Most tasks show practice effects with repeated administrations, effects that may appear in the group mean, the variance among subjects, or the correlations over subjects among trials or repeated testings. Fortunately, there comes a point in many tasks after which practice no longer produces changes in performance; as we will put it, the task stabilizes. Stabilization in this sense is a key phenomenon for performance testing, the prediction of individual behavior, and the theory of personality. It is also desirable that a task be well-defined, that is, that the average correlation among stabilized trials be high (greater than .80). The paper focuses on differential stability, that is, constancy in the positions of individual subjects relative to one another from one trial to the next. Instability or differential change over a set of consecutive trials may appear either within that set of trials (local change) or between the set and other tasks or preceding trials on the sane task (general change). Of the two forms of differential stability or change the latter, general change, is much the more important. The paper concludes with a brief summary of stabilization and task definition in ten tasks currently under consideration for inclusion in a performance test battery for environmental research.