duplicate portions of foods consumed; food-consumption diaries, in which daily fish intakes are recorded quantitatively; recall methods, such as 24-hr recall of fish consumption; diet histories of usual consumption at various meals; and food-frequency measures of usual frequency of consumption of fish and shellfish. Duplicate-diet collections and food-consumption diaries are prospective approaches, and the others are retrospective approaches.
General considerations for duplicate-diet studies were recently discussed by Berry (1997) and Thomas et al. (1997). In duplicate-diet studies, participants collect an identical portion of the food they consume and provide it to the investigator for laboratory analyses. In theory, duplicate-diet studies have the potential to provide the most accurate information on the ingested dose of MeHg, because the mass of fish and other nutrients and contaminants, in addition to MeHg, can be measured directly. The fact that only the fish portion of any given meal will contain MeHg simplifies the burden of duplicate-diet collection. In practice, however, this approach is limited by the demands it makes on the participants, the difficulty in identifying individuals who are willing to carry out such a study, the influence exerted by investigator observation, and the potential change in diet resulting in response to the burden of food collection. Thomas et al. (1997), working with nine highly motivated households, was able to collect duplicate samples for 97% of meals and 94% of snacks over a 7-day period. The number of uncollected meals, however, tripled after the first 3 days, and participants strongly recommended that future studies be limited to a maximum of 3-4 days. When such studies are confined to fish consumption, 3-4 days of collection might be useful only for populations with very frequent and highly regular patterns of fish consumption. Because of the practical limits on the length of the collection period, the authors recommended that duplicate-diet studies for risk-assessment purposes should be done over multiple intervals of time. Moreover, when the calorie content of collected food was compared with the estimated energy requirements of participants, duplicate portions were found to be underestimated.
Duplicate-diet studies have been specifically applied to the estimation of MeHg exposure by Sherlock et al. (1982) and Haxton et al. (1979). Sherlock et al. (1982) carried out a 1-week duplicate-diet study with 98 participants selected on the basis of frequent fish consumption. In