Female twin pairs were identified from birth records, and their families invited to participate in a prospective study of the determinants of alcohol problems in women. We investigated sampling biases arising because of failure to locate families, or non-cooperation of families. Out of 2644 families with a live-born pair (born between July 1975 and December 1986) who survived beyond infancy, contact was established and a brief screening interview completed with 90% (N = 2380). Fewer than 6% of located families declined to participate in the initial screening interview. Predictors of failure to locate a family or to obtain a screening interview were identified from information recorded in birth records, and from neighborhood characteristics identified from 1990 US Census block group data for the family residence when the twins were born. African-American families were under-represented in the final sample, but this effect was barely significant when other variables were controlled for. Under-represented were families where the mother was 19 or younger at the birth of the twins, where the mother herself was born out-of-state, or where information about biological father was not reported in the birth record. Non-participating families on average came from neighborhoods with a higher proportion of residents living in poverty, and with a higher proportion of African-American residents. Sampling biases were however small. The unusual cooperativeness in research of families with twins persists.