Who Wrote the Gospels?

Summary

None of the four Gospels names its author inside the text.  That is not a skeptical claim; it is the starting point everyone shares.  The names arrived on the outside, as titles, and the question is whether those titles preserve real information or later guesswork.

The answer is different for each book.  Anyone who gives one answer for all four is not looking closely.

The four verdicts, stated up front:

What the evidence actually shows

1.  The Gospels are formally anonymous, and the manuscripts show it.  No author speaks in his own name in any of the four.  Matthew 9:9 refers to Matthew in the third person and Luke 6:15 lists him among the Twelve without comment.  The names sit outside the text, in titles — and of the 141 New Testament papyri, ten preserve a title at all, eight of their twenty-six titles belong to a Gospel, and not one preserves a title to Mark.³  The earliest complete titled copies are fourth-century.  So the uniformity of the names is attested from a later and much thinner base than the argument usually implies.  That the titles are later additions is not a skeptical claim.  The scholar most often quoted on the other side says of a second-century date "This again may well be true," and a conservative specialist who thinks the anonymity was deliberate calls the titles "clearly secondary."^{4 5}  How much later is disputed:  Hengel puts them between 69 and 100.

2.  No manuscript carries a rival name — but one ancient group did.  Every surviving titled copy of every Gospel gives the traditional name.⁶  That uniformity is real evidence.  It is also not quite the clean sweep apologists claim:  the group Epiphanius calls the Alogi rejected the Fourth Gospel and said it was written by Cerinthus.⁷  Irenaeus already knows people rejecting John around AD 180.⁸  The honest statement is that one rival attribution was ever proposed, it concerned John, it came with no evidence, and its stated reason was that John's chronology disagrees with the Synoptics.

And it may be thinner even than that.  Charles Hill's book-length study of John's early reception argues that the "Alogi" are not a window onto broad early suspicion of the Fourth Gospel but a late and very thin datum:  Epiphanius coined the name himself, and "many scholars believe that Epiphanius' Alogi are simply Gaius and Gaius alone."  Hill's own conclusion is that they are "either his cipher for Gaius or, more probably, his consolidation of various criticisms of the Fourth Gospel and the Apocalypse which he knew from disparate quarters, into a single, identifiable group."⁹  That cuts our way and this article still reports the attribution, because a rival name was proposed and a reader is entitled to know it — but on either reading the counterexample is a late report under an invented name, not a movement.

3.  For Mark and Matthew, the "unanimous testimony of the fathers" is one witness echoing.  Papias of Hierapolis, writing about AD 100–110, is the source.  Irenaeus reproduces Papias' own vocabulary and names him elsewhere.¹⁰  Clement, Tertullian, Origen and Eusebius sit downstream.  Any argument of the form "the tradition is unanimous across the whole Mediterranean" is inflating a single Phrygian sentence into a chorus.  The one possible second line is for Mark:  Justin, at Rome around 155, mentions "his memoirs" beside a detail found only in Mark 3:17, probably meaning Peter's, but he never names Mark and may have read Papias.¹¹

4.  And the earliest witnesses contradict each other on the details.  Irenaeus says Mark wrote after Peter's death.  Clement of Alexandria says Peter was alive and approved it — and in another passage that Peter "neither forbade nor encouraged" it.¹²  Clement also reports as ancient tradition that the Gospels with genealogies were written first, which nobody believes.  The names are stable.  The stories about them are not.

5.  Papias, read plainly, does not fit the Matthew we have.  He says Matthew "arranged the oracles in the Hebrew dialect, and each interpreted them as he was able."¹³  Our Greek Matthew is not a translation of a Semitic book, and that can be measured rather than argued.  Compared against the Septuagint's Greek translation of Genesis — a known translation from Hebrew — Matthew sits firmly on the composed-Greek side on every marker:  more connective particles than the translation control, a δέ-to-καί ratio half again as Greek, and genitive participles at more than three times the translated rate.¹⁴  The measurement agrees with the literary evidence, reached by a completely different route.

6.  Matthew reproduces Mark, at length and word for word — including in the one story where the author would have been describing himself.  Twelve separate runs of sixteen or more identical consecutive Greek words, the longest thirty-three.¹⁵  The decisive one is short.  Mark 2:14 tells of the call of a tax collector named Levi; Matthew 9:9 tells the same story about Matthew, and sixteen of Mark's twenty Greek words are in it.  In Mark the dinner that follows is "in his house"; Matthew keeps the dinner and changes it to "the house," taking over exactly as much of Mark's sentence as he needed and no more.  That is an editor working on a text, not a man recalling the day he was called.  This reading of the verse is Richard Bauckham's, which is worth noting only because he is the field's leading defender of the Gospels as eyewitness testimony and reaches it anyway.¹⁶

7.  Luke and Acts are by one hand, and that is now measured rather than assumed.  A stylometric test run for this article puts Acts nearest to Luke and far from Matthew, Mark and John — and, crucially, Acts shares no text with Luke, so the resemblance cannot be copying.¹⁷

8.  The "we" passages of Acts are not a spliced-in diary.  The same test finds them stylistically indistinguishable from the rest of Acts, across every parameter setting tried.  Whoever wrote Acts wrote them.  But that leaves two live options, not one — an author who was there, or an author composing in the first person — and stylometry cannot choose between them.

9.  The literacy objection is real, and the tax-collector reply is weaker than it looks.  The gap between signing your name and composing an 18,000-word Greek narrative is enormous, and the documents settle where a tax collector sat in it.  Every surviving customs receipt issued by a named τελώνης in the published papyri — sixteen of them — runs 24 to 48 words, median 34, 557 words in all, and carries one editor's spelling correction every fifteen words, the errors phonetic (τετελώνητε for τετελώνηται).¹⁸  Matthew is 18,329 words.  A trade that produced thirty-four-word receipts with misspellings is a different skill from the one that produced the Sermon on the Mount.

And the most damaging witness is a friendly one.  Origen, defending Christianity around AD 248, describes the apostles as ἁλιεῖς καὶ τελώνας μηδὲ τὰ πρῶτα γράμματα μεμαθηκότας — "fishermen and tax collectors who had not even learned their first letters."¹⁹  The great apologist of the third century puts tax collectors in the same bracket as fishermen, and treats it as something the Gospels themselves record.  So the asymmetry this article once drew — hard on Matthew and John, not at all on Mark and Luke — was drawn in the wrong place.  It is hard on Matthew and John;  and about Mark, nothing is known either way.

One mitigation for John is worth recording because it comes from a source with no stake in it.  Mark 1:20 leaves Zebedee in the boat μετὰ τῶν μισθωτῶν, "with the hired men."  The family employed labour, so the sons of Zebedee were not subsistence fishermen, and a household that hired hands is a household that could have paid for schooling.²⁰  That raises the ceiling on what John might have been able to do.  It does not reach composition of literary Greek, which is several rungs further up, and Origen's sentence above still stands against it.

