odd unit of measure in a ruling:  "man must speak in the language of his teacher."  The same passage settles a dispute by the testimony of two weavers from the Dung Gate who spoke "in the name of Shemaiah and Avtalion," teachers of the first century BC.²⁸  That is a striking picture of laymen preserving a teacher's exact ruling.  It was compiled about AD 200-220, roughly a century and a half after the Temple fell.  It shows what one later Jewish school valued.  It does not show that Jesus' Galilean disciples, a generation before the rabbinic academies, worked the same way, and the rule against writing such teaching down (b. Gittin 60a) is later still.¹⁰

What the practices establish

Together these witnesses establish that the tools for faithful transmission existed:  memorisation as the normal method of learning, the receiving-and-handing-on vocabulary Paul used, short formulas fixed enough to quote, and the habit of taking notes from teachers.  They establish that Paul and Peter met in the mid 30s and that fixed formulas were being taught by about AD 50.  They do not establish that Jesus made his disciples memorise his sayings, that anyone took notes during his ministry, or that a central body checked what was told.  Those claims rest on analogy.  The surviving texts are direct evidence of its products, and Parts 5 and 6 turn to them.

Part 3: What memory research measures, and what it cannot

Experiments on memory are often quoted on both sides of this debate, usually as slogans:  "memory is reconstructive" on one side, "the core holds" on the other.  The studies themselves measure particular things in particular samples.  Four words need to be kept apart.  Accuracy is agreement with an independently established event or record.  Consistency is agreement with the person's own earlier report, which may itself have been wrong.  In the Challenger study the first-day report serves as the best available baseline, so its authors' accuracy score is a proxy rather than an independent record of each person's circumstances.  Confidence and vividness are what the person feels about the memory.  The table gives the studies this article relies on, with their actual samples.

Three levels of memory

Three distinctions help interpret these studies.  They compare different tasks and do not establish a universal hierarchy for every memory.

The first is exact wording, which spontaneous recall often fails to preserve.  Dean was praised as "the human tape recorder," and Neisser found that for the meeting of 15 September 1972 "hardly a word of Dean's account is true.  Nixon did not say any of the things attributed to him here."²  Long identical runs of words therefore need a separate explanation:  memorisation of short fixed forms, or copying.  Part 6 uses this.

The second is the particulars of a single episode:  where one was, who was there, who said what first, the hour, the order.  These drift and are often replaced.  A quarter of Neisser and Harsch's retested students contradicted their first-day reports on every major point of their own Challenger circumstances, and those most certain were no more accurate:  the 13 who gave the top confidence rating had "about 25% at zero and the median at 2," the same as everyone else.⁴  Hirst's 9/11 respondents gave answers about their own circumstances that matched their first-week reports 63% of the time after eleven months.³¹

The third is the general situation and selected central facts, which were retained better in these examples.  In the ten-year 9/11 study, fifty outside raters scored each fact for how "critical" it was to the story;  the crash sites and the number of planes were rated most critical.  Their accuracy went from .95 and .95 in the first week to .89 and .88 ten years later, while the airline names fell from .88 to .56.³²  Neisser said of Dean, after showing how wrong his reports of particular conversations were, that "his testimony had much truth in it, but not at the level of "gist."  It was true at a deeper level.  Nixon was the kind of man Dean described, he had the knowledge Dean attributed to him, there was a cover-up."²

Even the third level is not perfect.  At ten years, the average score was about .89 for crash sites and .88 for the number of planes.  These are separate item scores, not the proportion of people who answered both correctly.  "Central memories do not change" is not what the data show.  "These central facts were retained better under these study conditions" is.

The fuller 2009 sample also records substantial event-memory loss:  the mean overall score across its five probes fell from .88 in the first survey to .77 at about eleven months, then .78 at thirty-five months.  The strongest-retained items should not be substituted for the complete result.³¹

Retelling can correct an error or fix it in place

Repetition is often treated as a guarantee.  The data say it works both ways.  Hirst's team found that "inconsistent flashbulb memories were more likely to be repeated rather than corrected over the ten-year period;  inaccurate event memories, on the other hand, were more likely to be corrected."³²  Outside information was associated with better event-memory scores.  Respondents who had seen Michael Moore's film Fahrenheit 9/11, which shows where President Bush was that morning, recalled his location correctly 90% of the time in the third survey, against 80% for those who had not.³²  The comparison was observational:  people were not randomly assigned to watch the film.  It supports the possibility of public correction, while warning against treating this highly publicised event as a test of unaided memory over ten years.

Vansina, writing about oral traditions in Africa, reached the same point from fieldwork:  "frequent repetition does not itself guarantee fidelity of reproduction.  Tales told many nights in the month may in fact change faster than tales which are told more infrequently."³³  What matters is the intent of the performance and whether anyone checks it.

Passing a story on: what chain experiments measure

The studies above largely test people recalling an event they experienced or watched.  A second family of experiments tests what happens when a story is passed to someone else.  They use three different designs, and the designs answer different questions.

