Validating Parallel Similarities and Source Citations Through Thematic Chaining
My writing on Medium usually focuses on two distinct areas: AI epistemology and biblical textual criticism. Until now, those subjects haven’t crossed. But they intersect today because of a recurring problem in textual analysis: when a later author repurposes symbolism, motifs, or allusions from older literature, how do we determine which specific text served as the true progenitor when multiple candidates fit the bill?
Admittedly, this sounds like a niche historical puzzle. Yet the underlying epistemic challenge appears in many places. Whether you are untangling copyright infringement or patent prior art, performing historical linguistics, auditing forensic evidence, or catching AI hallucinations in a legal brief, the core difficulty is identical, which is distinguishing genuine structural lineage from superficial resemblance. We need an epistemic guardrail that forces rigorous source validation.
The First Problem: Parallelomania
In his 1961 presidential address to the Society of Biblical Literature, published the following year as “Parallelomania” (Journal of Biblical Literature 81, 1962), Samuel Sandmel warned against a habit he saw everywhere in his field: scholars who first overdo the supposed similarity between two passages, and then describe one as the source of the other, as though the direction of borrowing were obvious.
The history of biblical scholarship is full of examples. At the beginning of the twentieth century, Friedrich Delitzsch’s Babel und Bibel lectures traced much of the Old Testament to Babylonian sources. When the Dead Sea Scrolls were discovered, some early interpreters saw the Qumran “Teacher of Righteousness” as a forerunner of Jesus. In each case, real resemblances were inflated into sources, and the conclusions outran the evidence.
The Second Problem: Source Tracing and Citing
The problem is not limited to biblical studies. In copyright law, courts have long had to separate real copying from shared building blocks. When Ed Sheeran was sued over “Thinking Out Loud,” the plaintiffs pointed to its resemblance to Marvin Gaye’s “Let’s Get It On,” but a jury found in 2023 that the shared chord progression and rhythm were common musical elements, not evidence of copying.
In Mata v. Avianca, lawyers in New York were sanctioned in 2023 after filing a brief citing court cases that ChatGPT had hallucinated. The citations looked exactly like real ones, with plausible names, courts, and dates, but nothing connected them to real decisions. The common error is the same in every case. Resemblance is treated as evidence of connection, and each added resemblance is treated as making the case stronger. Without greater epistemic rigor, AI can dig for fool’s gold and think it’s the real thing.
The Solution: Thematic Chaining
So, how do we tell a real source from a mere resemblance? The methodology I developed is what I call “thematic chaining” and it starts from a simple observation: writers and creators don’t borrow isolated details. They borrow ideas that do work in their texts, and those ideas come as linked sequences. Thus, instead of asking how many details two texts share, thematic chaining asks whether a proposed source supplies the whole chain. The test works in steps:
- Isolate the chain. Identify the sequence of linked ideas in the passage being studied.
- Check priority. For each link, ask which source first supplies it. If an earlier text, such as Isaiah or Daniel, already supplies a link, it doesn’t count toward a later one.
- Check for contradiction. If a proposed source conflicts with any link in the chain, it’s ruled out for that chain, however many other parallels it shares.
- Test the joins and the weight. Does the source supply the connections between the links, or only separate pieces? And does the borrowing carry the passage’s argument, or is it just a passing image?
- Compare economy. Prefer the source that explains the whole chain with the fewest added assumptions.
To put it simply, parallelomania counts isolated matching pieces, Thematic chaining tests whether those pieces were assembled in the same structural sequence. This is what I’ve called earned confidence applied to sources: a claimed connection has to survive scrutiny before it’s accepted, rather than being accepted because it sounds plausible and has been repeated often.
Get Vir Multiplicis’s stories in your inbox
Join Medium for free to get updates from this writer.
If you would like to use this framework with your large language model, I have distilled the principles into a GitHub mark down for general textual critical, source citation, as well as specific legal and medical domain usage applications.
A Practical Example: Lincoln and Kennedy
Most readers have seen the famous list of “coincidences” between Abraham Lincoln and John F. Kennedy. Lincoln was elected to Congress in 1846 and Kennedy in 1946. Lincoln became president in 1860 and Kennedy in 1960. Both were shot in the head on a Friday, with their wives beside them. Both were succeeded by a Southerner named Johnson. Kennedy’s secretary was named Lincoln. The list goes on, and it feels uncanny.
