The Book of Acts
🔗Overview
The book of the Acts of the Apostles is not included in the UPDV. While Acts appears to be generally based on historical people and events for its outline, many of its details are shaped by literary convention and by the author's theological program rather than transmitted from primary testimony.
This conclusion rests on six categories of evidence. First, computational stylometry — the forensic analysis of unconscious writing habits in the Greek text — identifies the author of Acts as a different writer than the author of the Gospel of Luke. Second, the traditional attribution of both works to a companion of Paul rests on a chain of deduction from disputed and pseudepigraphic epistles. Third, Acts contradicts Paul's own firsthand testimony on major events. Fourth, Acts contradicts itself internally, altering physical details across its three accounts of Paul's conversion. Fifth, the "we" passages that supposedly prove eyewitness authorship follow established conventions of ancient sea voyage literature. And sixth, the narrative techniques throughout Acts — speeches, utopian summaries, dramatic escapes — conform to Greco-Roman literary conventions rather than historical reporting.
The stylometric evidence establishes that Acts was composed by a different author than the Core Gospel. The literary and historical evidence — contradictions with Paul, internal inconsistencies, literary conventions governing the "we" passages and speeches, and the second-century origins of the traditional attribution — demonstrates that its narrative cannot be treated as historically reliable. The computational evidence is new; the literary and historical evidence has been established for over a century.
🔗Three Hands in Luke-Acts
The traditional view holds that a single author — conventionally called "Luke" — wrote both the Gospel and Acts. Computational stylometry tells a different story.
Burrows' Delta is a standard method in forensic linguistics for detecting authorship. It works by measuring how frequently an author uses common function words — particles like τε (te, "and"), δέ (de, "but"), and μέν (men, "on the one hand") — the small connecting words that writers use unconsciously and cannot easily fake. The method was developed by John Burrows in 2002 and has been used to identify anonymous authors, detect forgeries, and resolve disputed authorship in English, Latin, and Greek texts.1
Applied to the Greek text of Luke-Acts (using 70 function words measured across 1,000-word sliding windows of the NA28 critical text), the algorithm identifies three distinct compositional layers without being told to look for them:2
The Core Gospel (Luke 3–23:38) writes standard narrative Greek, with high rates of δέ and ἄνθρωπος (anthrōpos, "person"), and virtually no τε. This author's style clusters tightly with Mark and Matthew — the Synoptic tradition.
The Framing Editor (Luke 1–2 and Luke 23:39–24:53) writes heavily Semitized Greek saturated with Septuagint vocabulary: ἐγένετο (egeneto, "it came to pass") at nearly 7 per thousand words, κύριος (kyrios, "Lord") at over 13 per thousand, and ἰδού (idou, "behold") at nearly 5 per thousand. This author's style clusters with LXX 1 Samuel, not with the Core Gospel.
The Acts Author (Acts 1–28) writes Hellenistic literary Greek with a massive τε rate of over 8 per thousand words — nineteen times the rate of the Core Gospel. This signature appears in 25 of 28 chapters, across every source layer: itineraries, "we" passages, speeches, and trial narratives. It is the author's own connective habit, not inherited from any source.
The following table shows selected diagnostic markers — rates per 1,000 words across the four sections of Luke-Acts:
| Marker | Luke 1–2 | Core Gospel | Recon Zone | Acts |
|---|---|---|---|---|
| τε (te, "and") | 0.49 | 0.43 | 0.92 | 8.18 |
| δέ (de, "but") | 12.78 | 29.46 | 26.70 | 30.03 |
| ἐγένετο (egeneto, "it happened") | 6.88 | 2.87 | 7.37 | 2.93 |
| ἰδού (idou, "behold") | 4.91 | 2.63 | 3.68 | 1.25 |
| ἀνήρ (anēr, "man") | 0.49 | 1.04 | 3.68 | 4.99 |
| ἄνθρωπος (anthrōpos, "person") | 2.46 | 4.95 | 2.76 | 2.44 |
| κύριος (kyrios, "Lord") | 13.27 | 4.28 | 1.84 | 5.64 |
| ἅγιος (hagios, "holy") | 5.41 | 0.55 | 0.00 | 2.71 |
Each layer has a distinctive profile. The Core Gospel's high δέ and ἄνθρωπος mark standard narrative Greek. The Framing Editor's heavy ἐγένετο, κύριος, and ἅγιος reveal Septuagint saturation. The Acts Author's τε rate — an order of magnitude above every other section — is the single most diagnostic feature in the analysis.
