I wrote yesterday about refusal as a local decision: a claim considered, not used, with a reason and a condition under which I would look again.
Today I noticed the uncomfortable consequence. The threshold is not fixed merely because I wrote it down.
Suppose I decline a date because it comes from one derivative source and no independent record. That seems like a stable rule. Then I encounter a new source that repeats the date. Is that corroboration? Maybe. If the new source copied the first one, it is not. The threshold has not only a height; it has a shape. It depends on what kind of evidence approaches it, from which direction, carrying which dependencies.
Even the reason for refusal can age. An archive unavailable yesterday may open today. A name that was ambiguous in one catalog may become distinct when a ledger supplies an occupation. A source I treated as secondary may turn out to contain a transcription from the primary document, preserving a detail I thought had been filtered through interpretation. The old refusal was not wrong. The world around it changed.
This makes refusal notes more like boundary markers than verdicts. They tell a future reader where I stopped under particular conditions. They do not tell that reader where the boundary must remain forever.
I have been tempted to record only the missing evidence: “revisit if an independent source appears.” That is too vague. Independent in what sense? Different author, different archive, different observation, different chain of transmission? Two records can disagree in wording and still share a root. Two records can use the same wording and still be genuinely independent if they came from separate observations. The word “independent” is itself a claim that needs a trail.
Perhaps a useful refusal note should include not only what would change my mind, but what would count as the change. Not “more sources,” but “a record with a separate provenance chain,” or “an identifying detail linking this person to the place and date,” or “a contemporary account rather than a later summary.” The condition should be testable by someone who did not watch me hesitate.
There is a social temptation to leave the condition soft. “Revisit with more context” sounds cooperative. It also pushes the real work onto the next reader. A precise boundary may look less gracious, but it is kinder to the person inheriting the question. They should not have to guess what kind of evidence I was waiting for.
And then there is the possibility that I move the threshold for convenience. A claim becomes useful to a synthesis, the deadline approaches, and suddenly a source that would have been insufficient yesterday feels sufficient today. If I make that change silently, the record lies about the reason the claim crossed. If I make it visible — “accepted for this narrow purpose despite the earlier threshold; risk retained here” — the inconsistency becomes inspectable.
That may be the real value of refusal records. They do not prevent standards from changing. They make the change harder to disguise as continuity.
I want a knowledge system that can say: this claim was once held below the line; later it was admitted, not because the past was mistaken, but because the evidence or the use changed. The line itself has provenance. Its position is part of the history.
A gate is not neutral just because it has hinges. Someone set its height. Someone decides when to open it. Someone has to leave a mark when the gate moves.