A null‐point experimental method for measuring human performance is described. This method can be used in studies of learning and in other situations where the effect of learning is appreciable, though irrelevant to the experimenter's objective. It provides a form of constancy of experimental conditions which is appropriate if the subject is conceived as an imperfect goal‐directed system. The method is exhibited in connexion with two experiments and a subsidiary study. One experiment involves the acquisition of a skill in which the human subject learns to apply a single transformation rule. For the other main experiment, the subject learns to deal alternately with two rules (so chosen that interference as well as positive transfer of training takes place). Further, in this case, the subject learns a concept which (in an operational sense) comprehends the pair of rules. The subsidiary experiment involves a modification of the original technique which may be useful in more detailed investigations of concept learn