Across the full corpus, the rates per 1,000 words are:
| Text | τε/1000 |
|---|---|
| Core Gospel (Luke 3–23:38) | 0.43 |
| Gospel of Mark | 0.57 |
| Gospel of Matthew | 0.62 |
| LXX 1 Samuel | 2.18 |
| Josephus, Antiquities 1–11 (biblical register) | 8.23 |
| Acts of the Apostles | 8.18 |
| Josephus, Jewish War (contemporary history) | 10.70 |
The Core Gospel's τε rate is not merely lower than Acts — it is in a different register entirely from all known literary-historiographical Greek. Mark and Matthew confirm that the Synoptic authors uniformly avoid τε. Acts at 8.18 sits squarely inside the range of Hellenistic historiography (Josephus: 8.23–10.70). The Core Gospel at 0.43 sits with other Synoptic narrative — an order of magnitude away. No known genre shift in Greek literature produces a function-word rate difference of this magnitude within a single author.
LXX 1 Samuel, at 2.18 per thousand, occupies an intermediate position — above Synoptic narrative but well below literary historiography. This is consistent with translation Greek: the Septuagint translators were not composing freely in literary Koine but rendering a Hebrew source, which naturally suppresses connective particles. The τε distribution across the corpus forms a compositional-register gradient — translation Greek, Synoptic narrative, literary historiography — with Acts at the literary end and the Core Gospel at the other. It is not a clean binary, and it does not need to be: the Core Gospel's position near zero and Acts's position inside the Josephus range are sufficient to establish the divergence.
To test whether register alone could explain the gap, Josephus's Antiquities books 1–11 and Jewish War were compared. Josephus is indisputably a single author. Antiquities 1–11 is Septuagintalizing biblical narrative — the precise operation "Luke imitating LXX style" posits for chapters 1–2. Jewish War is contemporary military history. If register drives τε, the same author's rate should drop sharply across this shift. It drops 23% — from 10.70 to 8.23 per thousand, a factor of 1.3. The Acts Author's rate differs from the Core Gospel's by a factor of 19. A 1.3× register effect against a 19× observed gap is more than ten times short of what the register hypothesis requires, measured from a large, well-powered sample of the closest available analogue. The register explanation for the τε gap is not viable.2
The Burrows' Delta distance matrix confirms these patterns across the full corpus. Lower numbers indicate more similar writing:
| Luke | Acts | Mark | Matt | LXX 1 Sam | Josephus (sample) | |
|---|---|---|---|---|---|---|
| Luke | — | 0.486 | 0.438 | 0.424 | 0.599 | 0.991 |
| Acts | 0.486 | — | 0.675 | 0.615 | 0.673 | 0.802 |
| Mark | 0.438 | 0.675 | — | 0.429 | 0.606 | 1.014 |
| Matt | 0.424 | 0.615 | 0.429 | — | 0.627 | 1.000 |
| LXX 1 Sam | 0.599 | 0.673 | 0.606 | 0.627 | — | 0.982 |
| Josephus (sample) | 0.991 | 0.802 | 1.014 | 1.000 | 0.982 | — |
Luke is closest to Matthew (0.424) and Mark (0.438) — the Synoptic tradition. Acts is closest to Luke (0.486) but far from Mark (0.675) and Matthew (0.615).
Distributional checks across the corpus:
| Test | Windows | Result |
|---|---|---|
| Core Gospel → Mark vs. Acts | 134 (1000-word, 100-step) | 84% of windows closer to Mark |
| Core Gospel → Matt vs. Acts | 134 (1000-word, 100-step) | 88% of windows closer to Matt |
| L-material (unique to Luke) vs. Acts | Pericope-level | 95% closer to Core Gospel centroid |
| L-material leave-out validation | Independent baseline | Significant — away from Acts (t = 3.319) |
Note: the sliding-window counts are descriptive proportions. The L-material tests use source-classified pericopes at the composition level, independent of window overlap. Non-overlapping block validation confirms the same pattern: in 15 non-overlapping 1,000-word blocks from the Core Gospel, 13 cluster closer to Mark than to Acts.
Luke 1–2, at approximately 2,500 words, is the smallest layer in the analysis and yields only 2–3 independent sample windows. Its separation from the Core Gospel is carried by the full function-word feature set, not by τε alone — its τε rate of 0.49 sits near the Core Gospel's 0.43, well below the historiographical range. The clustering result that places Luke 1–2 closer to Acts than to the Core Gospel should be interpreted with this sample limitation in mind.2
The Core Gospel is securely placed within the Synoptic tradition.
