Short definition

The state of research in which new iterations bring only confirmation or redundancy, without changing the representation that has been built.

Full definition

The term comes from qualitative research: in grounded theory, theoretical saturation is reached when new interviews or new observations no longer bring out any new category.

In the articles on this blog, it means the same thing applied to a document corpus, with two clarifications. Saturation is always relative to a mission and a scope, never to the subject in general. And it is observed on the extracted knowledge — notes created, notes changed, contradictions revealed — not on the number of documents read.

Usage in the field

Used when deciding to stop desk research, a prior-art search or a loop handed to an agent.

Synonyms and variants

"convergence", "zero yield from new sources". "Convergence" emphasises the model settling down, "saturation" the drying up of new contributions.

Not to be confused with

• Stopping criterion — a condition set in advance that decides a piece of work is finished. Saturation is an observed state; the stopping criterion says from which observation you stop.
• exhaustiveness — complete coverage of an enumerated set. Saturation is no proof of it: it records a yield, not completeness.

Examples

"At the fifteenth iteration, no new source important enough to change the conceptual model of the five themes was retained."

Ambiguities / debates

The threshold isn't fixed: how many iterations without effect does it take to speak of saturation? The usage adopted here requires several failed break searches, not just a drying up.