Frequencies of Severe Mental Retardation (SMR) and Mild Mental Retardation (MMR) were obtained from pilot surveys of severe childhood disability in 8 less developed countries. Approximately 1,000 children aged 3 to 9 years were surveyed in each location. The surveys used a Ten Question (TQ) door-to-door interview, usually of the mother, as a screening procedure and a systematic medical and psychological assessment for the diagnosis. Diagnostic categories of SMR (IQ less than or equal to 55) and MMR (IQ greater than 55, less than or equal to 70) were assigned by well trained local psychologists, using formal and informal techniques of assessment. Contrasting frequencies and distributions for MMR compared with SMR are shown for each location. No consistent pattern for MMR versus SMR emerged, neither regarding frequency, male/female ratio, average age nor socio-economic status of household. By contrast, MMR did differ from SMR consistently regarding consanguinity of parents, the presence of associated impairments and the positive report of symptoms at interview. Also, the families of all MR children were lower in SES than comparison families. An interpretation of these findings is offered: the more severely disabled children tend to be assessed as SMR, but so do other children who might, in better circumstances, be assessed as MMR. The relevance of this interpretation is discussed, in terms of assessment and of rehabilitation, and as a guide to further epidemiologic studies.