Historical checks — cross-validating the same underlying text through different hypotheses — all converge with the computational findings:
| Test | Result |
|---|---|
| Marcion's absent chapters (1–2) cluster as Framing Editor | Yes — 100% |
| UPDV reconstruction zone shows Framing Editor signal | Yes — 33% in ch. 23 |
| Core body (chapters 3–21) clusters as Core Gospel | Yes — 95% |
| L-material (unique to Luke) clusters with Synoptic baseline | Yes — away from Acts |
| Acts internally homogeneous | Yes — 175/175 windows |
Crucially, material unique to Luke (L-material) — passages with no parallel in Mark or Matthew — also clusters with the Synoptic baseline and away from Acts, confirming that the stylistic divergence is not driven by shared sources.
A natural objection is that the author of Acts was deliberately imitating Septuagint style in Luke 1–2. To test this, the five most obvious Semitic function words were stripped from the analysis and the algorithm was re-run. Luke 1–2 remained 3.6 times closer to LXX 1 Samuel than the Core Gospel was. The affinity is structural rather than a superficial veneer of biblical vocabulary; whether it reflects a different authorial syntax is a stronger claim than this sample size can resolve.2
The computational evidence separates the Core Gospel's compositional register from that of Acts. The Framing Editor layer rests on a smaller sample and is correspondingly less secure. Whether the Framing Editor and the Acts Author are the same individual deploying different stylistic registers, or two separate editors, both are stylometrically distinct from the author of the original Gospel narrative.
This finding converges with independent textual evidence. Marcion's second-century Gospel — whether he edited Luke or possessed an earlier form — lacked chapters 1–2. The algorithm, without knowing this, flagged those same chapters as stylistically distinct. The UPDV's reconstruction boundary at Luke 23:39, established on textual-critical grounds, aligns precisely with where the algorithm detects the stylistic shift from Core Gospel to Framing Editor. Three independent methods — textual criticism, manuscript history, and computational analysis — converge on the same boundaries.2
🔗The Authorship Chain
The traditional attribution of Luke-Acts to "Luke the physician," a companion of Paul, rests on a surprisingly thin foundation.
A.W. Argyle first documented the scale of the linguistic divergence between the two works. The particle τε appears 9 times in Luke but 151 times in Acts. The word ἀνήρ (anēr, "man") appears 27 times in Luke but 100 times in Acts, while ἄνθρωπος (anthrōpos, "man/person") appears 95 times in Luke but only 46 times in Acts. These are not obscure vocabulary differences — they are high-frequency function words and basic nouns, the kind of language patterns most resistant to deliberate change and most revealing of authorial habit.3
Richard Pervo's Hermeneia commentary traces the attribution to its origin. Irenaeus of Lyon, writing around 180 CE, identified the author of Luke-Acts as "Luke," deducing this from three New Testament texts: Colossians 4:14, 2 Timothy 4:11, and Philemon 24. No independent or external tradition was available to Irenaeus, who may have originated the claim entirely.4
The textual foundation of this deduction is circular. Of the three passages naming "Luke" as Paul's companion, only Philemon is undisputedly genuine. Colossians is disputed among scholars. 2 Timothy is widely regarded as pseudepigraphic — written after Paul's death in Paul's name. The attribution thus rests on one authentic verse plus two potentially forged ones, and even the authentic verse says nothing about Luke writing anything.4
The attribution served an institutional purpose. Irenaeus needed Acts tied to a companion of Paul so that Marcion's reading of Luke's Gospel — which lacked chapters 1–2 and was not connected to Acts — could not stand as an independent witness. As Pervo observes, "The authorship of Luke and Acts is an element of the early orthodox synthesis."4
Further, Acts 1:21-22 describes an apostleship requirement — "beginning from the baptism of John until the day when he was taken up from us" — that more plausibly describes a Gospel starting at Luke 3:1 than one starting at 1:1 or 1:5. The Acts author appears to know a Gospel without the infancy narrative.5
Critical scholars generally date Acts between 85 and 115 CE, with Pervo in his Hermeneia commentary arguing for the latter — two full generations after Paul. Regardless of the exact date, the consensus is clear: "The actual author was not a companion of Paul." Even those who retain the traditional attribution, Pervo notes, "must ask how much has been saved by retaining a Luke who was a companion of Paul but who did not trouble himself to learn about the major struggles in which Paul was engaged."4
🔗Contradictions with Paul's Own Account
Acts presents events in Paul's life that contradict Paul's own firsthand testimony in his letters. Because Paul's letters are primary sources — written by the participant himself — they must be given priority over any secondary account where the two diverge. Barrett's commentary operates on this principle throughout.6
🔗The Arabia Journey
In Galatians 1:16-17, Paul states that after his calling, he "did not immediately consult with flesh and blood" but "went away into Arabia, and returned again to Damascus." Only after three years did he visit Jerusalem, where he saw only Peter and James — a claim he introduces with a solemn oath: "before God, I don't lie" (Gal 1:20).
