A meta-analysis of 35 interventions to reduce diagnostic testing revealed that testing was reduced overall by 17.6%. Stronger effects were found for immediate reductions (31.7%) compared to reduction at follow-up times (18.7%) and stronger effects were found for interventions targeted at specific tests (23.2%) compared to interventions designed to reduce testing in general (11.8%). An examination of the quality of needs assessment shows that interventions planned on the basis of careful analysis of need were more effective than those based on the innovators' intuitive impressions. The value of the use of learning theory in developing effective interventions is discussed. The contrast between the cognitive processes of experts compared to the processes of less experienced decision makers reveals that many interventions may not be conducive to helping medical residents develop their skills in diagnostic testing in the most efficient manner. Several innovative, on-line interventions are described, which include effective interventions that may have lasting impacts since they seem to be based on processes leading to effective learning. One table and one graph are included. (Author/SLD)