1. Repeated reproduction: one person retells the same story several times, weeks or months apart.
2. Social contagion: one person hears another person's recall, then recalls alone.
3. Serial reproduction, or the transmission chain: the first person reads a text and writes it out, the second reads only that version and writes it out, and so on, like the party game of telephone.

Five things happen in these experiments, and they are worth keeping apart.

1. Loss and compression. Detail goes first and fast. In the 2004 chains most individually stated actions were gone after one reader. A shorter summary can still express several of those actions, so the proposition count is not a percentage of facts turned false.³⁶
2. Reconstruction. What survives is rewritten toward what the teller expects. Bergman and Roediger replicated Bartlett's finding that people "introduced rationalization and distortion into their accounts of the story, with increases in the proportion of material distorted as retention interval increased." After six months, "most of what people recalled of the story was distorted in some form."³⁴
3. Selective preservation. The outline outlasts the detail. Mesoudi and Whiten report that although each routine shrank drastically, "there is an intuitive sense that the 'gist' or 'core' of the story has been preserved right through the chain."³⁶ Content matters too: in the 2006 chains, "what mattered for superior transmission was that there were a number of third party social agents interacting with one another."³⁷
4. Intrusion. New material enters. About a fifth of what the 2004 chains passed on was not in the original routine, and one fourth-generation reader merged three separate routines into a single day's story.³⁶ A partner's confident error spreads: "one person's memory can be infected by another person's errors."³⁵
5. Accurate reminders. The same channel carries accurate material. Objects that the partner named correctly were recalled 43% of the time, against 26% for objects neither had named.³⁵ In Bergman and Roediger's study, an early recall test reduced the number of wholly invented propositions a week later.³⁴

One result cuts against a common defence.  Errors made early were forgotten no faster than correct material.  Of the items correct on the first test, a mean proportion of .26 was forgotten by the second;  of the items already distorted, .24 was forgotten, and the two "did not differ statistically."  The authors conclude:  "Rather, the subjects introduced new errors into their accounts."³⁴  The authors call repeated recollection "a double-edged sword":  it increases accurate recall and limits major distortions, but if the first retelling contains errors it also increases "the later recollection of those errors as fact."³⁴

These numbers cannot be carried over to the Gospels as rates.  The two serial-chain studies used people who read a text once, without warning that their version would be passed on, and handed on a written copy.  Mesoudi, Whiten and Dunbar list these limits themselves:  their participants "were not instructed that the experiment was a memory test, that the material had come from another participant, or that their recall would be passed on to another participant.  The third parties described in the material were also not known to the participants."³⁷  These chains had no long delay;  the six-month result belongs to Bergman and Roediger's separate, individual-retelling study.  Those authors also note that studies of people's own experienced events found "little change across repeated tests," and leave open "how general this result is."³⁴

Jesus' disciples knew their teacher personally;  later Christian tellers did not all have that advantage.  Paul documents repeated instruction and contact with named leaders, but those facts do not establish the chain behind every Gospel story.  Some differences could improve preservation:  repeated teaching, familiar social content and access to better-informed people.  Others could encourage distortion:  expectations, agreement within a group and early errors fixed by rehearsal.  The studies identify mechanisms under stated conditions.  They provide neither a Gospel error rate nor a measured lower or upper bound for first-century transmission.

Memory research and Bauckham: a critique and its answer

Judith Redman's 2010 article in the Journal of Biblical Literature is a direct critique of Richard Bauckham's use of memory research.  She grants more than she is usually credited with:  "Psychological research into memory supports Bauckham's contention that at least some of the differences attributed to redaction in fact arise from the normal processes of remembering, and that the editing may have contributed far less to the content than the form critics suggest."³⁸  Her objection is timing.  The changes found in eyewitness research "occur within hours, days, or weeks of the event," potentially before a community had established a practice of guarding its retelling.  The timing of those practices in each early Christian setting is itself uncertain.³⁸  So the living eyewitnesses are "a guarantee only of the preservation of the community's agreed version, not of the exact details of the event itself."³⁸

Bauckham answered in the second edition of his book (2017), naming Redman among critics who, he says, "concluded that eyewitness testimony is comprehensively unreliable."  He replies that legal eyewitness research is "a special case ... and only marginally relevant to the case of the Gospels," and that "it was never my intention to claim that eyewitnesses always get everything right or that the eyewitnesses behind the Gospels always got everything right."¹⁰  Each side states the other too strongly.  Redman's text does not call testimony comprehensively unreliable;  it concedes the point quoted above.  Redman summarises Bauckham as holding that "eyewitness transmission guarantees accuracy."³⁸  His own stated claim is that the eyewitnesses "remained the authoritative sources of their traditions until their deaths," a claim about who controlled the tradition, and he denies the stronger reading.¹⁰  Their disagreement concerns which findings transfer to the Gospels, how early distortion could begin, and how far eyewitness authority could prevent or correct it.  Neither a warning about early error nor a denial of perfect recall answers those questions by itself.  The John Hogg case in Part 4 illustrates how an interpretation could enter an eyewitness retelling within days.