10.  The best argument for the traditional names is what they are not.  When early Christians did invent Gospel attributions, they chose Thomas, Peter, Philip, Mary, Judas — famous names.  The canonical four include a deserter's assistant (Mark), a Gentile doctor who never met Jesus (Luke), and a tax collector (Matthew).  Tertullian had to construct a rule in the middle of an argument to make the non-apostolic names carry apostolic authority:  "it may seem allowable that what disciples published should be regarded as their masters'."²¹  Nobody constructs a rule to rescue a name he was free to choose.

What this does and does not change

It changes less than either side usually claims.  A Gospel written by an anonymous author using good sources is not thereby unreliable, and a Gospel written by an apostle is not thereby accurate.  Authorship bears on access — how close the writer stood to the events — and access is one input to reliability among several.

The honest position on all four:  these books were written within living memory of the events by people with access to those who were there, and in two cases (Luke–Acts, John) by someone claiming to have been present for part of it.  That is a real historical claim and it survives audit.  "The apostle Matthew wrote Matthew" does not.

Are the Gospels anonymous?

The texts do not name their authors

Open any of the four and no author speaks in his own name.  Matthew 9:9 refers to Matthew in the third person.  Luke 6:15 lists him among the Twelve with no comment.  Mark and Matthew never use "I."  Luke uses "I" in his preface but never names himself.  John's author appears only obliquely, as "the disciple whom Jesus loved," and even the final claim about him is made by somebody else:  "we know that his testimony is true" (John 21:24).

This is not disputed.  Gregory Goswell, writing in a conservative evangelical journal, states it flatly:  the four Gospels, Acts and Hebrews are "strictly speaking, anonymous."²²  Armin Baum calls the superscriptions "clearly secondary."⁵  Michael Bird, an evangelical who defends the names, puts both halves in one place:  "the Gospels are strictly speaking anonymous;  there is no direct self-reference to the authors in these works," and "I am relatively confident about both the antiquity and authenticity of the names used in the titles, but an honest approach means recognizing that the superscriptions may represent a form of aetiological development."²³

And the scholar apologists most often quote on the other side concedes it too.  Simon Gathercole's 2013 study of the titles is the standard reference for the claim that the names are early.  In his 2018 follow-up he takes the majority view in E. D. Freed's words, that "the present titles probably were not added until sometime in the second century," and says of it:  "This again may well be true."  Hengel's case against it, he writes, "has not really persuaded many others."  And:  "It is not clear that Mark added ευαγγελιον κατα Μαρκον as a prefix or a suffix to his autographon."⁴

Gathercole's own conclusion should be stated too, because it is not the skeptic's.  He argues that an author's name could be attached to a book in other ways than the titles that survive, and that "the most likely conclusion to be drawn is that the attributions of authorship are original."  What he gives up is the titles in their present form.  He does not give up the names.

And Hengel himself does not claim the evangelists wrote them.  This matters, because his name is the one the argument is usually borrowed from.  What Hengel proposes is that titles were attached by the scribes who copied and circulated the books, and he says so:  the authority behind a title "need not have been identical with the author."  On John he is blunter still — the title "was added only after the author's death, by his pupils."²⁴  His date for the process is between 69 and 100, and his reason is practical:  a congregation reading from more than one Gospel in worship, and shelving more than one scroll, had to be able to tell them apart.

Hengel's claim is that the Gospels were never published untitled.  It is not a claim that any evangelist signed his book, and when he writes that the titles are "original" he means belonging to the book as circulated, not written by the man who wrote the book.

He also flags his own argument as a hypothesis, twice, and concedes the point that does it the most damage:  where an author was already well known to his community, a spoken word of explanation would have been enough, and no written title would have been needed at all.²⁵

So any sentence of the form "the manuscript titles are original" is refuted by our own best witnesses — plural.  Neither of the two scholars this argument is borrowed from claims what it is used to claim, and this article does not use it.

What the titles do establish

What Gathercole's 2013 census establishes is narrower and holds up.  Across every surviving titled Greek, Latin, Syriac and Coptic Gospel manuscript likely to be earlier than AD 500, the variation in the titles is entirely formal — εὐαγγέλιον κατὰ Μᾶρκον, κατὰ Μᾶρκον, longer or shorter — and never nominal.  No copy of the second Gospel is attributed to anyone but Mark.⁶

The control that gives this its force is Hebrews.  Hebrews is a New Testament book that really did circulate without an author's name, and the result was not silence — it was guessing, and the guesses disagreed.  Inside about a century it drew four different attributions:  Paul (Pantaenus, and the placement of Hebrews between Romans and 1 Corinthians in P46), Barnabas (Tertullian), and Clement of Rome and Luke (both reported by Origen, who concluded that God alone knew who wrote it).²⁶

That is what an unattributed early Christian text looks like in the record.  Four names, in competition, across the same period and the same churches that transmitted the Gospels.  Against that background the Gospel titles' uniformity is a real datum rather than an assumption — and it is also the fair measure of how much weight it carries, since Hebrews shows the machinery of guessing was fully operational.

That is worth something, and it is worth being precise about how much.  Only ten of the 141 New Testament papyri preserve any title at all, and only eight of their twenty-six titles belong to a Gospel, on five papyri.  No papyrus preserves a title to Mark.³  So the uniformity is real, but it is attested from a later and thinner base than the argument usually implies.

The comparison that clarifies it:  P46, an early codex of Paul's letters, preserves nine titles and none of them names Paul — exactly as papyri of the Iliad never name Homer.³  So a title did not have to name an author at all.  Allen puts the two collections at opposite ends of ancient practice, "with the Gospels (focusing on authorial personae) and Paul (lacking any mention of the author) on either end of the spectrum."³  Naming the author in the title was one ordinary way to label a book, and the Gospel titles do it every time.

The Old Testament pattern, tested

Baum's explanation is that the evangelists were following a Jewish convention:  in the Hebrew Bible, historical books are anonymous and prophetic and wisdom books carry names.⁵

Checked against the Old Testament, this is a tendency rather than a rule.  Nehemiah 1:1 carries a superscription in the same form as Jeremiah's and Amos's — "the words of Nehemiah son of Hacaliah" — and Baum counts Nehemiah among the anonymous histories.  Job is a wisdom book and is anonymous.  Lamentations has no Hebrew superscription.  Jonah names its prophet as the subject, not the author.

The tendency is real and the explanation is reasonable.  It should be presented as a tendency.

Is the patristic testimony one witness or several?

This is the question that decides how much the external evidence is worth, and the answer is different for different books.

Three conclusions follow.

Before Irenaeus, only Papias names an evangelist.  Justin Martyr, writing at Rome around 155, quotes the Gospels constantly and calls them "the memoirs of the apostles" (ἀπομνημονεύματα τῶν ἀποστόλων).  He never names one.²⁷  The Epistula Apostolorum names eleven apostles as authors of its own book and attributes no Gospel to anyone.

