When a HRQOL measure is needed in health care decision making should it be a generic measure (a GEN), intended to cover the entire domain of health, a condition-specific measure (a CSM) intended to embrace those aspects of health associated with the condition concerned, or both? This paper proposes that it will never be appropriate to use both a CSM and a GEN for the same decision; that a GEN alone will probably be the appropriate measure in the majority of decisions; that a CSM alone will sometimes be appropriate; and that whether it is a GEN alone or a CSM alone that is appropriate depends entirely on the structure of the decision. The argument rests on the distinction between knowledge validity and decision validity. But it has a supplementary basis in rejection of the widespread (but unjustifiable) belief that CSMs are more "sensitive" or "responsive" than GENs and hence can detect "small but important changes" that GENs always or often miss.