Abstract
In 1972, psychophysiological data were collected on 1800 3‐yr‐old children with the aim of selecting a subsample of whom some would have characteristics suggesting that they might later break down with schizophrenia and hence be at high‐risk. This approach contrasts with that usually adopted in high‐risk studies where a familial risk definition is employed. The paper examines the extent to which the necessarily crude initial selection procedure appears to have worked in terms of later more extensive analysis and the reliability of the procedures used over an 18 month period. Analogues between the technique used and those employed in psychometric analyses are suggested.