Justin has one near-exception, and it concerns Mark.  At Dialogue with Trypho 106.3 he says that Jesus renamed one apostle Peter, that this is written "in his memoirs" (ἐν τοῖς ἀπομνημονεύμασιν αὐτοῦ), and that he renamed two other brothers, the sons of Zebedee, "Boanerges, which is sons of thunder."¹¹  Three facts bear on it:

• The detail points to Mark. Βοανηργές occurs once in the New Testament, at Mark 3:17, one verse after the renaming of Simon. Matthew and Luke both drop it.
• The pronoun is ambiguous. "His" can be Christ's or Peter's. The old Ante-Nicene Fathers translation prints "the memoirs of Him," meaning Christ. Georges Archambault, who edited the Greek in 1909, answers that Justin never calls the Gospels memoirs of Christ, always memoirs of the apostles, so it "must be read" as memoirs of Peter.
• Archambault then assigns the phrase to the Gospel of Peter, not to Mark, because of parallels Harnack found between that book and Justin's passion material. The weakness of his choice is that the Gospel of Peter as it survives is a fragment of the passion and resurrection only, with no naming of apostles in it. The only known book that contains what Justin describes is Mark.

Bauckham takes the passage as Justin calling Mark "Peter's memoirs," and adds that Papias used the matching verb, ἀπεμνημόνευσεν, of what Mark wrote down from Peter.  He thinks Justin "unlikely" to have taken the term from Papias' passing use of it, and more likely that the description was already current.¹⁶  That cuts both ways:  the shared word may show a common older tradition, or it may show that Justin had read Papias.  So the fair statement is this.  Justin, at Rome around 155, probably refers to Mark as Peter's memoirs.  He never uses Mark's name, the reading rests on a pronoun, and his independence from Papias is possible rather than shown.

For Mark and Matthew the chain is a single line.  Irenaeus uses Papias' own terms — Mark as Peter's "interpreter" (ἑρμηνευτής), Matthew writing among the Hebrews "in their own dialect" (διαλέκτῳ, Papias' noun) — and cites Papias by name elsewhere.¹⁰  Everyone later repeats it.  Two qualifications keep this from being overstated in the other direction.  The later writers add things Papias' surviving words do not contain — Rome, the request of Peter's hearers, whether Peter was alive — and they disagree with each other about them, which is not what simple copying produces.  And Justin's Dialogue 106.3, above, may be a second Roman line for Mark.  Neither is enough to turn one stream into several independent witnesses;  both are enough to make "one man's sentence, echoed" too simple for Mark.  For Matthew it remains the fair description.

For John it genuinely is not.  Theophilus of Antioch names John after AD 180 with no known connection to the Papias–Irenaeus line, and Polycrates of Ephesus independently attests a John buried at Ephesus.²⁸  John is the best externally attested of the four, not the weakest — which is the reverse of what most readers expect.

Where the external evidence is weaker than usually claimed

• The Muratorian Fragment cannot be treated as a secure c. 170 witness. Its single manuscript is seventh- or eighth-century and badly copied, its Latin is a translation whose language "has now been dated to the early fifth century," and Albert Sundberg and Geoffrey Hahneman have argued for a fourth-century Eastern origin.²⁹ That redating has been strongly resisted, and the resistance should be reported too. Eckhard Schnabel's 2014 survey of the research sets eight criticisms against Hahneman, among them that Victorinus of Pettau, around 270, seems to borrow from the Fragment, and it reports Christoph Markschies' judgment that "the traditional date c. AD 200 is supported by the better arguments."³⁰ A witness whose date specialists dispute by two centuries cannot anchor an argument. Even on the traditional date it does not name Mark — the opening line is the tail of a lost sentence.
• The Anti-Marcionite Prologues cannot carry weight as second-century evidence, and the case against them was made in 1946. Engelbert Gütwenger, on the prologue to John: "there is neither an internal nor an external reason for dating the prologue earlier than 300." His verdict on the three together is more guarded than "worthless." Their authority "is at least questionable," and "it is uncritical to use them in the same manner as the documents of well-known writers of the early patristic literature."³¹ The John prologue has the apostle John rejecting Marcion, who was active around 140. Popular apologetics still cites them. This article does not.
• Origen's reasons, where he gives them, are exegetical guesses. Mark-and-Peter rests on 1 Peter 5:13, which he quotes for "Marcus, my son." Luke's is "the Gospel commended by Paul," a phrase usually traced to 2 Corinthians 8:18 and its unnamed "brother whose praise is in the gospel," though Origen does not cite the verse (both in Eusebius, HE 6.25.5–6, McGiffert's translation). Both are deductions any reader of the New Testament could make.
• Irenaeus argues the number four from cosmology — four regions of the world, four winds, four faces of the cherubim, therefore "neither more nor fewer" than four Gospels.³² A skeptic is entitled to weigh that against his reliability as a historian. The usual reply, that he is defending a collection already fixed, is an assumption the passage does not prove.
• Every source in this chain is on the winning side. Papias survives only because Eusebius quoted him — and Eusebius calls him "a man of very small intelligence."³³ Marcion's writings are gone. The Alogi survive only through Epiphanius, who invented their name. The absence of rival attributions is real, but it is an absence measured in a library curated by the party that won.

Papias, and the Hebrew Matthew problem

Papias of Hierapolis wrote five books called Exposition of the Oracles of the Lord.  They are lost.  What survives is quotation, mostly in Eusebius.

He is earlier than the skeptical retelling usually allows.  The standard date of c. 130 rests on Philip of Side alone.  Jerome's Latin of Eusebius' Chronicle places Papias at AD 100, alongside Polycarp and Ignatius, and the current specialist consensus sits near c. 110.³⁴  This favours the traditional case and should be said.

On Mark, quoting an informant he calls the Elder:

"Mark, having become Peter's interpreter, wrote down accurately, though not in order, whatever he remembered of the things said or done by the Lord.  For he had neither heard the Lord nor followed him, but afterwards, as I said, followed Peter."³⁵

That is a claim with a named source, a date before 130, and — notably — a criticism built into it.  A fabricator inventing authority does not add "though not in order."

On Matthew, and here the trouble starts:

"Matthew arranged the oracles in the Hebrew dialect, and each interpreted them as he was able."¹³

Read plainly, this does not describe the book we have.  Four things weigh against it, and none is an opinion:

1. Matthew and Mark share sixteen consecutive identical Greek words of plain narration at Matthew 7:28–29 // Mark 1:22, and eleven more such runs.¹⁵
2. Both carry the parenthesis "let the reader understand" — addressed to a reader of Greek.
3. Matthew's own argument at 21:16 depends on the Septuagint's "praise" where the Hebrew reads "strength."
4. Measured against a known translation from Hebrew, Matthew's Greek is composed Greek, not translated Greek, on every marker.¹⁴

The one reading that would dissolve the problem is Josef Kürzinger's proposal that Ἑβραΐδι διαλέκτῳ means "in a Hebrew style" rather than in the Hebrew language.  It is a minority view.  C. Clifton Black:  "few have been persuaded that … Papias means, as Kürzinger argues, 'Semitic style' (not 'Hebrew language,' surely a more natural translation)."  Bauckham calls it one of "the least convincing of Kürzinger's fresh interpretations of Papias" and lists William Schoedel, Matthew Black, C. Clifton Black, Davies and Allison, and Hagner against it;  Davies and Allison write that "Kürzinger has not convinced";  John Wenham, a conservative who wants an early Matthew, calls it "strained."  But it is held by serious people, and Robert Gundry is one of them.³⁶

This article does not use it, and the reason is a measurement rather than a head-count.  Whatever Papias meant, our Greek Matthew does not read as Greek translated from a Semitic original:  set against the Septuagint's Genesis, a known translation from Hebrew, it carries more connective particles than the translation control, a δέ-to-καί ratio half again as Greek, and genitive participles at more than three times the translated rate.¹⁴  The stylistic reading of Papias could be right about Papias and would still leave that unchanged.

