A number of areas of social and personality psychology have focused on relations-interpersonal relations and person-object relations-as basic units of information. These relations may take the form of behaviour in which x acts on y , or the form of opinions or attitudes,2 in which x is in a cognitive-affective relation with (liking, approving 09 y. Relations tend to be regarded as basic in the sense of being prior to and responsible for trait attributions, evaluations, etc. In particular, the loose collection of theoretical analyses which have been grouped together under the title 'consistency theory' (e.g., Abelson, et al., 1968), deal with patterns of inter-relationships among people, objects, and events. The congruence of the grouping of objects within this network of relations is regarded as fundamental to questions of relational stability, attitude change, etc.Considering the fact that relational links between an individual other persons or objects play a major role in a number of theoretical formulations concerning attitudes, cognitive structure, and patterns of attribution and evaluation, it is surprising that the formal analysis and classification of these relations has received only minor attention in social psychology. T w o systematic attempts to classify such relations-Heider's sentiment-unit classification scheme and the affective-influence categorization stemming from research on cognitive schemata-will be reviewed and evaluated here. Both forms of categorization will be shown to be inapplicable to a wide range of potential relations. It is suggested that the limited focus of much of the consistency research on one class of relations-likes-dislikes-has resulted from the lack of a broad theoretical conceptualization of the properties of relations. Such a formulation, based upon an analysis of the components of relations, is proposed and detailed. J. Theory SOL. Behaviour 7 , 2 , Printed in Great Britain