Bannister and Salmon (1966), in a repertory grid study, reported that 'thought-disordered schizophrenics lost significantly more reliability and social agreement when shifted from object to people construing than normal'. This conclusion is of doubtful validity, since the between-grids variable, people vs. objects, was confounded with at least five other variables of possible relevance, three of which have been found by subsequent workers (McPherson & Buckley, 1970; Williams, 1971; Heather, 1976) to affect performance on grids. The main aim of the present experiment was to determine the effects of as many as possible of these variables. The subsidiary aim was to replicate, or otherwise, the findings of Frith & Lillie (1972) and Haynes & Phillips (1973) that the effective discriminator between thought-disorder schizophrenics and other groups on repertory grid tests is not intensity of relationship between constructs but pure (element or internal) consistency. Six grids involving all feasible combinations of rating familiar and unfamiliar people and objects on psychological and physical constructs were administered to 10 thought-disordered schizophrenics, 10 non-thought-disordered schizophrenics and 10 normals. As regards the main aim of the experiment it was found that there was no significant grids x diagnoses interaction on pure consistency, but that there was one on intensity, even with pure consistency partialled out. Taking previous work into consideration, it appears that this effect is weak, inconsistent from experiment to experiment, and anomalous in the present study. As regards the subsidiary aim, the findings of the two previous experiments cited above were confirmed. The discrepant results of McPherson et al. (1973) are considered, and it is noted that they held only for a subsample of the schizophrenic group in that experiment, the results for the total sample of schizophrenics, manics and depressives not being inconsistent with those of other experiments. The theoretical and practical implications of these and previous finding are indicated.