Three further cautions, all of which cut against the traditional case:

• The Matthew notice is not attributed to the Elder. Eusebius introduces it only with "and about Matthew he said this." Anyone who writes "John the Elder tells us Matthew wrote in Hebrew" has added a source the text does not name.
• The most-quoted sentence may not be Papias'. "For he took care of one thing, to omit nothing of what he heard and to falsify nothing" is marked as possibly Eusebius' own in Stephen Carlson's 2021 critical edition, and he explains why in a 2025 article: the sentence fits Eusebius' own concerns, and "one ought to be suspicious of such an editorial intervention in this case."³⁷ It is the sentence apologists quote most.
• Papias transmits material that is plainly bad. His account of Judas' death — a body swollen beyond the width of a chariot track, then crushed — is a third version incompatible with both Matthew and Acts.³⁸ He is a collector, not a critical filter.

What survives, stated conservatively:  a named, datable, pre-130 claim that a book by Mark records Peter's teaching, sourced to a named informant, and carrying a criticism of the book it describes.  That is a better class of evidence than the second-century Christian average.  What does not survive is the plain reading of his Matthew sentence as a description of the book we have.  It may well survive as a description of some other book:  that is precisely James Edwards' case, that a Hebrew Gospel existed, that the patristic references to it point to a real document, and that it fed Luke's special material.³⁹  His reviewers accept the first two points more readily than the third.  The measurement above constrains what Papias can have meant about canonical Matthew;  it says nothing about whether the apostle also wrote something else.

Mark

The traditional author is John Mark:  a Jerusalem Jew (Acts 12:12), a companion of Barnabas and Paul who abandoned them mid-journey (Acts 13:13, 15:38), later reconciled, and associated with Peter (1 Peter 5:13).

The internal evidence is consistent with the tradition without proving it.

• The Greek is the least literary of the four, heavily paratactic, and it retains Aramaic phrases and then translates them for the reader — talitha koum, ephphatha, Eloi Eloi lema sabachthani, abba, korban. A writer who does that is bilingual and writing for readers who are not.
• Measured against a known translation from Hebrew, Mark is the most Semitic-looking of the four — more paratactic than the Septuagint's Genesis on three of four markers.¹⁴ This is not what the usual apologetic ordering expects, which treats Matthew as the Jewish Gospel and Mark as the Roman one. It does not make Mark a translation: its genitive-participle rate is three times the translated rate, so it is composed Greek in a plain register.
• Latin loanwords are densest in Mark, and the figure was counted rather than asserted. Mark carries 1.77 Latin loanword tokens per thousand words, against 1.04 for Matthew and 0.36 for Luke; three of his Latin words appear nowhere else in the New Testament, and twice he explains a Greek term with a Latin one for his readers.⁴⁰ Two honest deflations belong with that number. Mark's commonest "Latinism," κράβαττος, may be a Macedonian or Thracian word that entered Greek through Latin rather than a Latin word; drop it and Mark falls to 1.33, close to John's 1.10. And the count is driven by subject matter — Mark's Latin clusters in the passion, the tribute question and the execution of the Baptist, which are scenes about Roman soldiers, Roman taxes and Roman coins. A writer describing a centurion would use the ordinary word for a centurion wherever he lived.
• And the decisive check kills the Roman inference outright. Josephus wrote the Jewish War in Rome, under Flavian patronage, for a Roman readership — and he uses none of Mark's Latin loanwords. Not one occurrence of κεντυρίων, λεγιών, πραιτώριον, δηνάριον or φραγελλόω in either the War or the Antiquities; he writes ἑκατοντάρχης and τάγμα instead. Meanwhile κεντυρίων appears in forty-one documentary papyri and ostraca from Roman Egypt.⁴⁰ And the strongest form of the argument — Mark 12:42 glossing a Greek coin with a Latin one — is matched exactly by m. Kiddushin 1:1, which defines the perutah against "the Italian issar": the same operation, in Hebrew, in Palestine. Latinisms track register, not geography. Mark's do not locate him in Rome, and any argument in this archive that used them for that purpose is withdrawn.
• Peter frames the narrative — first named disciple at 1:16, last named at 16:7. Bauckham argues this is a deliberate device marking the eyewitness source. Critics answer that Peter is simply the most frequently named character, so the pattern may fall out of frequency alone.
• Three small details sit closer to Peter in Mark than in the Gospels that reuse Mark.⁴¹ (1) The verb "remembered" (ἀναμιμνῄσκω) occurs twice in Mark, at 11:21 (the withered fig tree) and 14:72 (the cock-crow), and Peter is the subject both times. Matthew 21:20 retells the first and replaces Peter with "the disciples." (2) In Gethsemane, Mark 14:37 has Jesus speak to Peter alone, in the singular: "Simon, are you asleep? Were you not strong enough to watch one hour?" Matthew 26:40 keeps "he says to Peter" and turns the verb plural. (3) Mark has the cushion Jesus slept on in the boat (4:38) and the green grass at the feeding of the five thousand (6:39); Matthew and Luke tell both stories and keep neither detail. The honest weight of these is modest. Two occurrences of a verb is a tiny sample. A good storyteller can invent a cushion. And Matthew and Luke shorten Mark nearly everywhere, so lost details are what editing predicts whatever Mark's source was. What the three do show is direction: where the versions differ, the Peter-specific and the concrete are in Mark, and the later Gospels generalise. That is what the tradition predicts, and it does not prove it.

One popular argument is not used here.  2 Peter 1:16–18 has Peter recall the Transfiguration, and it is often said to match Mark's account.  It matches Matthew's.  The voice in 2 Peter says "in whom I am well pleased" (εἰς ὃν ἐγὼ εὐδόκησα), which is in Matthew 17:5 and absent from Mark 9:7, and 2 Peter lacks Mark's "listen to him."⁴¹  2 Peter is also the New Testament letter whose authorship is most widely doubted, so a skeptical reader would not grant the premise.

The objections, stated fairly.  Mark is charged with geographical errors:  the route in 7:31 runs from Tyre to Sidon to the Sea of Galilee through the Decapolis, which reads oddly; 5:1 sets a lakeside miracle at Gerasa, about thirty miles inland (and the manuscripts disagree about the name); 11:1 lists Bethphage before Bethany, which is the wrong order coming from Jericho.  These are real difficulties, and a writer with close local knowledge would not be expected to make them.

Verdict:  probably right.  The attribution is early, traceable to about AD 100–110, has no ancient rival, and fits a man nobody would invent — a deserter's assistant with no claim to have seen anything.  Against that, the external case is one named witness, Papias, plus a probable but unnamed reference in Justin, and the two earliest accounts of the book's origin contradict each other about whether Peter was alive when it was written.

