AbstractThe confusion of mental testing practice cannot be cleared up by rule‐of‐thumb methods or by reference to the theory of aptitudes. It needs a comprehensive new experimental method, which might be built up around needs and goals in the hierarchy of patterns of relations in society and would be put into effect by the extensive study of situations, rather than persons or traits. Trait rating systems are inefficient because they substitute the indifferent dimensions of the factorial hyperspace for the complexly connected dimensions of behaviour. It is therefore likely that they will fail to meet their own claims, and this is demonstrated by analysis of the results of one system, in which a factorial solution claimed to haw simple structure is found to have a confused and inefficient structure. In the same way, the analysis of abilities has led only to confusion because it has sought for reified factors of the mind instead of trying to classify tests as situations which are similar or different. A method of re‐classification for tests of abilities is suggested. The difficulty of coherent teaching in mental testing is met by showing present practice in its historical setting and in its relation to future structurally designed experi