Performance tests are an integral component of assessment for competitive cyclists in practical and research settings. Cycle ergometry is the basis of most of these tests. Most cycle ergometers are stationary devices that measure power while a cyclist pedals against sliding friction (e.g. Monark), electromagnetic braking (e.g. Lode), or air resistance (e.g. Kingcycle). Mobile ergometers (e.g. SRM cranks) allow measurement of power through the drive train of the cyclist's own bike in real or simulated competitions on the road, in a velodrome or in the laboratory. The manufacturers' calibration of all ergometers is questionable; dynamic recalibration with a special rig is therefore desirable for comparison of cyclists tested on different ergometers. For monitoring changes in performance of a cyclist, an ergometer should introduce negligible random error (variation) in its measurements; in this respect, SRM cranks appear to be the best ergometer, but more comparison studies of ergometers are needed. Random error in the cyclist's performance should also be minimised by choice of an appropriate type of test. Tests based on physiological measures (e.g. maximum oxygen uptake, anaerobic threshold) and tests requiring self-selection of pace (e.g. constant-duration and constant-distance tests) usually produce random error of at least approximately 2 to 3% in the measure of power output. Random error as low as approximately 1% is possible for measures of power in 'all-out' sprints, incremental tests, constant-power tests to exhaustion and probably also time trials in an indoor velodrome. Measures with such low error might be suitable for tracking the small changes in competitive performance that matter to elite cyclists.