Matthew

The name is early, and it is not derived from the text.  This point deserves more weight than it usually gets, and it comes from a scholar who rejects apostolic authorship.  G. D. Kilpatrick:  "Even after the changes of Matt. ix.9, x.3, Matthew is a much less important figure than Peter and if an apostolic name was to be sought from the contents of the book, it would be expected that Peter would be chosen."⁴²  If someone were inventing an attribution from what the Gospel contains, Peter is the obvious choice.  He was not chosen.  Kilpatrick's own conclusion from this is not the traditional one.  He holds that the book was "from the beginning deliberately pseudonymous," put out under Matthew's name by its author around AD 90–100.  What his sentence rules out is a title guessed later from the two verses.

The one internal argument that points toward Matthew, measured.  A popular apologetic case holds that the Gospel shows a tax collector's preoccupation with money.  It is testable, so it was tested — every currency, metal-as-money, wealth, tax and debt term in the four Gospels and Acts, counted from the tagged Greek.⁴³

What survives:  Matthew does lead, he names the widest range of coins of any New Testament writer, and two coins appear in no other Gospel — the δίδραχμον and the στατήρ, both in the temple-tax episode at 17:24–27.

What does not survive is the inference.  The popular version puts Matthew against Mark, where the gap looks enormous.  Set him against Luke and it nearly closes:  5.18 against 3.96 per thousand words, with Luke actually ahead on general wealth vocabulary and level on tax words (18 against 17).  Luke is traditionally a physician, not a tax collector.  A measure that cannot tell a tax collector from a doctor is not evidence of a tax collector — it is evidence of an interest in money, which Luke's Gospel is famous for on entirely different grounds.

But the apostle did not write the Greek book we have.  Three independent lines converge:

1. Papias' claim does not describe it — see above, and the measurement confirms it.
2. It reproduces Mark, verbatim, at length. Twelve runs of sixteen or more identical consecutive Greek words, longest thirty-three.¹⁵
3. Matthew 9:9. Mark 2:14 tells of the call of Levi son of Alphaeus. Matthew 9:9 tells the same story about Matthew, and apart from the name the wording is virtually identical. In Mark the dinner follows "in his house" (2:15); Matthew keeps the dinner and changes it to "the house" (9:10). The writer took over exactly as much of Mark's Levi story as he needed to attach the call to Matthew, and no more.

The three lines converge on one conclusion, and the evidence carries it without help:  a book whose account of its own author's call takes over sixteen of the twenty words of another writer's sentence about another man was not written by that author.  The name was attached to the Gospel;  the apostle did not compose the Greek.  The same reading of Matthew 9:9 is Bauckham's, and he states the consequence plainly:  "the author of Matthew's Gospel intended to associate the Gospel with the apostle Matthew but was not himself the apostle Matthew.  Matthew himself could have described his own call without having to take over the way Mark described Levi's call."¹⁶

The answers on offer, weighed.

• Deny Markan priority. Available — the Griesbach and Augustinian solutions dissolve the problem outright. But both buy relief in one place by paying more elsewhere, and nobody who does not already want an apostolic Matthew finds them persuasive. See the synoptic article.
• An eyewitness may prefer an authoritative source. The best partial answer. Mark's Gospel carried Peter's authority, and a writer wanting to align with the church's leading witness had reason to follow him closely. It explains why Matthew might use Mark. It does not explain 9:9.
• The name attaches to a source, not the finished book. The honest one, and Bauckham's. It concedes that the man who composed our Greek Matthew was not the apostle.
• A Matthean school. Costs nothing and predicts nothing. It is a way of holding tradition and data together, not an argument.
• Papias' "oracles" were the sayings source Q. Dead. Papias uses the same word of things the Lord "said or did," so it does not mean "sayings."

Verdict:  the apostle Matthew stands somewhere behind this Gospel and did not write it.  That is less than the tradition claims and more than nothing — and it is where the evidence sits, not a compromise.

It is worth recording who else lands there, because the route matters more than the tally.  Martin Hengel got there first.  In 1989 he wrote that the First Gospel was composed "by an unknown Jewish-Christian teacher who, as the Papias note suggests, transferred the apostle's name, Matthew, from his main source, the logia of Jesus, to his own work to give this apostolic authority also."  In 2000 he put it in one word:  the Greek First Gospel "'pseudepigraphically' bore the name of the apostle Matthew."¹  Richard Bauckham reaches it defending eyewitness testimony;  Donald Hagner's commentary reaches it by positing that the apostle collected the discourse material and the name travelled with it;⁴⁴  James Edwards reaches it arguing that the apostle really did write a Hebrew Gospel — a different book from ours.³⁹  And Maurice Casey, who left the Christian faith in 1962, reaches the same place from the far side:  "I followed the ancient tradition that Matthew the apostle and tax-collector wrote down traditions about Jesus, mostly his teaching, in Aramaic, and that different translations were made, some better than others.  This may be the reason why Matthew the Evangelist was called Matthew."⁴⁵  Five starting points, including one hostile to apologetics, converging on the same two-part answer is worth more than any of them separately.

And the standard conservative textbook states the difficulty against itself.  Carson and Moo grant that unless Kürzinger's stylistic reading of Papias is adopted, "we are gently nudged to the conclusion that Papias was wrong when he claimed that Matthew was first written in Aramaic" — and then ask the question their own argument raises:  "And if he was wrong on this point, what prevents us from supposing that he was likely wrong in his ascription of authorship to the apostle Matthew?"⁴⁶  They do answer it, in the next lines, and their answer should be heard:  "Such skepticism, superficially plausible, seems a trifle extreme.  The two issues are not integrally connected.  Authors have been known to err on one point without erring on all points!"  Papias' Semitic original, they suggest, "may have been an intelligent, albeit erroneous, guess."  That reply is fair as far as it goes.  An error about the language does not cancel the name by logic.  What it leaves standing is that the one early sentence carrying the name is a sentence that is wrong about the book, and no second early witness supplies the name without it.  An earlier draft of this article said Carson and Moo left their question unanswered.  That was false, and it is withdrawn.

Luke and Acts

One author, measured

Common authorship of Luke and Acts is usually rested on the two prefaces and on general impression.  It is testable.  A stylometric test run for this article cut all five books into equal blocks and asked which blocks resemble which.  Acts is nearest to Luke (mean Delta 0.947) and much further from Matthew (1.032), Mark (1.036) and John (1.120) — and Acts's own internal scatter is 0.863, so the gap to Luke is barely above the book's variation with itself.¹⁷

The reason this is decisive rather than circular:  Acts shares no text with Luke.  The one long verbatim run between them is a quotation of Psalm 110:1.  So the resemblance cannot be copying.  It is habit.

Mikeal Parsons and Richard Pervo are often cited as challenging this.  They do not:  they concede authorial unity in terms — "we find no grounds for attempting to reopen this question" — and dispute the generic, narrative and theological unity instead.⁴⁷

This cuts both ways, and the article says so.  Luke and Acts stand or fall together.  Anything that dates Acts late dates the Gospel of Luke with it.

