Laboratory tests are purported to affect patients even if they have no diagnostic values. We tested this hypothesis by measuring clinical outcomes of 176 patients thought clinically to have nonspecific chest pain. They were randomly allocated either to have a routine electrocardiogram and serum creatine phosphokinase tests (test group) or to have all diagnostic tests withheld (no-test group). Fewer patients in the tests group (20%) reported short-term disability after the index visit than patients in the no-test group (46%) (p = 0.001). Logistic discriminant analysis confirmed that the use of diagnostic tests was an independent predictor of recovery. Patients in the test group felt that care was "better than usual" more often (57%) than patients in the no-test group (31%) (p = 0.001). After the index visit, the two groups were equally worried about serious disease and equally sparing in their use of other medial care for chest pain.