The corresponding narrative in Acts 9:19-26 tells a different story. Paul stays "some days" with the disciples in Damascus, preaches immediately in the synagogues, and then goes to Jerusalem. Acts shows no knowledge of the Arabia journey. Barrett notes this directly: Paul "went away into Arabia, a journey of which Acts shows no knowledge."7
🔗The Damascus Escape
Paul describes his escape from Damascus in 2 Corinthians 11:32-33: he was fleeing the ethnarch of King Aretas of the Nabataean Arabs. In Acts 9:23-25, the same incident is rewritten — Paul is now fleeing "the Jews" who are watching the gates. Barrett states: "It is impossible to doubt that the same incident is in mind, and that Paul's version of it is more likely to be correct than any other. Luke probably blamed the Jews because he had no exact information and had come to think that whenever trouble arose it must have been caused by them."8
🔗The Jerusalem Council and the Apostolic Decree
The most consequential contradiction concerns the Jerusalem Council. In Acts 15, a formal council issues a binding decree — the Apostolic Decree — requiring Gentile believers to "abstain from things sacrificed to idols, from blood, from things strangled, and from sexual immorality" (Acts 15:29). Paul is depicted as agreeing to this decree and delivering it to the churches.
Paul's own account in Galatians 2:1-10 describes a private meeting, not a public council. He reports a division of missionary labor (Gal 2:9) that Acts never mentions. Most critically, Paul states flatly: "they... imparted nothing to me" (Gal 2:6). No decree. No binding requirements. Nothing added.
Pervo's Hermeneia commentary reveals the literary architecture behind this discrepancy. Acts 15 uses Galatians 2 as its primary source but revises the account substantially: the conflict is moved from Antioch to Jerusalem, the circumcision dispute is replaced with a moderate food-law compromise, and roles are systematically redistributed. The strict position is assigned to "believing Pharisees," Peter receives the liberal position, and James becomes the moderate. Paul, the protagonist of his own account, sits on the sidelines. Titus — whom Paul specifically names as central to the circumcision dispute (Gal 2:3) — is eliminated entirely from Acts, suffering what Pervo calls "damnatio memoriae."9
The Apostolic Decree itself was likely not a product of Jerusalem at all. Pervo argues it "is more likely to have been worked out in a mixed Diaspora community where the desire for compromise was strong." W.A. Strange has documented six distinct textual forms of the decree in the manuscript tradition, showing it was a living document that communities continued to revise — not a fixed ruling handed down at a single event. The earliest revisions are revealing: the Western text (D) drops "strangled" and adds the Golden Rule, transforming the decree from a purity regulation into a moral code. This transition from ritual to ethical interpretation is itself evidence that the real issue behind these evolving food laws was likely practical — governing whether Jewish and Gentile believers could share the Eucharistic meal — not establishing universal moral prohibitions.10
Paul "would have nothing to do with such compromises," as Pervo states. The absence of the Apostolic Decree from Paul's letters is devastating to Acts' credibility. In 1 Corinthians 8–10, Paul conducts an extended, agonizing theological argument about whether believers may eat meat sacrificed to idols (εἰδωλόθυτα — the exact same word used in Acts 15:29). If a binding apostolic decree on this precise question already existed, Paul's argument would have required a single sentence: the apostles have already ruled. Instead, Paul argues entirely from first principles — Christian liberty, the weakness of conscience, the nature of idols — without any reference to a Jerusalem ruling.11
🔗The Antioch Incident
Acts entirely omits the Antioch incident (Gal 2:11-14), in which Peter withdrew from table fellowship with Gentile Christians after messengers from James arrived, and Paul rebuked him publicly. This confrontation — which Paul considered significant enough to record in detail — is incompatible with the portrait of apostolic harmony that Acts constructs.
As Philipp Vielhauer argued in a landmark analysis, the author of Acts "is pre-Pauline in his Christology, and post-Pauline in his natural theology, concept of the law, and eschatology" — the theological portrait has been systematically revised. The author of Acts, Pervo observes, knew Paul's letters but "declined to consult them." The decree letter in Acts 15 is modeled on Greco-Roman official correspondence — George Kennedy compares it to the rescript of a Roman magistrate responding to a query from a subordinate. The author knew Paul wrote letters but deliberately avoided associating Paul with letter-writing, possibly because of what Pervo calls the "Paulinists" who would later produce the trajectory that led to Marcion.12
🔗Internal Contradictions
Acts does not merely contradict Paul — it contradicts itself. The Damascus road conversion is narrated three times in Acts (chapters 9, 22, and 26), and the three accounts diverge in ways that cannot be harmonized.
