Objectives. To assess the validity of a 213-item semi-quantified food frequency questionnaire (FFQ) in estimating habitual energy and protein intake in a sample of older South Africans. Repeatability of the FFQ was assessed by comparison of reported intakes after a 6-month period. Design. Crass-sectional analytic study. Methods. Twenty-one subjects were selected from a baseline sample of 200 non-institutionalised subjects aged 65 years and over in Cape Town, who had previously been randomly selected for a nutrition and health survey using a two-stage cluster design. Reported dietaryenergy and protein intakes, estimated by means of the FFQ method, were compared with 24-hour energy expenditure, measured by the heart-rate monitoring technique and 24-hour urinary nitrogen excretion, respectively. Results. Spearman correlation coefficients for reported energy intake (using the FFQ) versus measured energy expenditure were 0.31 (P = 0.482) and 0.36 (P = 0.345) for men and women, respectively. Men tended to underreport energy intake, while women tended systematically to overestimate energy intake by 21% and 25%, respectively. In men, reported protein intake using the FFQ closely matched urinary protein excretion and a good association between the two measures was found (r = 0.62, P = 0.061). In women, no association was found between reported protein intake and urinary nitrogen excretion. The FFQ resulted in a twofold overestimate of protein intake, based on urinary nitrogen excretion. In women, correlations between 6-month repeated measures of energy and protein intake using the FFQ were 0.69 (P = 0.067) and 0.61 (P = 0.053), respectively; however a poor between-measure association was found in men. Conclusions. The study findings demonstrate that the semi-quantified FFQ method underestimated food energy intake in older men and overestimated both energy and protein intake in older women.