Sampling problems related to twin problems may be of at least three kinds: small samples, self-selection and unrepresentative ascertainment. The article mostly discusses the third type of sample problem. Data from a nationwide Norwegian twin study show that the results of twin studies will be quite different depending upon the ascertainment procedure. According to both the ICD-9 as well as the DSM-III classification system, only samples ascertained from mental hospitals treating severe cases are able to demonstrate hereditary factors of any strength.