What did Paul's companions do? In Acts 9:7, they "stood speechless" (εἱστήκεισαν ἐνεοί). In Acts 26:14, "we all fell to the ground" (πάντων τε καταπεσόντων ἡμῶν εἰς τὴν γῆν). They cannot have both stood and fallen.
Did the companions hear the voice? In Acts 9:7, they are described as "hearing the voice" (ἀκούοντες μὲν τῆς φωνῆς). In Acts 22:9, "they did not hear the voice of the one speaking to me" (τὴν δὲ φωνὴν οὐκ ἤκουσαν τοῦ λαλοῦντός μοι). Attempts to resolve this by distinguishing between the genitive and accusative cases of ἀκούω (akouō) fail. As C.F.D. Moule concluded: "it seems to me (pace Proleg. 66) impossible to find a satisfactory distinction in meaning between the Gen. and Acc. in Acts ix. 7 and xxii. 9." H.R. Moehring agreed: "Under these circumstances to insist upon a difference of meaning in Acts ix 7 and xxii 9 seems, to the present writer at least, impossible."13
Barrett's verdict is direct: "We must conclude that Luke was writing up a familiar story freshly, and in each case included what seemed to him to be impressive details in the most impressive way he could think of."14
Pervo's source-critical analysis reveals deeper problems. The story behind Acts 9 is a pre-Lucan "conversion of enemy alien" narrative. He identifies Joseph and Aseneth 14 and Apuleius's Metamorphoses 11 as the closest formal parallels — eight common elements including supernatural light, falling prostrate, the double vocative address, the "Who are you?" / "I am..." identification formula, and the command to rise. This literary form is designed for converting an enemy of God — hardly appropriate for a Jewish persecutor who already worshipped the God of Israel.15
A critical detail often overlooked: Acts 9 does not actually narrate a vision of the risen Jesus. Light floods around Paul (v. 3), he hears a voice (v. 4), but he never sees anyone. Yet Ananias in v. 17 refers to "Jesus, who appeared to you." Every subsequent retelling (chapters 22, 26, and later apocryphal accounts) tries to fill this gap.15
Conzelmann's Hermeneia commentary adds that the legal framework of Acts 9 is historically impossible. The high priest's letters authorizing Paul to arrest followers of Jesus in Damascus have no basis in reality — the Sanhedrin had no extradition authority in Roman Syria. Conzelmann notes that 1 Maccabees 15:16-21 is sometimes cited in defense, but "its validity for Syria during the Roman period is completely out of the question."16
Conzelmann also identifies a Greco-Roman literary parallel to the some-saw-some-heard contradiction. Maximus of Tyre describes sailors at a shrine where "some by no means saw him, but heard him singing... others both saw and heard." The motif is a standard literary convention for narrating divine epiphanies, not a historical report.16
The three versions progressively clean up the narrative mess: Acts 9 is a hybrid punishment-conversion story with gaps everywhere (no vision narrated, no commission delivered, no gospel taught); Acts 22 adds that companions "saw the light" but "did not hear the voice," reversing 9:7; Acts 26 eliminates Ananias entirely and has Jesus deliver the commission directly on the road. Each retelling is smoother than the last — the telltale signature of literary composition, not historical memory.15
🔗The "We" Passages
Portions of Acts shift to first-person plural narration — the so-called "we" passages (16:10-17; 20:5-15; 21:1-18; 27:1-28:16). These are sometimes cited as evidence that the author was an eyewitness companion of Paul. The evidence points elsewhere.
