The aims of the present study were (1) to compare three methods of assessment of dietary Se intake, i.e. chemical analysis of duplicate diets, diet records and a food-frequency questionnaire (FFQ) designed specifically for Se, and (2) to determine dietary Se intakes of residents of Otago, New Zealand. The FFQ was completed by 110 free-living adults. Diet records (3 d) and duplicate diet collections were carried out by forty-three of these subjects chosen on the basis of low blood Se concentration, and during a period when consumption of the high-Se foods fish, kidney, liver and Brazil nuts was discouraged. Mean Se intakes were similar for duplicate diet analysis (29 (SD 13) micrograms/d) and diet record assessments (28 (SD 15) micrograms/d). Estimates of intakes from the FFQ for the subgroup of forty-three subjects were higher (51 (SD 26) micrograms/d) than those from duplicate diets and diet records. Values from duplicate diet analysis and diet record assessments were strongly correlated (r 0.7, P = 0.0001), but difference plots indicated a lack of agreement between the two methods. Thus, diet record assessment was not adequate for predicting dietary Se intakes of individuals. Significant correlations were found for relationships between Se intake from duplicate diets (microgram/kg body weight per d) and plasma Se, Se intake from diet records (microgram/d and microgram/kg body weight per d) and plasma Se; and Se intake from the FFQ and whole-blood Se. Se intakes from duplicate diets and diet records were similar to those reported previously for New Zealanders, but lower than the recommended intakes in the USA (National Research Council, 1989), Australia (Truswell et al. 1990) and the UK (Department of Health, 1991) and the World Health Organization/Food and Agriculture Organization/International Atomic Energy Agency (1996) normative requirement.