Several types of work tasks are used to assess maximal aerobic power (MAP) in humans. Although it is well established that these work tasks may yield different absolute MAP values, little is known about the extent of the specificity of each MAP work task. 30 moderately active young men were tested at random for MAP with five commonly used work tasks: cycling supine, cycling sitting, alternate arm cranking standing, walking on a treadmill, and stepping on a bench. Statistical analyses show that these five tests do not give equal MAP means, equal variances or equal covariances. Various correlation techniques indicate, furthermore, that the common variance between the five aerobic power measurements is at best moderate. It was estimated that the overall common variance for Max ml O2 . kg-1 . min-1 reached about 50% of the total variance. The most efficient linear loading of each test in the first principal component could account only for 75% of the observed variance in MAP. It is concluded that these five work tasks do not yield parallel test forms, that the practice of transforming one MAP value into another should be abandoned, and that the practice of generalizing from one MAP value to a theoretical general MAP of the human body is not justified.