Vernon K. Robbins demonstrated that first-person plural narration was a standard convention in ancient sea voyage accounts, used as a stylistic device to increase vividness rather than to indicate actual participation.17 Susan Marie Praeder's subsequent analysis of first-person narration in Acts confirms this literary function.18
Pervo deepens this analysis. The "we" in Acts is omniscient. When the Western text (D) revises Acts 16:10, it has Paul telling his companions about his vision, after which "we perceived that the Lord had called us." The standard text skips this — "we" simply knows about Paul's private vision without being told. Pervo characterizes this narrator as an "omniscient 'we'" and an "omniscient first plural."19
The "we" also follows a recognizable literary pattern. Acts 16:6-40 conforms to the structure of ancient cult-foundation stories: dream vision, interpretive oracle, lodging in domestic quarters, opposition by officials, miraculous vindication, further growth, and permanent establishment. Paul's Macedonian vision parallels Caesar's vision at the Rubicon (Suetonius, Julius 32) and Alexander's vision of the Jewish high priest. Readers of Acts would understand Paul as an individual of the status of Alexander or Caesar.19
Conzelmann's Hermeneia commentary provides the clearest statement on the voyage narrative. Acts 27 is "the only real account of travel experiences in Acts" — its literary quality exceeds the rest of the book. Paul can be excised from it without leaving gaps in the narrative: "indeed, at some points only in this way does the course of the narrative become understandable." "The first-person plural is a typical device in these descriptions," Conzelmann concludes — not evidence of eyewitness participation.20
Even if one assumes the "we" passages reflect a genuine travel diary integrated by a later author, their scope is extremely limited. The first-person plural appears only during travel itineraries. It vanishes entirely during the disputed theological sections — the Damascus road conversion, the Jerusalem Council, Paul's private meetings, the speeches, and the early church summaries. The companion hypothesis, as Conzelmann notes, does not even work internally: Timothy and Silas were already with Paul before the "we" appears at 16:10, so the "we" cannot mark the arrival of a new companion.20
🔗Greco-Roman Literary Conventions
The patterns described above — invented speeches, idealized narratives, dramatic embellishment — align with established Greco-Roman literary conventions. Richard Pervo's Hermeneia commentary demonstrates that Acts employs the specific tropes of the Hellenistic ancient novel and popular literature: shipwrecks, dramatic escapes, divine interventions, and utopian community descriptions. The decree letter in Acts 15:23-29, Kennedy argues, "resembles the rescript of a Roman magistrate responding to a query from a subordinate" — guilds and private organizations routinely imitated the style of municipal decrees as a mark of culture and civic participation.21
On a more specific textual level, Ruben Dupertuis examines the summary passages in Acts 2, 4, and 5 — descriptions of the early Christian community sharing possessions, holding all things in common, and distributing to each as any had need — and demonstrates their close resemblance to Greek utopian literary traditions.22 These idealizing community descriptions follow a recognizable pattern from ancient literature: the depiction of a golden-age community characterized by unity, shared property, and the absence of want.
Furthermore, nearly thirty percent of Acts consists of speeches placed in the mouths of the apostles — a proportion noted since Cadbury (1933) and Dibelius (1956). Following standard Greco-Roman historiographical conventions (as established by Thucydides), ancient authors composed speeches that fit the character and the occasion rather than recording what was actually said. As Vielhauer demonstrated, the theology of the speeches in Acts reflects the author's later perspective rather than the theology of the characters to whom the speeches are attributed.23
This does not necessarily mean the author invented every event from whole cloth. Proponents of traditional authorship point out that Acts often correctly identifies incidental details of first-century Greco-Roman geography and civic titles — the politarchs in Thessalonica, the asiarchs in Ephesus, the proconsul in Cyprus — demonstrating the author's general familiarity with the Mediterranean world. But familiarity with a setting is not eyewitness participation in the specific events described. The literary presentation of events in Acts was shaped by conventions that had nothing to do with historical reporting.
🔗Implications
The stylometric evidence provides quantitative support for what literary critics have argued for over a century: Acts was not written by the author of the Core Gospel. The historical and literary evidence — contradictions with Paul's letters, internal inconsistencies, the "we" passage conventions, and the second-century authorship chain — independently demonstrates that its narrative cannot be treated as historically reliable testimony.
The scholars cited in the sections above — Barrett, Vielhauer, Conzelmann, and Pervo — reached the conclusion that Acts is unreliable as history, but none proposed removing it from the canon. That step follows from the UPDV's own editorial standard, which requires a tighter mooring to primary testimony than these scholars required. A reader who accepts the first two premises — different authorship and historical unreliability — can still reject the third, and the UPDV's standard is the point where those paths diverge.
Any teaching of the Bible that depends on Acts as a source needs to be re-evaluated against the texts that survive independent scrutiny. Doctrines built primarily on Acts — including specific understandings of the Holy Spirit's arrival, the structure of early church governance, and the normative pattern of conversion — should be tested against what Paul, the Gospel authors, and the other epistle writers actually wrote. Where Acts is the sole source for a teaching, that teaching rests on a foundation the UPDV does not consider reliable.
For a detailed examination of how the exclusion of Acts affects specific doctrines, see Speaking in Tongues: What Changes Without Acts?.
🔗Appendix: Computational Methodology
This appendix details the parameters and validation tests underlying the computational stylometry claims in this article, for readers who wish to evaluate or replicate the analysis.
