鈥� Accurate detection of unprescribed drug use by addicts in treatment may facilitate their rehabilitation. Many clinics collect urine samples at random, using fixed-interval collection schedules, which are not free from sampling error. Random-interval schedules minimize sampling error and consequently increase detectability of drug use by eliminating safe periods during which drug use cannot be detected. We compared these two methods by observing rates of detected opiate- and quinine-positive samples preceding and following implementation of random-interval schedules. Detected drug use doubled initially. As detection and clinical sanctions became more certain, drug use declined to well below its former level. Programs that use fixed-interval schedules may underdetect drug use by more than 50%. If patients can reliably predict safe periods, the possibility of using drugs without fear of detection may impede their rehabilitation.