The objective of this study was to compare the values obtained for total body fat obtained with deuterium dilution, anthropometry and bioimpedance in 41 institutionalized elderly individuals (65-90 years old). The values obtained with each technique were compared using the graphic analysis proposed by Bland and Altman, that plots the difference between measurements with both methods against their average. In men (n = 20) and women (n = 21), the best degree of agreement was obtained between the values measured by deuterium dilution and those calculated from skinfolds (mean difference = 1.4% and 6.9%, respectively). The limits of agreement (+/-2SD), for skinfolds reached a maximum of 14.8% in men, and 16.8% in women. These values tend to underestimate fat in the obese and overestimate it in thinner subjects. For bioimpedance and deuterium dilution, the inter-method difference is significantly greater: 9.3% in men and 14.7% in women. This lack of agreement is attributed to the fact that the bioimpedance equipment utilizes equations validated for younger adults. In conclusion, estimation of body composition using skinfoids has the smallest difference compared to deuterium dilution, even though individual measurements are not clinically acceptable. Caution is recommended when using measurements of body composition in the elderly, due to large errors in the determinations.