What the experiments cannot do

None of these studies measures a story passed from person to person over thirty or sixty years, and none involves people whose teacher's sayings were the centre of their communal life.  They are samples of students and survey volunteers, with heavy attrition:  Hirst's ten-year sample is under a tenth of those who answered the first survey.³²  They show which kinds of content tend to survive and which tend to drift.  They cannot turn that into a probability for a given Gospel story, because nothing in them says how many links stood between a story and its witness, which details a community treated as central, or whether a witness was present to correct it.

Two common uses of this research go too far.  A defender who says "they would surely have remembered something so striking" needs more than the memory's vividness or the witness's confidence;  these studies show that both can accompany substantial error.³⁰  A critic who says "memory is unreliable, so the Gospels are unreliable" is applying results about personal circumstances and exact wording to the outline of a public career that many people watched.  Ehrman's excerpt from the Dean study emphasises the errors in particular meetings, ending with "He was not such a man."  Ehrman also explicitly accepts some reliable gist memories elsewhere, so this is not a dispute between a claim that all memory fails and one that all memory succeeds.⁹  Neisser also judged that Dean "was right about what had really been going on in the White House," but explicitly excepted matters where the significance of Dean's own role was at stake.²  That combination is relevant:  a witness can convey a real situation while misreporting conversations and presenting himself favourably.  One involved witness is not a general experiment on a movement's memory of its founder.

Part 4: A checkable case, the John Hogg stories

Bailey's model rests on his observation of Middle Eastern villages, and almost none of it can be checked, because he published no recordings or field notes.  One part can.  He illustrated how new material enters a tradition with stories about John Hogg (1833-1886), a Scottish missionary who founded many Protestant churches in Upper Egypt, and he said Hogg's daughter Rena had "dipped into this same oral tradition" for her 1914 biography.¹⁶  Theodore Weeden compared Bailey's stories with that book in a 2009 paper titled "Kenneth Bailey's Theory of Oral Tradition:  A Theory Contested by Its Evidence";  James Dunn and Craig Keener replied.³  Most writers on both sides quote one another.  The book itself is in the public domain and settles much of the dispute.

What Rena Hogg actually printed

In her prologue Rena Hogg describes the tradition about her father:  "tales have been preserved which gain in glamour with the years, about a man whom the people call "Hoj."  ... At his words ignorant Mohammedan robbers were transformed into honest Christians in a night."  She adds that "without thought of untruth, whatever they consider good they attribute to him," and in the same sentence that "his sermons are remembered;  his love of song, his powers of physical endurance, the illustrations he employed to point a lesson, his very gestures."¹

Later she tells the robbers story twice.  First comes the popular version, "as related by a fine old patriarch," given explicitly "to show how fact and fancy mingle in such current lore."¹  In it Hogg sets out alone at night, "accosted by a robber band who demanded his gold watch and purse."  He hands them over and offers a greater treasure, a small book.  "Before morning dawned the whole band had been converted," they try to return his goods, he refuses the purse, and "he supported them liberally from that time forward out of his own pocket until they had learned to earn an honest living and had become respected and God-fearing members of the Church!"¹

Then she prints "the original story itself ... as recounted by the chief actors Dr. Hogg and Mr. Shenoodeh Hanna, his companion."¹  After a long day of preaching at Tahta, the two set out on foot for their boat, could not find it, and were frightened by an armed man on the riverbank.  Running inland, they met "a group of smokers, three men and a boy, two of them armed."  Hogg called out, "Call off your dogs, and I shall tell you a story that will make you laugh," and they spent the night in the circle.  Shenoodeh told a Bible story about murder, and one listener confessed he had nearly shot him.  At dawn a guard led them to the boat for a tip.¹  The two witnesses agreed that one man had been about to fire.  They disagreed about who the men were.  Shenoodeh said this "made us certain that these men were highway robbers," while Hogg said they had been warned of "two suspicious characters" and had come to guard "their melon-crop against the marauders."  And "the only real sequel to the story was that Mr. Shenoodeh wove it into an ingenious and thrilling sermon, which greatly moved his audience when it was preached in Assiut a few days later."¹

The timeline

What this case shows, and what it does not

A legend formed within one lifetime, while the eyewitnesses' account existed in writing.  The robbery, the watch, the conversion of the whole band and Hogg's support of them are not in the participants' account.  Bailey's later account retains the legend without Rena's correction.  Without his original field records, that does not establish exactly what villagers told him independently of the book.

The legend had an eyewitness seed.  The participant who called the men robbers was one of the two people there, and he preached it within days.  Keener draws the same conclusion:  "Some key changes in the story may even go back to the telling of one of the eyewitnesses."³  Eyewitness origin is not the same as accuracy.

