An empirical law or evidence which supports a theory tends to have the greatest scientific value when it seemed improbable before it was investigated. Evidence which falsifies a theory tends to have the greatest value when it seemed probable that the investigation would confirm the theory. A scientist who wishes to optimize his contribution to knowledge probably will investigate most frequently hypotheses which seem neither very probable nor very improbable of being confirmed. This strategy leads neither to a maximum number (...) of confirmations of a theory, nor to a set of tests which are maximally severe. (shrink)