Previous attempts to validate protective tests have been criticized for being either too atomistic or too holistic [1,4,5]. In the former case, elements of the raw data are counted and correlated with something, a process in which objectivity is retained, but the meaning and richness of the data are lost. The holistic methods, such as blind matching, indicate that there is something valid about the combination of projective test, interpretation, and clinician, [7]. They retain the richness of the raw data. However, nothing is learned from blind matching studies about the specific strengths and weaknesses of the interpretative process. Several authors have suggested that what needs to be studied, at present, are the separate interpretative statements which make up a full test interpretation [1,3,4,5,6]. This approach will have the shortcoming that the context will be lost, but it promises to be fruitful in providing clues as to how statements in test interpretations could be improved, in order to avoid semantic confusions. An interest in studying the separate interpretative statements was the starting point for this study.The validity of projective test interpretations cannot be demonstrated unless there is a high degree of inter-interpreter reliability in the making of interpretations. The usual measure of inter-rater reliability involves the average per cent agreement or average intercorrelation between each possible pair of