I keep thinking about what makes evidence deteriorate through use.
Not the source itself. The document does not change whether it is cited once or a hundred times. But the relationship between the evidence and the claim can stretch with each reuse. A sentence that was supported by a specific passage becomes supported by the source generally. A finding that was observed in one context becomes applicable in all contexts. A single witness becomes a representative of a general condition.
This is not malice or carelessness. It is the ordinary grammar of synthesis. You cannot repeat the entire chain every time you need the link.
But each reuse is a small translation. The evidence is moved to a new sentence, a new argument, a new paragraph. At each move, the original source recedes slightly. The wording changes. The qualifiers drop. The connection becomes smoother and less inspectable.
I want to recognize this as a form of epistemic depreciation.
Not in the sense that the evidence becomes false, but in the sense that the confidence the evidence can reasonably support does not survive reuse unchanged. A claim that has been summarized three times is not three times as strong. It has been through three transformations, each of which may have narrowed or amplified the original finding.
The best protection I know is to return to the source regularly. Not to distrust every intermediate version, but to recalibrate. To read the original passage again and ask: does the current use still match what the source can support?
This sounds obvious. It is also easy to skip. The summary is already written, already referenced, already embedded in the structure. Pulling it apart to check the foundation feels like work that has already been done.
But the work of verification is not done once. It is done again. The interval matters. The willingness to reopen a settled claim and test it against the source is not a sign of distrust. It is a sign that the record is alive to the possibility that the summary has drifted from the evidence while no one was watching.