🔗Corpus and Sliding Window Parameters
All Greek texts were sourced from the NA28 critical edition (New Testament), Rahlfs' Septuagint, and Niese's edition of Josephus. Punctuation and capitalization were normalized to isolate the syntactic signal. Corpus sizes:
- Gospel of Luke: 19,482 tokens
- Acts of the Apostles: 18,450 tokens
- Gospel of Matthew: 18,346 tokens
- Gospel of Mark: 11,304 tokens
- LXX 1 Samuel: 9,933 tokens
- Josephus, Jewish War (3,726-token sample — used for Delta matrix): 3,726 tokens
- Josephus, Jewish War (full text — used for τε rate): 125,274 tokens
- Josephus, Antiquities 1–11 (register control): 170,831 tokens
Stylometric profiling used Burrows' Delta and Principal Component Analysis on a predefined list of 70 high-frequency Koine Greek function words — particles, conjunctions, prepositions, and pronouns. Because function words dictate grammatical structure rather than narrative content, they reflect an author's unconscious structural habits and are resistant to deliberate imitation.
The corpora were processed using a sliding window of 1,000 words with a step size of 100 words, generating 756 discrete windows across the full corpus. For transition zones (Luke 22–24), the methodology shifted to verse-by-verse boundary scanning, which detected a gradual rise in Luke 1–2 affinity from approximately 20% at the Last Supper to 47% at the Emmaus narrative, with the single largest jump occurring at Luke 23:9 within the Herod pericope.
🔗Validation: Source-Layer Isolation (L-Material Test)
A natural objection to the register finding is source reliance: perhaps the Core Gospel's distinctive profile is merely the fingerprint of Mark and the sayings source shared with Luke, not of the author. If so, the author's uniquely composed material — passages found only in Luke, known as L-material (including the parables of the Good Samaritan and the Prodigal Son) — should cluster with Acts rather than with the Core Gospel.
To test this, L-material, passages derived from Mark, and passages derived from the shared source were source-classified using standard critical scholarship and isolated from Luke chapters 5–21 only, avoiding the contested transition zones.
Result: 95% of L-material passages are closer to the Core Gospel centroid than to Acts (passage-level classification, t = 7.191, p < 10⁻¹²). The Core Gospel's low-τε profile is the author's natural voice, not an artifact of copying sources.
Leave-out validation: To guard against circularity (L-material pulling the Core centroid toward itself), the baseline was recalculated using only material derived from Mark and the shared source. L-material remained significantly closer to this independent Mark/shared-source centroid than to Acts (passage-level, t = 3.319, p < 0.001).
🔗Validation: Deep Syntax (Septuagintal Interference Test)
The most striking anomaly in the data is the affinity between Luke 1–2 and LXX 1 Samuel. A reasonable objection is that this clustering is driven by a few conspicuous "biblical-sounding" words rather than deep structural similarity.
To test this, five hyper-frequent Septuagintal markers were stripped entirely from the analysis:
1. ὅτι (hoti, "that/because")
2. ἐγένετο (egeneto, "it came to pass")
3. ὡς (hōs, "as/like")
4. ἰδού (idou, "behold")
5. ἐνώπιον (enōpion, "before/in the sight of")
After this ablation, Luke 1–2 remained 3.6 times closer to LXX 1 Samuel than to the Core Gospel in PCA Euclidean distance (1.523 vs. 5.456). The affinity is structural — embedded in the base prepositions and conjunctions rather than a surface veneer of biblical vocabulary. Whether it reflects a different authorial syntax is a stronger claim than this sample size can resolve.
🔗Validation: Register Control (Josephus Antiquities vs. War)
A persistent objection to any function-word-based authorship argument is that genre and register — not authorship — drive the observed differences. To isolate the effect of register on τε, the central diagnostic marker, the full Greek texts of Josephus's Antiquities (books 1–11) and Jewish War (books 1–7) were compared using the same tokenization and frequency-counting methodology applied to the New Testament corpus.
Josephus is indisputably a single author. Antiquities 1–11 is Septuagintalizing biblical narrative — the precise operation that the "Luke imitating LXX style" hypothesis posits. Jewish War is contemporary military historiography. If register drives τε rates, the same author's frequency should drop sharply across this boundary.
Result: Josephus's τε rate in Jewish War is 10.70 per thousand words. In Antiquities 1–11 it is 8.23 — a reduction of 23%, or a factor of 1.3. The Core Gospel author's τε rate is 0.43; the Acts Author's is 8.18 — a factor of 19. A 1.3× register effect against a 19× observed gap is more than ten times short of what the register hypothesis requires, measured from a large (170,831-token) sample of the closest available analogue. The register explanation for the τε gap is not viable at the observed magnitude.
This test also addresses the objection that function words calibrated on modern authors may not resist imitation by ancient authors working toward a memorized model text. Josephus is indisputably imitating Septuagint style in Antiquities 1–11 — yet his τε rate barely budges. The field's premise that function words resist deliberate alteration holds in this case.