The printed versions resemble each other, but the route is uncertain.  Bailey's version and the patriarch's share a long sequence of points.  Keener lists nine and calls them "extraordinarily close, too close to be independent."³  Dunn says the "very close parallels" are "strikingly similar to the synoptic parallels which typify a Synopsis of the first three Gospels."¹⁷  Ehrman says Bailey's versions and "the written account in the 1914 biography" were "vastly different."⁹  That is fair if the comparison is with the participants' account, but overstates the difference if it includes the village legend printed in the same book.  The distinction between those two accounts is essential.  But the stability may be weaker than it looks.  Keener himself thinks it "much more likely" that Bailey was recounting Rena Hogg's printed version and noting its similarity to what he had heard, and that Bailey "probably did not take the sort of careful field notes an ethnographer or other anthropologist would."³

Bailey presented the legend as preserved tradition and did not tell his readers what Rena Hogg said about it.  He wrote that he had found "the same stories told in almost the same way," and that "we discovered verifiable evidence of authenticity in oral transmission at least from 1914 to 1960."¹⁶  The robbery story examined here is one that the printed source labels as fact mixed with fancy.  Dunn, defending Bailey, conceded that Weeden "is justified in drawing attention to the issues here and provides an appropriate warning against reading historical events too quickly from the stories to which the events gave rise."¹⁷

What survived was real.  Hogg, a night journey in dangerous country, armed men, a night spent talking with them, a story told to them, and a peaceful end are all in the participants' account.  So is the theme the community cared about, the preacher who meets danger with words.  What changed was the plot and its meaning.

This is an analogy, not evidence about Galilee.  The Hogg case shows what one nineteenth-century Egyptian community did with an adventure story about its founder.  Early Christians valued both Jesus' teaching and claims about his deeds, death and resurrection;  the latter were central to Paul's formulas too.  Rena Hogg's statement that people remembered her father's sermons has not been tested against their texts, so it does not demonstrate that a teaching layer escaped change.  The analogy identifies a possible mechanism of embellishment, not its frequency in a different community.  The case predicts nothing about a particular Gospel story.  It does show that a tradition's stability and its accuracy are separate questions, and that the second needs a source closer to the event.

Eric Eve's review reaches a similar distinction.  He concludes that "Taken together, the Bailey anecdotes we have discussed do not demonstrate that oral tradition in the Middle East is typically controlled in a way that reliably preserves historical fact," while the reshaping involves "not so much invention de novo, as reinterpretation, combination and idealization."  He also judges that "Weeden's criticism of Bailey is a little harsh."²⁰  This is a judgement about what Bailey's evidence demonstrates, rather than a general verdict on all oral tradition.

What anthropology adds: oral history before oral tradition

Jan Vansina, whose Oral Tradition as History (1985) is the standard method for using African oral traditions, draws a line that matters here.  Oral traditions are "oral messages based on previous oral messages, at least a generation old," and "there must be transmission by word of mouth over at least a generation.  Sources for oral history are therefore not included."³³  Material still told by people who lived through the events is oral history, made of reminiscences that "always include hearsay along with eyewitness accounts."³³  On that definition a Gospel author drawing on a living participant would be using oral history;  an author using accounts passed between generations would be using oral tradition.  Both could exist at the same time.  The time since the event alone does not identify the chain behind a particular passage.

Vansina also describes the controls some societies set on performances, and how little they prove.  In Rwanda, the reciters of certain royal ritual texts could be put to death for deviating, but "the only acceptable standard was the remembered performance of the teacher of the performer ... No failure is remembered."³³  His test of fidelity is not the existence of a rule:  "whether or not the intent to be faithful succeeds, and to what degree, will appear from the variability of the versions collected at various performances."³³  That is the test Part 6 applies to the Gospels.

Part 5: Aramaic traces, and what they can identify

Aramaic was part of Jesus' and his followers' linguistic setting, while the surviving Gospels are Greek.  The retained Aramaic expressions and other linguistic features raise questions about translation and bilingual transmission.  They do not require every saying to have started in Aramaic.  Three kinds of evidence get mixed together in this search, and they prove different things.

Aramaic words kept in the Greek.  Mark leaves several sayings untranslated and then translates them:  talitha koum (5:41), ephphatha (7:34), abba (14:36), and the cry from the cross (15:34), along with names such as Boanerges (3:17).  Paul's letters, written in Greek to largely Gentile churches, use abba twice (Galatians 4:6;  Romans 8:15), the prayer marana tha, "Our Lord, come!" (1 Corinthians 16:22), and Peter's Aramaic nickname Cephas throughout 1 Corinthians and Galatians.  Joseph Fitzmyer, a specialist in first-century Aramaic, notes that Paul "has preserved two of the Aramaic expressions normally attributed to Jesus and the early Palestinian Christian church."⁷  This is the hard evidence.  It shows that Aramaic prayer words had reached Greek-speaking congregations by the early 50s and were used there without translation.  That requires an Aramaic-speaking stage that Greek churches looked back to.

