Sources of error in diet assessment studies include inaccurate estimation of portion sizes, incorrect description of food items, incomplete recall of food items eaten and inclusion of extraneous foods. Data from free-living subjects on self-selected diets usually cannot separate the sources of error because of uncertainties about actual food intake. The present study was conducted to identify the contribution of portion size estimations of different food groups to the estimation of food group nutrient intake. Because subjects in the Dietary Effects on Lipoproteins and Thrombogenic Activity (DELTA) Study were fed well-controlled experimental diets, this was an excellent opportunity to compare self-reported portion sizes with weighed portions provided by the research kitchen.

Subjects were normal weight males (n=25) and females (n=18) from three DELTA field centers (Columbia University, Penn State University and University of Minnesota) Twenty-four hour recalls on three randomly selected days were collected for each subject at the beginning of the feeding period. The recalls were conducted by telephone interview at the Penn State Diet Interview Center using the University of Minnesota's microcomputer Nutrient Data Systems. Research kitchen staff observed subjects consume portions. Portion sizes recalled by subjects were converted to gram weights. Individual foods in the experimental menus were grouped as major food groups and subgroups. Self-reported and actual portion sizes for all food groups were compared.

Self-reported estimates of portion size versus actual quantities provided by the research kitchen ranged from –43% (underestimation in gram quantities) for condiments to +156% (overestimation) for pasta. In only 5 subgroups (i.e. juices, bread, cookies, milk/milk products and condiments) were portion sizes underestimated while intake from 14 subgroups was overestimated. The overestimation of food groups that resulted in the greatest relative overestimation of protein were pasta, nuts and mixed dishes. The overestimation of food groups that resulted in the greatest relative overestimation of fat were meats, baked beans, nuts, gravy and fat spreads. The overestimation of fruits, pasta, nee, mixed dishes and puddings resulted in the greatest relative overestimation of carbohydrate.

This study establishes that subjects overestimate portion sizes of foods they were given to eat and recalled having eaten. In addition, portion sizes of certain food groups are more likely to be overestimated than others. This error in portion size estimation is one important contributor to the errors associated with dietary assessment studies. However, the extent of under- or overestimation of total nutrient intake is also affected by other sources of error.