perform an experiment or measurement, one of whose possible outcomes can be explained by one hypothesis (and accompanying cohort) but not the other.
It is in general very hard to do this because you don't know the membership of the cohort. The Quine-Duhem problem is real.
