Many issues related to the selection, reliability, and validity of mental stress testing in cardiovascular research are discussed. Five categories of mental stress testing are distinguished: problem-solving tasks, information-processing tasks, psychomotor tasks, affective conditions, and aversive or painful conditions. A series of practical and theoretical criteria are outlined for the selection of appropriate tests, and the measurement of a range of dependent variables is emphasized. The temporal stability of cardiovascular responses to mental stress tests is examined through an analysis of test-retest correlations (weighted for sample size) in 28 comparisons with intervals between sessions varying from 1 day to more than 1 year. Heart rate reactions to tasks show an average-weighted Z of 0.732 +/- 0.031 (r = 0.62), with Z = 0.575 +/- 0.034 (r = 0.52) for systolic blood pressure and Z = 0.313 +/- 0.035 (r = 0.30) for diastolic blood pressure. It is argued that the validity of mental stress tests can be judged in relation to several different aspects, specifically, methodological, ecological, diagnostic, prognostic, and therapeutic validities. The nature of these standards is described, and pertinent literature is presented.