Although we are slow to learn, even from our mistakes, the frequency with which proper experiments have shattered accepted beliefs in the value of a host of standard preventive, therapeutic, and rehabilitative maneuvers has convinced all but closed minds of the need to rigorously validate new clinical procedures. Strategies for improving compliance, like other clinical maneuvers, consume resources and expose patients to potential toxic reactions and side effects (in the form of preoccupations with medicines, illness, and infirmity). Accordingly, potential compliance-improving strategies should undergo the same experimental testing as new drugs or operations, and the report from Spector and his colleagues elsewhere in this issue (p 36) is an addition to a small but burgeoning group of compliance trials. What should the clinical reader look for (or look out for!) in such reports? The execution of a human experiment, even with random allocation to alternative treatments, is no guarantee of