"Semitisms" in the Greek.  Some Gospel Greek has features unusual in ordinary Greek and normal in Hebrew or Aramaic.  Fitzmyer warns that "the word "Semitism" can be abused and can turn out to be a weasel-word":  many such features could come from Hebrew, from Aramaic, or from the Greek Old Testament that Greek-speaking Jews read.⁷  And Sanders found that the tradition became "both more and less Semitic" over time, so Semitic colouring is not a reliable sign of an early date.⁸

Retroversion.  A scholar translates a Greek saying back into Aramaic and finds a pun, a rhythm or a clearer sense that the Greek lacks.  Joachim Jeremias argued from such features that Jesus had a recognisable manner of speech, an ipsissima vox, and drew from it a rule:  "In the synoptic tradition it is the inauthenticity, and not the authenticity, of the sayings of Jesus that must be demonstrated."³⁹  He also set the limit himself:  a characteristic style "does not relieve us of the necessity of examining each individual instance," and "the question of authenticity cannot, therefore, be settled in a purely schematic way on the basis of the linguistic and stylistic evidence."³⁹  Fitzmyer adds two controls.  The comparison must use Aramaic of the right period, Qumran Aramaic and first-century inscriptions, not the later Aramaic of the Targums and Syriac.⁷  And whole-Gospel retroversion has failed:  C. C. Torrey's attempt to show the Gospels were written in Aramaic "has convinced no one.  It is in reality no proof of the Aramaic substratum itself."⁷

What a good Aramaic reconstruction does and does not show

A plausible Aramaic reconstruction can support an Aramaic stage when it explains a distinctive feature better than alternatives such as ordinary Greek composition or imitation of biblical style.  Ease of translation alone is insufficient:  many Greek sentences can be rendered naturally into Aramaic.  Even a convincing reconstruction does not locate it uniquely in Palestine or Syria, because Aramaic speakers lived elsewhere too.  It does not show:

1. who said it first. The early Jerusalem church also spoke Aramaic. A saying a Christian prophet formed in Aramaic in the 40s would retrovert as well as one Jesus spoke in the 20s.
2. that an eyewitness passed it on. Language says where a saying was shaped, not by whom it was carried.
3. its date. Aramaic speakers were part of the church for decades. Fitzmyer lists among the things any study must reckon with "the coloring of the tradition by a later faith-experience of the early Christians" and "likely additions to the traditional collections of sayings."⁷
4. that the Greek is a translation of a written source. It may be the translation of an oral saying, by a bilingual teacher or by the evangelist.

So the Aramaic evidence supports a modest, firm claim:  Aramaic expressions reached Greek-speaking churches by the 50s.  This does not establish the same transmission history for every saying or narrative.  It does not, by itself, separate Jesus' words from those of his early followers.  For that, the case rests on content, coherence with other material and multiple attestation, which Hard Sayings and Gospel Differences takes up.

Part 6: Testing the Gospels' own variation

Vansina's test is the comparison of versions.  The Gospels and Paul supply versions, and their Greek can be measured.  The question for each pair is which mechanism the agreement and disagreement fit:  copying from a written text, a fixed formula repeated in worship, a story retold in different words, or no contact at all.

The measures, and a control

Three measures are used, all on the Greek text of the SBL edition.⁴⁰  Shared words counts how many of one passage's words also appear in the other, each matched once.  Longest run is the longest stretch of identical consecutive words.  Rare shared words are dictionary forms found in both passages that occur five times or fewer in the whole New Testament.  The script and its full output are in the article's research folder.⁶

Three selected comparisons give a limited baseline.  Different episodes, such as Jairus' daughter in Mark beside the man born blind in John, share 23 to 32% of their words depending on which passage supplies the denominator, because Greek narrative repeats articles, conjunctions and verbs.  Their longest identical run is two or three words, and they share no rare words.  These three examples are not a probability distribution for chance agreement.  Shared-word percentages need to be interpreted alongside content, passage length, longer runs and distinctive vocabulary.

The table uses exact normalised forms, retaining inflection.  Counting dictionary forms raises agreement, including the controls, but does not by itself distinguish oral transmission from written use.⁶

Copying: long runs of identical words

The studies above give reasons not to assume that spontaneous recollection preserves exact wording.  Deliberate learning and repeated performance require separate assessment.  Robert McIver and Marie Carroll tested this directly with students who reproduced passages from memory or with the text in front of them.  In their main experiment, 43 students performed three tasks in different orders:  writing from prior knowledge, reading and returning a source before writing, and writing with the source available.⁴¹  None in the returned-source condition "had a sequence of words greater than 15," and they proposed that "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."⁴¹  Two later papers in the same journal show the number should not be used as a hard line.  The experiments were in English, and Mark Goodacre points out that English needs more words than Greek, "and so the sixteen-word criterion is much too high for Greek."⁴²  John Poirier found that the lengths of shared runs in the Gospels do not fall into two groups:  "the numbers plot a continuous curve."⁴³  So no cut-off separates memory from copying.  Goodacre argues that the English threshold would miss copying in Greek.  Poirier's objection is broader:  a continuous distribution of run lengths does not sort passages into written and oral sources.  He shows how an editor's altered word, omission or addition can break a longer shared run into several short ones;  a short run therefore does not exclude written dependence.⁴³  Neither paper validates a universal copying test for these examples.