Now run it through thematic chaining. Is there a chain? No. Each item is isolated: the dates of elections don’t lead to the day of the week of the assassinations, which don’t lead to the successors’ surnames. Nothing joins one coincidence to the next. And when we check the items themselves, some fail outright. The list usually says John Wilkes Booth was born in 1839, a century before Lee Harvey Oswald, but Booth was born in 1838. It often claims Lincoln had a secretary named Kennedy, but he didn’t; his secretaries were John Nicolay and John Hay. What remains is what we’d expect from two lives with thousands of recorded details: a handful of chance matches, selected after the fact. The list persuades because of its length, not because of any actual connection.
AI’s Natural Parallelism Weakness
In the previous article, 1 Enoch and the Galilean Theological Layer Within the New Testament, I argued that the influence of 1 Enoch is concentrated in New Testament texts with Galilean sources, and I promised that the next article would test the Book of Revelation against that pattern. To test Revelation fairly, I needed a tool that could tell a real source from a mere resemblance, and the reason became clear in a conversation with a friend.
This friend, who holds a more traditional view of Revelation’s authorship than I do, ran our discussion through an AI model and came back with a list of parallels between Revelation and 1 Enoch. The list looked impressive. However, when I traced each item back to its earliest source, most of them dissolved. The details were real, but most came from Isaiah, Ezekiel, and Daniel, and the few genuine contacts with 1 Enoch had nothing to do with its core Watchers story.
For my friend, the AI did what statistical models do by default: it pulled the most heavily discussed resemblances from its training distribution and presented them with fluent confidence, entirely skipping source validation. As I’ve noted in my work on epistemic hygiene and earned confidence, this isn’t an occasional glitch; it’s the structural default of large language models.
The Test: Revelation
Revelation is the New Testament’s most apocalyptic book, full of fallen stars, a bottomless abyss, bound and released spirits, and final judgment. If any book should draw on the Watchers story, it should be this one. Loren Stuckenbruck and Mark Mathews make the strongest scholarly case, arguing that Enochic tradition shaped particular words, phrases, and motifs in Revelation, and some contacts may well be real. Thematic chaining asks a narrower question: does Revelation build on the Watchers framework, or only share imagery with it?
Revelation 20:1–10, the binding of the dragon, is the strongest candidate. Its chain runs from binding and imprisonment in a pit, through release “after a little while,” to Gog and Magog and final destruction. It resembles 1 Enoch 10, where Raphael binds Asael in darkness. But every link already appears in the prophets: prisoners shut in a pit and “after many days” punished (Isa 24:21–22), the same phrase leading into Gog’s invasion (Ezek 38:8), a sealed stone (Dan 6:17), and a beast thrown into the fire (Dan 7:11). And 1 Enoch contradicts the chain. Asael is never released, so 1 Enoch 10 is ruled out as its source.
As the chart above shows, most of the other proposed parallels fare no better. A few are genuine but limited: the seven angels before God are closest to Tobit, the fallen star with the key to the abyss mixes canonical and Enochic elements, and the blood up to the horses’ bridles may echo the Epistle of Enoch. None of them involves the Watchers story. So, the result is clear. Revelation shares some imagery with 1 Enoch, but under thematic chaining its main chains run through Isaiah, Ezekiel, and Daniel, and it has none of the Watchers framework that marks the strongest Enochic register.
Disclaimers
To be clear, Thematic Chaining is a filter, not a complete historical methodology. It narrows the field of viable parallels, but secondary variables like dating, linguistic transmission, and redaction history may require separate proofs. Its priority test assumes the traditional dating of Daniel. Under the critical dating (c. 165 BCE), Daniel still predates the Parables of Enoch but not the Book of the Watchers, so more of the weight falls on Isaiah, Ezekiel, and the contradiction test, which hold under either dating. I’m happy to discuss any of these points in the comments section.
Who Wrote Revelation?
If Revelation’s author was not formed in the Galilean Enochic milieu, then who was he? Its Greek is rough and unedited, and its grammar shows a Hebrew rather than Aramaic substrate. When it labels a word “Hebrew,” the word is actually Hebrew, while the Gospel of John uses the same label for Aramaic words. And its authorial voice is a solitary “I,” in contrast to the collective “we” of the Gospel and 1 John. In the next article, I’ll bring these lines of evidence together to argue that Revelation was written not by the Galilean fisherman John, son of Zebedee, but by another, distinct John of early Christian tradition.
