Editor—In 1998, Crow and Tolmie1drew attention to the problem of giving recurrence risks to families with undiagnosed mental retardation. Published empirical risks vary widely and are based on data which have been collected in many different ways and are decades old. Clinical practice in the UK, as derived from their postal survey,1 showed similarly wide variations. Thus, in a standardised situation where the index patient is a boy with either mild or severe mental retardation, published recurrence risks vary from 3.5 to 17.8% and the UK clinicians used a similar wide range of figures. We would like to make some suggestions for approaching this common problem based on new empirical data and to offer some theoretical guidelines for estimating risks for the sibs of both male and female index cases.

We have just completed a clinical genetics review of 429 subjects with mental retardation in the Australian Child and Adolescent Developmental (ACAD) project, which is a longitudinal study of behaviour in the mentally retarded. A comprehensive, representative, community based sample of children was recruited into the study in 1990/91 for a first behavioural assessment.2 3 The genetics review was at the time of their second assessment some five years later when the subjects ranged in age from 10 to 24 years. In about two thirds the IQ had been measured as less than 50, 8% had not been tested, and the rest were over 50. All were seen by one or more of three clinical geneticists except for 74 patients in whom the diagnosis was clear cut (for example, documented trisomy 21). We found that when all those patients in whom the cause of their mental retardation was known and all those with recognised clinical entities without a known cause (for example, cerebral palsy, …