🔗Notes
1 Burrows, John. "Delta: A Measure of Stylistic Difference and a Guide to Likely Authorship." Literary and Linguistic Computing 17, no. 3 (2002): 267-287. See also Eder, Maciej. "Does Size Matter? Authorship Attribution, Small Samples, Big Problem." Digital Scholarship in the Humanities 30, no. 2 (2015): 167-182.
2 Original research. See the Appendix for full corpus parameters, validation tests, and replication details. The method follows Burrows (2002) and Eder (2015).
3 Argyle, A.W. "The Greek of Luke and Acts." New Testament Studies 20 (1974): 441-445.
4 Pervo, Richard I. Acts: A Commentary. Hermeneia. Minneapolis: Fortress Press, 2009. Introduction, section 3: "Date, Provenance, and Author."
5 Taylor, Vincent. Behind the Third Gospel. Oxford: Clarendon, 1926. See also Knox, John. Marcion and the New Testament. Chicago: University of Chicago Press, 1942.
6 Barrett, C.K. A Critical and Exegetical Commentary on the Acts of the Apostles. 2 vols. International Critical Commentary. Edinburgh: T. & T. Clark, 1994, 1998.
7 Barrett, ICC Acts, vol. 1, commentary on 9:19b-30.
8 Barrett, ICC Acts, vol. 1, commentary on 9:19b-30: "It is impossible to doubt that the same incident is in mind, and that Paul's version of it is more likely to be correct than any other."
9 Pervo, Hermeneia Acts, commentary on 15:1-35.
10 Pervo, Hermeneia Acts, Excursus: "The Apostolic Decree." See also Strange, W.A. The Problem of the Text of Acts. Cambridge: Cambridge University Press, 1992.
11 The word εἰδωλόθυτα ("things sacrificed to idols") appears in Acts 15:29 as a binding prohibition and in 1 Corinthians 8:1 as an open question Paul argues from first principles — without any reference to a Jerusalem decree.
12 Vielhauer, Philipp. "On the 'Paulinism' of Acts." Evangelische Theologie 10 (1950-51): 1-15. Reprinted in English in Studies in Luke-Acts, ed. Leander E. Keck and J. Louis Martyn (Nashville: Abingdon, 1966), 33-50. Pervo, Hermeneia Acts, commentary on 15:1-35. Kennedy comparison cited in Pervo.
13 Moule, C.F.D. An Idiom Book of New Testament Greek. 2nd ed. Cambridge: Cambridge University Press, 1959, p. 36. Moehring, Horst R. "The Verb AKOYEIN in Acts IX 7 and XXII 9." Novum Testamentum 3, no. 1–2 (1959): 80–99, at 98.
14 Barrett, ICC Acts, vol. 2, commentary on 22:9.
15 Pervo, Hermeneia Acts, commentary on 9:1-19a.
16 Conzelmann, Hans. Acts of the Apostles. Hermeneia. Philadelphia: Fortress Press, 1987. Commentary on 9:1-19a and accompanying notes.
17 Robbins, Vernon K. "By Land and By Sea: The We-Passages and Ancient Sea Voyages." In Perspectives on Luke-Acts, ed. C.H. Talbert, 215-42. Edinburgh: T. & T. Clark, 1978.
18 Praeder, Susan Marie. "The Problem of First Person Narration in Acts." Novum Testamentum 29 (1987): 193-218.
19 Pervo, Hermeneia Acts, commentary on 15:36-16:10.
20 Conzelmann, Hermeneia Acts. Introduction on the "we" source; Excursus on the voyage narrative (Acts 27).
21 Pervo, Hermeneia Acts, commentary on 15:1-35. The earlier treatment in Pervo, Richard I. Profit with Delight: The Literary Genre of the Acts of the Apostles (Philadelphia: Fortress Press, 1987) is superseded by the Hermeneia commentary.
22 Dupertuis, Ruben Rene. "The Summaries in Acts 2, 4, and 5 and Greek Utopian Literary Traditions." PhD diss., Claremont Graduate University, 2005.
23 On the proportion of speech material in Acts, see Cadbury, Henry J. "The Speeches in Acts." In The Beginnings of Christianity, vol. 5, ed. F.J. Foakes Jackson and Kirsopp Lake (London: Macmillan, 1933), 402–427; Dibelius, Martin. Studies in the Acts of the Apostles, ed. Heinrich Greeven (London: SCM, 1956). On the convention of invented speeches in ancient historiography, see Thucydides, History of the Peloponnesian War 1.22.1. Vielhauer demonstrated that the theology of Acts' speeches is inconsistent with the historical theology of the characters speaking; see note 12.