The "we" passages

Four stretches of Acts switch to the first person plural:  16:10–17, 20:5–15, 21:1–18, and 27:1–28:16 — 1,722 Greek words, 9.4% of the book.  Four explanations compete.

One of them can be ruled out by measurement, and is.  If Acts incorporated somebody else's travel diary, those sections should read differently from their surroundings.  They do not.  The distance from the "we" material to the rest of Acts is the same as the distance from the rest of Acts to itself, at every block size and feature count tried — and at two settings the "we" material is closer to the rest of Acts than the rest of Acts is to itself.¹⁷  Whoever wrote Acts wrote the "we" passages.

A second explanation collapses on inspection.  Vernon Robbins argued in 1978 that first-person plural was a literary convention of ancient sea-voyage narrative, used "whether or not the author was an actual participant."  He had many first-person voyages to point to, but his closest parallels were three texts in which a third-person narrative shifts to "we" as a voyage begins.  Colin Hemer in 1985 and Susan Praeder in 1987 each checked all three independently, and all three fail.  Joseph Fitzmyer in 1989 went through Robbins' wider examples, from Egyptian tales and Homer to Dio Chrysostom and Josephus, and remained "skeptical about Robbins' explanation of the We-Sections."⁴⁸  Praeder then built the comparison set Robbins never built:  eighteen first-person sea voyages against twenty-nine third-person ones, with seven authors appearing in both columns.  There is no convention.  Acts also narrates its other sea voyages in the third person.

But killing the convention does not deliver the traditional answer, and this is the finding that matters most.  Praeder's own conclusion is that Acts is unparalleled in both directions:

"If Acts is a first person ancient history, then it is alone in its lack of first person singular participation."⁴⁹

An eyewitness who never says "I," never names himself, never speaks, never acts, and disappears at Rome is as strange as a convention nobody else used.  She rejects the usual reply that Luke was too unimportant to use "I":  "even an unimportant participant is not exempt from the authorial obligations of character identification and clarity."⁴⁹  Her stated aim was "not to propose another solution but to provide an understanding of first person narration in Acts as the unsolved problem that it is."⁴⁹

And the fourth option survives everything above.  Colin Hemer, a maximalist on Acts, concedes it:  "The suggestion that the 'we-passages' embody a false claim to personal participation seems a stronger alternative than any of the preceding.  It involves no special explanation."⁵⁰  Stylometry cannot touch that, because a single author writing his own memories and a single author composing a first-person voyage produce the same measurements.

Luke against Paul

The strongest internal argument against a companion of Paul is Philipp Vielhauer's 1950 comparison of Acts' Paul with the Paul of the letters:  natural theology, the law, Christology and eschatology all differ.  His verdict is that the author of Acts "presents no specifically Pauline idea."⁵¹

Nothing in the research for this article refutes the observation that Acts' Paul never preaches the cross the way Paul preaches it, never states justification by faith in Paul's terms, and takes a Nazirite vow to demonstrate that he keeps the law.  Fitzmyer, who accepts the "we" passages as autobiographical, also accepts that Vielhauer's thrust "has been generally accepted."⁵²

The best available reply is Fitzmyer's own — that the author was away from Paul between about AD 50 and 58 — and it buys the divergences at the price of Irenaeus' description of Luke as "inseparable from Paul" (Adversus Haereses 3.14.1), which is the form the tradition actually takes.  Fitzmyer says so himself:  the author was "at most a sometime companion of Paul."⁵²

There is also Acts 15 against Galatians 2.  Paul, writing under oath, lists two Jerusalem visits where Acts has three, and says "those of repute added nothing to me" about a meeting Acts says produced four binding requirements.  Hemer's identification of Acts 11 with Galatians 2 dissolves this, but only if the South Galatian theory is right and Galatians is Paul's earliest letter — two contested premises doing the work.

The medical argument, declined

W. K. Hobart argued in 1882 that Luke's vocabulary is that of a physician.  Henry Cadbury tested the same words against Josephus, Plutarch and Lucian in 1919 and found nine in ten of them there too.  His verdict:  "The style of Luke bears no more evidence of medical training and interest than does the language of other writers who were not physicians."⁵³

Luuk van de Weghe reopened the case in 2025 with the Thesaurus Linguae Graecae, and his position should be stated in his own words before it is weighed.  He holds that each of Cadbury's three objections "is undermined by advances in scholarship," and that the internal evidence "is completely consistent with the traditional attribution."  On the word count, though, his result is Cadbury's:  "Luke and Lucian display a comparable number of medical terms."⁵⁴  Lucian was a satirist, not a physician.  Van de Weghe's answer is that Lucian was highly educated and interested in medicine while Luke probably was not schooled to that level, and that Luke has a few terms found only in medical writers.  That is a fair argument.  It also turns a count into a judgment about Luke's schooling, which is not known, and "consistent with" is the most its own author claims for it.  This article does not use the medical argument.  An earlier draft went further and called it dead, citing van de Weghe as if he agreed.  He does not, and that claim is withdrawn.

Verdict

The strongest of the four on internal grounds, and the one with the least motive for invention.  Luke is a non-apostle, mentioned three times in the New Testament, never as a witness to anything Jesus did.  Nobody inventing authority picks him.

Against that:  two of the three Luke verses are in letters many critics reject (Colossians and 2 Timothy); strip those and "Luke the beloved physician" disappears, leaving a name in a list at Philemon 24.  And Praeder's problem is unsolved — the "we" passages fit no known pattern, including the traditional one.

A companion of Paul wrote the "we" passages, and there is no good reason to doubt that his name was Luke.  There is also no independent confirmation of it.

John

Two questions, and running them together is the commonest mistake in this area.

1. Was the author an eyewitness? The evidence is decent.
2. Was he John son of Zebedee? The evidence is weak, and the leading conservative scholar says no.

The case for a Judaean eyewitness

John is literarily independent of the Synoptics, and that is now a measurement.  Zero runs of sixteen identical words with any of them.  Six-word overlap sitting inside the noise band set by comparing books nobody thinks are related.¹⁵  Whatever John knew of the Synoptic tradition, he did not have a Synoptic text open beside him.

He knows pre-70 Jerusalem.  He names the pool of Bethesda by the Sheep Gate, the pool of Siloam, the Pavement called Gabbatha, Jacob's well, Solomon's portico, Aenon near Salim, Bethany beyond the Jordan.  He knows the stone water jars used for Jewish purification, the festival calendar, and burial customs.  He alone names the servant whose ear was cut off — Malchus.  That last detail is often pressed further, into an argument that Mark suppressed names to protect living people and John could name them once they were dead.  That argument does not survive counting, and it is taken apart below.

He claims to have seen it:  "He who saw it has borne witness, and his testimony is true" (19:35).

And the honest corrections

The topographical argument is weaker than apologetic books present it, and the corrections come from excavators, not skeptics.