John the Baptist's preaching provides strong evidence of a shared Greek formulation:  sixty-two of Matthew's 77 words also appear in Luke, including an unbroken run of 24 identical words.  Independent translations of an Aramaic speech would be unlikely to match that closely.  Written dependence is a strong explanation, consistent with the wider Synoptic evidence.  A Greek formulation repeatedly learned or recited is another possible source of agreement;  this measurement alone does not eliminate it.

Shared dialogue in different narrative frames

The centurion's servant is Dunn's model case of oral performance, and the measurement shows why.  In the exchange between Jesus and the centurion the two Gospels share a run of 25 words:  "having soldiers under me, and I say to this one, 'Go,' and he goes, and to another, 'Come,' and he comes, and to my slave, 'Do this,' and he does it.  Hearing" this, Jesus marvelled.  Around that core the stories diverge.  In Matthew the centurion comes himself.  In Luke he sends Jewish elders, then friends, and never meets Jesus.  The shared core is the dialogue;  the frame is told differently.

Two readings fit.  On Dunn's, the dialogue was the fixed centre of an oral story whose setting each teller supplied.  On the standard written-source reading, Matthew and Luke both copied one Greek text and each rewrote the setting.  A 25-word identical run is compatible with copying and is not evidence of independent translation.  It can also fit repeated learning of a Greek formulation.  The run therefore supports a shared wording but cannot, on its own, decide between Dunn's oral model and the written-source model.  Dunn concedes the cost of a third version.  John 4:46-54 tells a healing at a distance for an official's son.  If this is another version of the centurion incident, rather than a different healing, Dunn writes that "the memory appears to have become vaguer," so that "we today can have no certainty as to the original suppliant's status."¹⁷

Fixed formulas used in worship

The Last Supper words are the clearest case of tradition kept for use.  Paul's version and Luke's share a run of 10 words, "after supper, saying, 'This cup is the new covenant'" in Paul's order, and the phrase "do this in remembrance of me," with two words found only a handful of times in the New Testament.  Mark's version and Matthew's form a second family:  "this is my blood of the covenant, which is poured out for many."  Paul and Mark share no run longer than three words, while Mark and Luke share five.  These are patterns of closer and looser wording, not two completely separate textual families.  Yet all four keep the same structure:  bread, a blessing or thanks, "this is my body," a cup, covenant and blood.  Paul says he had delivered his form at Corinth before he wrote, so one form was fixed by about AD 50.  Repeated worship could help preserve such formulations.  The textual groupings alone do not identify which church supplied a Gospel, however, and Paul and Luke may also be related through written use.  One caution:  Luke 22:19b-20 is missing from Codex Bezae and some Old Latin manuscripts, and Bruce Metzger's commentary on the critical text sets out why the longer reading, which most editions print, is still preferred.⁴⁴  If the longer text were a later scribe's borrowing from Paul, the Paul-Luke agreement would be scribal, not traditional.

The Lord's Prayer shows the same mixture.  Matthew's form is longer.  Luke lacks "who art in heaven," "thy will be done" and "deliver us from evil."  But the two share an 8-word run and the word epiousios, usually translated "daily," found nowhere else in the New Testament and of uncertain meaning.  The shared unusual Greek term supports a common Greek formulation, later varied in use or editing.  It does not identify the original Aramaic word or decide between an oral and a written Greek source.

Paraphrase with a stated source

Paul's divorce ruling in 1 Corinthians 7:10-11 shares only five words with Mark 10:11-12, and no run longer than one word.  He gives the substance, that a wife should not separate from her husband and a husband should not divorce his wife, in his own words.  What he keeps exactly is the attribution:  "not I, but the Lord."  This is the earliest datable use of a Jesus saying, written about AD 54, and it shows the content held and the wording free.

Shared vocabulary without long identical runs

Mark and John tell the anointing at Bethany with no run longer than five words, and their shared-word figure is close to the controls.  Yet they share four words that occur exactly twice in the New Testament, both times in these two passages:  nardos (nard), pistikos (an obscure adjective describing it), triakosioi (the "three hundred" denarii it could have been sold for), and entaphiasmos (burial).  That strengthens the case that these passages belong to a shared tradition.  Four rare terms are not a calculated probability, however:  the vocabulary is related to the subject matter, and independent accounts of the same incident could also name the same perfume and price.  Literary dependence and a shared oral telling remain possibilities.  It cannot show whether the contact was a text or a story told the same way.  The feeding of the five thousand and Peter's denials share the main facts in Mark and John, five loaves, two fish and two hundred denarii in one, a fire in the high priest's courtyard, three denials and a cock in the other, with no long runs and no rare words.  These are separate presentations of closely related stories.  Short shared runs do not establish independent sources or, by themselves, the historicity of the event.

