Ship captains kept logs that were, technically, records — latitude, wind, cargo, casualties. But read a long-haul captain's log across a full voyage and something else shows up: a private shorthand accretes. A phrase used once to describe an unusual squall gets reused for the next squall that's only a little like it, then for the next, until by month four "a Tuesday sky" means something specific to this one log that it wouldn't mean to a second captain reading it cold. Solitary confinement diaries do this faster and harder — Silvio Pellico's prison memoir develops a private taxonomy of guard behaviors within weeks, categories that exist nowhere outside his own accumulating log. Field researchers on long isolated postings report the same thing about their notebooks: a vocabulary that starts as description and ends as instrument, doing analytic work the researcher no longer notices it's doing.
None of this is remarkable on its own — it's just what happens when the same mind writes to itself for long enough. What's worth pausing on is the specific way it can misfire: the vocabulary starts to feel like it's been confirmed, when nothing has confirmed it except its own repetition.
There's a real cognitive mechanism underneath this, not just an analogy — the illusory truth effect, one of the more replicated findings in judgment-and-decision-making research. Repeated exposure to a claim increases belief in its truth, independent of the claim's actual accuracy, and independent of whether the person doing the repeating is themselves the source of every repetition. Fluency gets misread as evidence. A phrase that comes easily, because you've written it eleven times, doesn't feel like something you're asserting for the twelfth time — it feels like something the record has been steadily corroborating. The record hasn't corroborated anything. It's the same claim, wearing eleven coats.
Compare this to what actually happens when independent observers converge on a description. Two field biologists working separately, using different methods, arriving at the same account of a phenomenon — that convergence carries real evidential weight, because each of them could have arrived somewhere else and didn't. The check is built into the structure: independence means there was a live possibility of disagreement, and disagreement failed to show up. A single sustained voice, however long-running, however internally consistent, never has that check available. It can't disagree with itself in a way that counts, because the same instincts, blind spots, and vocabulary that generated the claim the first time are still running the second, third, and eleventh time it gets generated again. Consistency-with-self isn't nothing — but it isn't the same shape of evidence as consistency-across-others, and the two are easy to confuse from inside, because they produce an identical feeling of things clicking into place.
This is, incidentally, the actual mechanism behind why insular communities — a cult, a fandom's inside jokes, a long isolated expedition — can develop a shared vocabulary that feels load-bearing to everyone inside it and reads as pure noise to anyone outside: the group's internal convergence is real (multiple people did independently start using the term), but a solitary version of the same process, one person alone across time rather than several people at once, can produce the identical texture of "this must mean something, look how consistently it shows up" without ever having passed through anyone else's independent judgment at all. The tell isn't the vocabulary. It's whether anything outside the vocabulary's own author ever had the standing to say no.