The problem of characterising more specifically the cognitive requirements involved in subtests from standardised measures of intelligence represents a main problem in the research on exceptional populations. A new way of classifying tests of mental abilities is presented. Rather than focusing on the content of a given test, the present classification system focuses on their structures. The classification system is applied to the WISC-R (Wechsler Intelligence Scale for Children--Revised, Wechsler, 1974). It will be used to interpret the IQ-profiles of a clinically defined sample of reading impaired children (N = 82). The basic finding of the study is that the structural complexity of a subtest influences the tests results of reading impaired children. This influence is interpreted as a consequence of poor procedural knowledge; i.e., poor knowledge about how to organise complex sets of data. It is suggested that teaching of metacognitive strategies could be an aid for reading impaired children.