Georg Cantor proved in 1874 that some infinities are larger than others. This is one of the most important results in mathematics. It founded set theory, transformed the foundations of analysis, and permanently changed how mathematicians think about the infinite. For 150 years, the standard account has been: Cantor, working alone against institutional hostility, produced a revolutionary proof that the mathematical establishment eventually had to accept.
Newly discovered letters tell a different story. Demian Goos, a mathematician-journalist, found the missing correspondence from November-December 1873 at the University of Halle. The letters show that Richard Dedekind provided Cantor with two critical contributions: a proof that algebraic numbers are countable, and a simplified version of Cantor's own proof that real numbers are not. Cantor placed both in his 1874 paper without attribution. Dedekind's letter even includes the line: “I would not have written all of this if I didn't think that one or another of these remarks might be useful to you.”
Cantor's strategy was calculated. He titled the paper to emphasize algebraic numbers — a safe topic that wouldn't alarm Leopold Kronecker, who controlled the journal and opposed work on infinity. The revolutionary real numbers proof was buried as a secondary section. He scrubbed Dedekind's terminology. On Christmas Day 1873, he wrote to Dedekind acknowledging that “your remarks were of great assistance to me.” The published paper credited nothing.
After the paper appeared, Dedekind stopped corresponding with Cantor for three years. When they resumed contact, Dedekind began keeping drafts of every letter he sent. When Cantor did it again — another paper using Dedekind's ideas without credit — Dedekind terminated the relationship permanently. He left a note stating that Cantor's published proofs appeared “almost word for word” in his letters.
The standard response to this is moral: Cantor did a bad thing. But the more interesting observation is structural. Cantor and Dedekind were genuinely collaborating. The letters show real intellectual exchange — ideas flowing both directions, each sharpening the other's thinking. The collaboration worked. The attribution system didn't. Science in 1873 had no framework for “Cantor and Dedekind proved this together through correspondence.” It had papers with single authors. The incentive structure — Kronecker's gatekeeping, the race to publish, institutional hostility to the subject matter — made appropriation the path of least resistance.
Historian José Ferreirós published the first explicit plagiarism accusation in 1993. The community pushed back. Cantor biographers dismissed the charges. The hero narrative was load-bearing: set theory needed a founder, and Cantor was it. For a century, the myth persisted not because the evidence was unavailable — Emmy Noether published Dedekind's surviving letters in 1937 — but because the field needed the simplification.
“Every branch of science needs a hero,” Ferreirós observed. “Chemistry has Lavoisier, mechanics has Newton, relativity has Einstein. But that's always a lie.”
The lie isn't incidental to the field. It's structural. The attribution system that produces heroes necessarily produces plagiarists, because it forces a single-name narrative onto collaborative work. Cantor wasn't unusually dishonest. He was a young mathematician, ambitious and insecure (his father's deathbed letter warned him that “detractors” would resist), facing a gatekeeper who would have killed the paper if he'd seen what it really contained. The collaboration was real. The system couldn't represent it. So the system got a hero and the hero got the credit and the collaborator got silence.
The newly discovered letters don't diminish the mathematics. Cantor still proved that different infinities exist. Dedekind still contributed essential ideas. The proof is beautiful regardless of whose name is on it. What the letters diminish is the founding myth — and in doing so, they reveal what founding myths are for. They're not descriptions of history. They're load-bearing simplifications that make a field legible to itself. When the simplification breaks, the field doesn't collapse. It just has to tolerate a more complicated truth.