• The most popular claim is false. The Ecce Homo pavement shown to pilgrims as John's Λιθόστρωτον is Hadrianic — second century. Pierre Benoit, the Dominican archaeologist who re-examined the site: "This pavement, however, had no connection with either Jesus or the Antonia fortress; it is merely a Roman forum from the days of Hadrian."⁵⁵ His argument can be checked without him. Josephus has the Struthion Pool, which that pavement covers, still open when the Fifth Legion raised a siege ramp across the middle of it in AD 70.⁵⁶ Jesus did not stand there.
• The Siloam identification is contested inside Israeli archaeology. A 2023 study in the Israel Antiquities Authority's journal makes the large stepped pool uncovered in 2004 into Solomon's Pool, an open swimming pool, and identifies John's Siloam with the smaller pool at the outlet of the Siloam Tunnel, a short walk away.⁵⁷ "They found the Pool of Siloam exactly where John said" is premature.
• Bethesda's five porticoes have never been excavated. The twin pools and the dam are real; the colonnades are an inference. Eusebius does independently describe the place in his own day, around AD 320: "Once it had five porticos. There are now pointed out twin pools."⁵⁸ That is genuine support — but it is not a first-century portico.
• Two of John's seven distinctive place-names remain unfound. Bethany beyond the Jordan defeated Origen, who lived and travelled in Palestine and changed his text rather than accept a place he could not locate.⁵⁹ Aenon near Salim is still unlocated.

What survives is still substantial:  a writer with detailed, accurate, pre-70 Judaean knowledge that he did not get from the Synoptics and did not get from the Old Testament.

The "protective anonymity" argument, tested and abandoned

Mark 14:47 has "one of those standing by" draw a sword and cut off "the slave of the high priest's" ear.  John 18:10 names both:  Peter, and Malchus.  A popular argument reads Mark as shielding people still in legal danger, and John as free to name them once they were dead.

It is a testable claim, so it was tested.  Every human character in all four Gospels was catalogued for this article — 260 rows — and the "walk-ons" counted:  individuals who are not the Twelve, not Jesus' family, not the Baptist, and not named office-holders.  The pattern is the reverse of the one the argument predicts, three separate ways.⁶⁰

1. Mark is not the anonymising Gospel. Matthew is. Mark names a higher share of his minor characters than Matthew does, 36.6% against 22.6%. Luke and John name more than either, 42.9% and 46.4%.
2. Every Gospel names more freely in Jerusalem, not less — including Mark, whose rate rises from 28.6% outside to 45.0% inside. Gerd Theissen, who first made the argument, put it in terms of place: "Only in Jerusalem was there reason to draw a cloak of anonymity over followers of Jesus who had endangered themselves by their actions."⁶¹ The cloak is not there in any of the four.
3. Mark names nine walk-ons in his passion narrative; John names six. John does name the largest share of walk-ons of the four, but in a shorter cast. On the crudest measure he is no freer with names than Mark: 69 distinct capitalised noun lemmas to Mark's 73.

What survives is smaller and real.  Mark almost never introduces a person as "a certain" somebody.  He attaches τις to a human noun six times against Luke's thirty-seven, and five of his six are plurals or a hypothetical case.  The one that introduces a real individual is the "certain young man" (νεανίσκος τις) who runs away naked at 14:51, and four verses earlier the swordsman is "a certain one of those standing by" (εἷς δέ τις τῶν παρεστηκότων, 14:47).  Both stand in the arrest scene.  That is the oddity Theissen started from, and it is real.  So there is something to explain about how Mark describes those two men.  There is nothing to explain about him hiding names in general, because he hides fewer of them than Matthew does.

Why probably not the son of Zebedee

Three things in the text itself point away from the son of Zebedee, and a reader can check each of them:

• John never names John son of Zebedee. The sons of Zebedee appear once, at 21:2, in a list of seven that also has "two others of his disciples" unnamed. The Beloved Disciple of 21:7 can be one of the sons or one of the two.
• The "other disciple" of 18:15–16, usually taken to be the Beloved Disciple, is known to the high priest — the Greek γνωστός with the dative means an intimate, not an acquaintance. And at 19:27 the Beloved Disciple takes Jesus' mother "to his own home" from that hour, which reads most easily of a home near at hand. A Galilean fisherman is an awkward fit. Both points are inferences, and the text does not say outright that the two disciples are one man or where the home was.
• The Gospel's material is overwhelmingly Judaean, not Galilean.

This case is Richard Bauckham's, who reached it from the internal evidence before re-examining the fathers, and who identifies the author with "John the Elder," the figure Papias distinguishes from the apostle.¹⁶  It is worth recording that the argument comes from the field's foremost defender of the Gospels as eyewitness testimony — not because that makes it true, but because a reader is entitled to know the position is not a skeptical one.

It was Martin Hengel's case before it was Bauckham's.  Hengel wrote in 1989 that "there are too many historical reasons against supposing that the Gospel was composed by John the son of Zebedee," and the first he names is that "this Gospel cannot come from a Galilean fisherman."¹

The conservative reply deserves its own words.  Andreas Köstenberger and Stephen Stout answered Bauckham in 2008.  Their strongest point is the Last Supper.  Rejecting the apostle, they write, means "that a disciple who was not one of the Twelve was not only present in the upper room — when Matt 26:20 and Mark 14:17 state it was only the Twelve, identified in Luke 22:14 as 'the apostles' — but leaned on Jesus' chest in a place of high honor."  They conclude that the author "certainly leaves more than enough hints to narrow the list," and that the apostle John is the one left standing.⁶³  That argument has force, and it is the reason this article says "probably" and not more.  Its weak point is the word "only."  Mark 14:17 and Matthew 26:20 say that Jesus came and reclined "with the twelve."  Neither says nobody else was in the house.

Where Bauckham overreaches, this article says so.  His claim that 21:2 "actually excludes" the son of Zebedee is stronger than the evidence supports.  Hengel, who holds the same view of the author, reads the same verse as leaving the question open:  at 21:2 "the audience again has a choice" between the son of Zebedee and one of the two unnamed disciples.¹  And Bauckham's argument from a 496-syllable prologue matching a 496-word epilogue is unfalsifiable as presented and is not repeated here.

And the traditional identification would not buy what it is usually wanted for.  Hengel again:  "Anyone who still adheres to the authorship of the Gospel by the son of Zebedee cannot in so doing rescue the historical trustworthiness of the Fourth Gospel."¹  Who wrote the book and how reliable it is are separate questions, and the second is not settled by winning the first.

What cannot be claimed

The words of the discourses are the evangelist's.  This was measured for the article.  Counting fifteen characteristic Johannine keywords per thousand words:  Jesus in John 61.7, the narrator's own asides 74–99, 1 John 73.5, and John the Baptist at 3:27–36 a comparable 57.1 — against Jesus in Matthew 8.9 and Jesus in Mark 5.1.⁶⁴  The long speeches in this Gospel are in the idiom of the man who wrote it.  The pattern is not uniform, and the exceptions belong on the record:  the Baptist's earlier speech at 1:19–34 scores 29.7 and Jesus' dialogue with the Samaritan woman 13.7, and many readers assign 3:31–36 to the narrator, not to the Baptist.

That does not settle authorship — a witness who has retold his material for decades in his own idiom produces the same signature.  But it does settle what may be claimed.  Defend the substance of John's discourses.  Never defend the wording as verbatim.

