Main idea

A prior-art search can always carry on: there will always be one more article, one more patent, one more discipline, one more language, one more keyword. No state of progress tells you that you have finished, and waiting for that signal amounts to never stopping.

So the criterion adopted bears not on the volume found but on the effect of the find. A source no longer counts because it resembles what you are looking for: it has to change the map — be older than what you had, be markedly closer, bring together mechanisms that were separate until then, or force a change of conclusion. Everything else is a variant of an antecedent already placed.

When several consecutive sources only improve a date or only add an example, the search has reached saturation. That is not proof of exhaustiveness — it is an observation about yield.

Why it matters

This gives a way out that is neither fatigue nor the impossible demonstration of an absence. Without a criterion, stopping is decided at the moment you get tired, that is, at a moment with no relation to the state of knowledge.

And it disciplines the search itself: asking each source what it shifts prevents you from accumulating references whose only purpose is to show that you searched.

Nuances and limits

The criterion is judged against the map you already have, and therefore against what you have already found. A search that started badly saturates quickly, and its saturation is worth nothing.

And it does not protect against an isolated antecedent, very old and very close, sitting in a corpus you never opened: that one will never change the map, since it will never enter it.

Open questions

• How many sources without effect do you have to string together before concluding that saturation has been reached, and does that number depend on the scale of the subject?