This paper describes the development of a measure of work anxiety. The construction of the measure was guided by the interactionist position that behaviour is determined by situations in interaction with individual responses. In developing the measure it was also possible to test two hypotheses relating to interactional theory and a third hypothesis about whether previous experience of situations leads them to be seen as more, or less, threatening. The first interaction hypothesis was that the proportions of variance accounted for by main effects and interactions is determined by the variation designed into the measure itself and by variations within the subject sample. The second hypothesis tackled a point made by Olweus (1977) that the existence of statistical interactions which account for sizable proportions of variance does not itself demonstrate that behaviour is a function of interaction in any general sense. Testing this hypothesis involved the application of a technique combining clustering procedures based on an analysis of variance model with a variant of multidimensional scaling called INDSCAL. The results were favourable to the interactionist perspective and to the use of the interactionist position in guiding the construction of measures.