What the examples show

The variation in the Gospels is not one kind.

1. Written use is strongly supported by the wider pattern of Synoptic agreements, including extended common wording, order and editorial changes. It is argued in The Synoptic Problem. Where one Gospel copies another, the second is not an independent witness to the event; it is a witness to what the first said.
2. Fixed or repeatedly used formulations are a plausible explanation for the Last Supper and prayer material. Paul's Last Supper wording is already attested in the 50s; the shared Greek does not date every Gospel form to that decade.
3. Different narrative presentations surround related material, as in the centurion story, feeding and denials. Retelling and written editing can both produce this result; the measurements do not separate them.
4. Paraphrase with attribution appears in Paul, the earliest datable witness.

Repeated learning could preserve short forms, and retelling could vary the surrounding narrative.  An editor using written sources can produce the same combination, while agreement can also transmit a shared error.  The Greek does not show a tradition that was memorised word for word, and it does not show a tradition that floated free.  Before counting two passages as independent attestation, their possible literary and traditional relationship must be assessed.  An absence of long identical runs does not exclude paraphrase, shared sources or retelling of a written account.  Different details or different name forms in two accounts do not by themselves prove independence:  Matthew changes Mark's names and details while copying him.

Objections and responses

1.  "People in oral cultures had prodigious memories.  Middle Eastern villages still preserve their traditions word for word."  For short fixed forms this is true, and the Gospels show it:  the Last Supper words, the Lord's Prayer and certain sayings keep runs of identical words across versions.  For stories it is not, on the best available test.  The one checkable story in Bailey's essay had become a different story within about forty years.¹  Vansina found that frequent retelling "does not itself guarantee fidelity of reproduction."³³  And the Gospels' own narratives vary in wording, order and detail, which a verbatim tradition would not produce.

2.  "Eyewitnesses were still alive to correct mistakes.  Paul says most of the five hundred were still living (1 Corinthians 15:6)."  Living witnesses give an opportunity for correction.  They do not prove that every version was corrected, or that the Gospel writers met them.  John Hogg's own written account existed and was printed in 1914, and the village version Bailey heard decades later had not absorbed it.  Ehrman's point stands:  a correction reaches only the people who hear it.⁹  On the other side, the early Christian case differs from a scattered village legend.  Its leaders are named and locatable;  Paul went to Jerusalem to see Peter and later met James and John there (Galatians 1:18-19;  2:9).  So correction of the core, which those leaders taught, is more plausible than correction of every peripheral detail.  Neither can be shown story by story.

3.  "Memory research shows eyewitness memory is unreliable, so the Gospels are unreliable."  The famous results are about one's own circumstances when hearing news, exact words, and details suggested by a leading question.²⁹⁴  The same research reports a ten-year mean score of .89 for crash sites and .88 for the number of planes in the repeatedly publicised 9/11 attacks,³² and that John Dean, wrong about the words of a meeting, "was right about what had really been going on."²  The research shows both retention and distortion under particular conditions.  It neither establishes nor rules out invention of a particular Gospel episode.

4.  "The differences between the Gospels prove the stories were invented."  Different wording, order, numbers of people and secondary details are what honest retelling and editing of shared material produce, and the Greek shows both processes.  Invention needs evidence of another kind:  a version that can be compared with a source closer to the event, as in the Hogg case, or a saying that matches later church concerns and has no earlier support.  The Gospels do contain divergences that cannot all be harmonised, and some of them are probably real errors of detail.  That would matter for strict inerrancy.  Its effect on a particular historical argument depends on which claim is wrong and what independent evidence remains.

5.  "Gerhardsson showed that Jewish disciples memorised their teachers' words, so the disciples memorised Jesus' words."  The rabbinic compilations are later, and the Gospel narratives do not reproduce one identical text.  Gerhardsson nevertheless allows editing, so variation alone does not disprove his model.  Gerhardsson's claim about Jesus' own teaching method is an inference, in his words "it seems clear to me."¹⁵  Short sayings may well have been learned by heart, and the long verbatim runs in some sayings are consistent with that.  The narratives were not transmitted as one unchanged verbal text;  that does not identify the mechanism of every alteration.

