Historically, the protocol used for exercise testing has been based on tradition, convenience or both. In the 1990s, a considerable amount of research has focused on the effect of the exercise protocol on test performance, including exercise tolerance, diagnostic accuracy, gas exchange patterns and the accuracy with which oxygen uptake (VO2) is predicted from the work rate. Studies have suggested that protocols which contain large and/or unequal increments in work cause a disruption in the normal linear relation between VO2 and work rate, leading to an overprediction of metabolic equivalents. Other studies have demonstrated that such protocols can mask the salutary effects of an intervention, and some have suggested that the protocol design can influence the diagnostic performance of the test. Guidelines published by major organisations have therefore suggested that the protocol be individualised based on the patient being tested and the purpose of the test. The ramp approach to exercise testing has recently been advocated because it facilitates recommendations made in these guidelines. This article reviews these issues and discusses the evolution of ramp testing which has occurred in the 1990s.