Twenty departments of clinical physiology in Sweden, doing annually 30,000 exercise stress tests, mainly of patients, completed a questionnaire regarding how they carried out exercise testing. Bicycle ergometry was predominantely used. The criteria for inclusion of patients for exercise testing and for interruption of the test were generally wide, allowing the patient to work until symptoms limited the test. In a second part of the investigation, the departments continuously reported all complications that occurred during an 18-month period which included 50,000 exercise tests. The complication rate was 18.4, the morbidity rate was 5.2, and the mortality rate was 0.4 per 10,000 tests. The number of complications leading to permanent damage was low and it could not be proved that the exercise test had induced a higher complication rate than otherwise would have occurred during the observation period. Patients with aortic stenosis had a high risk for complications. With adequate safety measures and a well-trained staff, exercise stress testing can be regarded as a safe method to be used in the evaluation of even very ill patients.