6.  "The sayings go back easily into Aramaic, so they are Jesus' own words."  A plausible retroversion alone does not establish an Aramaic original.  Where additional linguistic evidence supports an Aramaic stage, it still does not identify the first speaker.  Jesus' first followers also spoke Aramaic, and Fitzmyer warns that the tradition was coloured by "a later faith-experience" and expanded by "likely additions."⁷  Jeremias, the strongest advocate of the Aramaic argument, wrote that authenticity "cannot, therefore, be settled in a purely schematic way on the basis of the linguistic and stylistic evidence."³⁹

7.  "The Hogg case proves the Gospels are legends."  It shows a less elaborate participants' account and a later village legend.  The printed forms resemble each other, but Bailey's possible dependence on the book limits the claim of demonstrated oral stability.  It is one story, from a nineteenth-century Egyptian village, about a minor incident from his travels.  Rena Hogg herself says the teaching layer of her father's memory, his sermons and illustrations, was kept.¹  The case is a warning about dramatic stories that honour a founder.  It does not measure how often embellishment occurred, or establish that a particular Gospel story is legend.

8.  "This is just the telephone game."  The children's game has a single chain, one hearing per link and no opportunity to check back.  The early Christian sources attest repeated teaching, formulas and named leaders;  the chain behind each narrative remains less visible.  But Ehrman does not rely on the game;  his argument is that most tellers were not witnesses and the chain is unknown.⁹  That point is fair.  The length and shape of the chain behind any single Gospel story is not known.

9.  "Matthew the tax collector took notes, so the sayings are verbatim."  It is possible.  It is not attested by any early source, and Bauckham, who thinks notebooks "more probable than not," concedes that notes and oral sources are "virtually impossible" to tell apart in the texts.¹⁰  An unattested notebook cannot serve as a demonstrated link in the transmission chain.

10.  "If the core of the tradition is reliable, the miracles happened."  Reliable transmission preserves what the first witnesses said and believed.  It does not settle whether their interpretation of what they saw was right.  The Hogg case shows an eyewitness's interpretation, that the men were robbers, entering the story within days.  Whether the Gospel miracles and the resurrection happened depends on other evidence, weighed in other articles, and on prior views about whether miracles can happen.  Memory science cannot decide it.

11.  "Admitting variation and error concedes too much to the critics."  It concedes what the texts show.  Matthew and Luke copied and changed Mark;  Paul and Mark preserve different Last Supper wordings;  Dunn identifies John's official story with the centurion tradition, although that identification itself is disputed.  A case for the Gospels that denies this will lose to any reader with a synopsis.  A case that admits it, and then shows how much survives the variation, is the stronger case.

Issues and unanswered questions

1. Weeden's own paper was not read. His 2009 JSHJ article was not obtained through the routes checked. His arguments are represented through Dunn's and Keener's replies and Eric Eve's survey, and the Hogg claim is checked against Rena Hogg's book directly.
2. Social-memory analysis does not supply a transmission rate. The advocates cited here describe how past events and present interests interact. Their conceptual arguments and modern historical examples do not measure the accuracy of an individual Gospel episode.
3. No experiment examined here matches the historical case. The serial-chain studies use one reading and written hand-offs, without long delays. The six-month result concerns one person's repeated recall. Neither design follows repeated teaching and retelling within an early Christian community for thirty to sixty years.
4. The sixteen-word test lacks a Greek validation in the sources examined here. Goodacre's and Poirier's objections prevent using the English threshold as a calibrated Greek test.
5. The width of the "gist" has no agreed definition. The Hirst team's method, outside raters scoring each fact for how central it is, could be applied to Gospel episodes: have readers who know none of the scholarship rate each detail of a story, then compare which details agree across the Gospels. No such study was located for this article.
6. Luke 22:19b-20. If the shorter text of Codex Bezae were original, the agreement between Paul and Luke would be a scribal harmonisation rather than a shared liturgy.
7. The route of contact in the anointing. Four rare shared words strengthen a relationship between the anointing passages but do not uniquely identify its route. Whether John knew Mark, or both knew a common telling, is argued elsewhere and is not settled here.
8. Rena Hogg's claim that her father's sermons were remembered is untested. If any record of the village memory of his sermons exists, it would test the teaching layer directly.
9. Bauckham's second edition does not discuss the Hogg evidence, though it devotes a chapter to Bailey's model. How he would answer Weeden on Hogg is not known from his book.

Verdict

Paul's letters already attest received traditions in the 50s, including teaching attributed to the Lord and wording for the Lord's Supper.  He reports contact with Peter, James and John, providing a concrete opportunity for exchange rather than an anonymous chain of unlimited length.  The Gospels retain substantial common wording in some passages and differ widely in others.  Repeated teaching, oral performance and written editing can all contribute to that pattern.  The text comparisons constrain possible explanations without uniquely recovering the process behind each passage.

The Hogg case and the empirical studies show why preservation and distortion must both remain in view.  A community can preserve an earlier error;  a witness can misremember a conversation yet convey a real situation;  public correction can improve some information without reaching every version.  These results leave room for early historical information in the Gospels and justify studying it with normal source criticism.  They do not establish a uniformly accurate core, a word-for-word record of Jesus' speech, or the truth of a particular miracle.  Those conclusions depend on the evidence for the individual claim.

The companion pages collect scholarly quotations and the literature reviewed.

Sources