When to stop collecting evidence

More information is not always safer when obtaining it delays a treatment whose benefit falls quickly. I decide in advance which finding would change the immediate plan and which would merely refine the explanation later. Tests become a form of avoidance when no possible result alters action.

I explain this distinction to families because stopping investigation can otherwise sound like indifference. The responsible endpoint is tied to a decision, not to the fantasy of complete knowledge. Before ordering another test, I write down the result that would change treatment now.

If no plausible finding alters the immediate plan, the investigation pauses and the record names the symptoms or response that would restart it. Another physician can therefore continue promptly when circumstances change without mistaking the earlier endpoint for indifference or for a claim that the diagnosis was complete. The stopping rule is written before the final test is ordered: what result would change treatment, and how quickly?

If neither answer can be stated, more investigation may satisfy anxiety while making the patient’s actual decision worse.

An archive can always offer another box, but additional documents do not automatically improve an argument. I stop when new material repeats the same relationship without changing its range or exceptions. A dramatic late discovery can still matter, so the endpoint must remain revisable.

My notes state which kinds of evidence might reopen the search. Closure is a reasoned pause, not a declaration that the archive has nothing else to say. When I close a search, my notes identify what new relationship the last boxes established and what kind of source would genuinely reopen the argument.

Another historian can distinguish a dramatic but repetitive discovery from evidence that extends the social range, chronology or exception set. The pause remains conditional, yet it protects the project from making accumulation look like interpretation. I close an archival search provisionally and name the discovery that would reopen it.

This keeps the argument revisable without pretending that another box must always be consulted before any historical claim can responsibly be made.

Fault-finding can expand until every component has been tested, even when the customer’s problem has already been reproduced and isolated. I compare the cost of another test with the risk of replacing the wrong part. A cheap measurement may still be wasteful if it cannot discriminate between remaining causes.

I stop when the repair has a clear verification procedure and a safe fallback. The endpoint includes how we will know the diagnosis was adequate after the vehicle leaves. The job sheet states the reproduced fault, the test confirming the remedy and the safe fallback if the symptom returns.

Once those three elements are in place, further component checks may add explanation but no longer answer the customer’s problem. A later mechanic can resume diagnosis from the recorded boundary instead of repeating an unlimited search or replacing parts without a discriminating reason. A repair is closed only after the reproduced fault disappears under the conditions that first produced it.

I also leave a fallback and a threshold for reopening the diagnosis, so stopping does not mean declaring the machine permanently understood.

A story becomes vulnerable when reporting stops only because the deadline arrives. I build a claim table showing which assertion depends on which source and where independent confirmation is still missing. Some gaps weaken a detail; others undermine the whole public allegation.

Publication is justified when the unresolved uncertainty is accurately represented and the central claim no longer rests on one interested account. Stopping is an editorial judgement about exposure, not exhaustion. I stop gathering interviews or documents when new material no longer changes the central allegation, its evidential support or the response owed to people named in the story.

The handover memo lists unresolved questions and marks which could still alter publication. A later editor can reopen a consequential issue while recognising that additional colour or background alone does not justify indefinite collection. Before publication, every central claim is linked to its strongest source and to the uncertainty that remains.

Additional reporting ends when it changes neither wording nor prominence, not merely because the calendar has run out.