Two rather different human factor approaches may be distinguished in efforts to optimize the performance of man-machine systems. One seeks, through the training of the operator, to adjust the human component to the requirements of the system. The other attempts to enhance system performance by adjusting the mechanical elements to fit the man Although much has been accomplished through training programmes, this approach is limited. In the first place, no amount of effort devoted to training operators can make the performance of some man—machine systems adequate to their tasks. A second limitation of the training approach is just beginning to be appreciated. In many instances it is possible, through operator training, to eliminate performance differences among man—machine systems of different intrinsic merit. This might lead one to choose an inferior design in the place of a better one since, under normal operation, they would all appear to be equivalent. However, if the operators were stressed, the fundamental inferiority of the chosen system might reassert itself. The study reported here has, in fact, shown that two different tracking systems, made equivalent through training, revert to their original order of merit when the operators are subjected to any one of a wide variety of ‘ task-induced ’ stress situations.