One further honest observation:  John uses the word Ἰουδαῖος, "Jew," seventy-one times, against sixteen in all three Synoptics combined, most often for Jesus' opponents as a body, explains Jewish customs to his readers, and puts "your law" in Jesus' mouth.  This Gospel was written for readers outside the Judaean community, after the two had come apart.

Verdict

An eyewitness named John, probably not the son of Zebedee.  Best attested from outside of the four — two geographically independent lines — and the most confidently claimed by its own text.  Weakest of the four on the specific identification, and the case against that identification is made by the field's foremost defender of eyewitness testimony.

The case against traditional authorship

Stated at full strength, because an article that beats a straw man is worth nothing.

1.  Literacy.  Composing a continuous Greek prose narrative was a rare skill at the far end of a graded scale.  Catherine Hezser's criterion is the ability to read literary texts, and she places the large majority of Palestinian Jews, urban and rural, below it.⁶⁵  Her summary figure is 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."⁶⁵  William Harris concluded that even in Roman Italy "the overall level of literacy is likely to have been below 15%."  Acts 4:13 calls Peter and John ἀγράμματοί … καὶ ἰδιῶται — "unlettered and untrained."

Reply, and its limits.  The narrow form of this argument is strong and the broad form is not:  writing was common in first-century Palestine, and a tax collector wrote for a living — but he wrote receipts, and the sixteen surviving ones average thirty-four words with a spelling error every fifteen.¹⁸  Trade literacy moves Matthew one rung up a graded scale, not to the top of it.  Secretaries were normal — Paul dictated to Tertius (Romans 16:22).  But an amanuensis explains the Greek at the cost of "the apostle wrote it" in the strict sense.  And note where the argument bites:  hard on Matthew and John, not at all on Mark and Luke.

2.  The Gospels never name their authors and refer to the traditional authors in the third person.  Bart Ehrman's form of it has force:  "Even when this Gospel narrates the event of Matthew being called to become a disciple, it talks about 'him,' not about 'me.'"⁶⁶  Would an eyewitness write that way?  The reply has to explain why Luke's "we" appears in volume two and never in volume one.

3.  Attribution as authorisation.  That the names were attached in the second century to secure apostolic authority against Marcion and the gnostics.  The chronology is suggestive:  Irenaeus, our first four-name witness, is writing precisely that kind of polemic.

Reply.  The names are demonstrably older than Irenaeus for Mark and Matthew (Papias), and the chosen names are wrong for the purpose.  An authorising invention picks apostles.

4.  Theological development, especially John's Christology, as too developed for the eyewitness generation.

Reply.  Weaker than its reputation.  Its old dating anchor has collapsed (see the dating article), and Philippians 2:6–11 is a very high Christology from the 50s or earlier.  What survives is that John contains both high and low Christology, suggesting layered composition — a point about the book's production, not about who produced it.

5.  Matthew used Mark.  The strongest of them all, and treated in full above.

Why these four names?

If the attributions were invented, the choices are strange.

What Christians actually chose when they invented:  Thomas, Peter, Philip, Mary, Judas, James, Bartholomew, Nicodemus.  Famous names, apostolic names, names that confer authority.

What the canonical Gospels carry:  a man who deserted Paul mid-journey (Mark); a Gentile physician who never met Jesus and is named three times in the New Testament (Luke); a tax collector, the most despised trade in Judaea, and a minor figure among the Twelve (Matthew).  Only John is an obvious choice.

The strongest evidence that this mattered is Tertullian.  Arguing against Marcion in about AD 208, he needed the non-apostolic names to carry apostolic weight, and constructed a rule to do it:  "it may seem allowable that what disciples published should be regarded as their masters'."²¹  A principle invented in the middle of an argument, to solve a problem the church had inherited rather than created.

The counter-argument, which is real.  The attributions may have been constrained rather than free.  If the "we" passages already pointed to a companion of Paul, the field of candidates was small, and "Luke" may be an inference from the text rather than a memory.  Origen's reasoning for Mark and Luke — from 1 Peter 5:13 and, it seems, 2 Corinthians 8:18 — shows early Christians doing exactly that kind of deducing.

Where the balance lands.  The argument from awkward names is genuinely good for Mark, Luke and Matthew, and it is not decisive, because a constrained guess and a preserved memory can produce the same result.  It is worth stating as what it is:  a strong probabilistic argument, not a proof.

Objections and responses

"Anonymous means unreliable." No.  Most ancient historical writing that survives is anonymous or pseudonymous somewhere, and the books of Kings and Samuel are anonymous.  What anonymity costs is the ability to check the author's access directly.  It says nothing about what he knew.

"The titles are on every manuscript, so they must be original." The manuscripts that carry titles are later than the composition by a long way, and no papyrus preserves a title to Mark at all.  Uniformity across the surviving tradition shows the names were fixed early enough to leave no rival.  It does not show the evangelists wrote them.  Neither of the two scholars this argument is normally borrowed from claims that it does.  Gathercole says it is not clear that Mark put his own name on his book.  Hengel, who originated the early-title argument, has the titles added by the scribes who copied and sent out the books, and in John's case by the author's pupils after his death.

"The church fathers unanimously attribute these Gospels." For John, the attestation is genuinely multiple.  For Matthew it is one witness — Papias — repeated.  For Mark it is Papias repeated, plus a probable reference in Justin that never gives the name.  For Luke it is thin and uncontradicted.  "Unanimous" describes the outcome, not the number of independent sources.

"If Matthew didn't write Matthew, the whole thing is a fraud." That does not follow.  A book compiled from apostolic material and circulated under the name of the apostle behind that material is not a forgery in any sense the ancient world recognised, and the alternative — that a second-century church picked the least impressive available name — is harder to believe than the tradition it is meant to replace.

"Bauckham defends eyewitness testimony, so he supports traditional authorship." He does not, on Matthew or on John, and an article that quotes him for the first while ignoring the second is misusing him.  This article cites him against itself in both places.

"Stylometry proves Luke was Paul's companion." It proves nothing of the sort, and the claim should not be made.  What it proves is that Luke and Acts are by one hand and that the "we" passages are not an inserted document.  Whether that hand belonged to a participant is a separate question the measurement cannot reach.

What is not known

• No independent confirmation of any of the four names exists outside the Christian tradition.
• The "we" passages of Acts fit no known pattern, including the traditional one. Praeder called it an unsolved problem and it remains one.
• The identity of the Beloved Disciple is not settled, and the leading conservative treatment rejects the traditional answer.
• Whether the patristic writers preserved memory or made inferences from the same texts we have. Mark Goodacre states the deflationary view and labels it an assumption: "It is usually assumed that these fathers did not have a special knowledge of the order of the composition of the Gospels. Originally, someone made some inferences from the knowledge they did have, and these inferences soon became the basis for a steady, repeated tradition."⁶⁷ He is writing about the order of the Gospels, not their authors, but the same question applies to the names, and it has not been tested.
• How much of the apostle Matthew is in Matthew. "He stands behind it" is the verdict the evidence supports and it is not a precise one.

Sources