This study gives results for comparisons between dietary history interviews repeated at short-term (4-8 months) and long-term (4-7 years) intervals in conjunction with the Finnish Mobile Clinic Health Examination Survey. Interviews surveying the whole range of consumable foods over the preceding year were completed in 1967-1976. Short-term study was accomplished among 93 adults, and long-term study among 1844 adults. Comparisons were made for intakes of 32 food groups and 32 nutrient indices. In the short term, the intraclass correlation coefficients for nutrient indices ranged from 0.16 to 0.80, with 90% of values higher than 0.5. The corresponding figures for repeated measurements at long-term interval were generally poorer, being in the range 0.12-0.60, with 45% of values > 0.5. When studied in population subgroups, long-term agreement in dietary data was not found to be affected by sex, age, body mass index or smoking status, but it may be reduced among heavier drinkers (> or = 20 g alcohol per day). The intraclass correlation coefficients for separate nutrients tended to be higher than those for different food groups. In conclusion, we suggest that the reproducibility of the dietary history method used was acceptable, and that the dietary patterns of examinees were sufficiently stable to be compatible with the needs of epidemiological follow-up studies.