The aim of the present cross-sectional study was to examine the agreement and disagreement between a 7 d diet diary (7DD) and a self-administered machine-readable food-frequency questionnaire (FFQ) asking about diet in the previous year, and to validate both methods with biomarkers of nutrient intake. The subjects were an age- and employment-grade-stratified random subsample of London-based civil servants (457 men and 403 women), aged 39-61 years, who completed both a 7DD and a FFQ at phase 3 follow-up (1991-1993) of the Whitehall II study. Mean daily intakes of dietary energy, total fat, saturated, monounsaturated and polyunsaturated fatty acids, linoleic acid, total carbohydrate excluding fibre, sugars, starch, dietary fibre, protein, vitamin C, vitamin E (as alpha-tocopherol equivalents), folate, carotenes (as total beta-carotene activity), Fe, Ca, Mg, K and alcohol were measured. Serum cholesteryl ester fatty acids (CEFA), plasma alpha-tocopherol and beta-carotene were also measured as biomarkers. Estimates of mean energy intake from the two methods were similar in men, and some 10 % higher according to the FFQ in women. Compared with the 7DD, the FFQ tended to overestimate plant-derived micronutrient intakes (carotenes from FFQ v. 7DD men 2713 (SD 1455) v. 2180 (SD 1188) microg/d, women 3100 (SD 1656) v. 2221 (SD 1180) microg/d, both differences P<0.0001) and to underestimate fat intake. Against plasma beta-carotene/cholesterol, carotene intake was as well estimated by the FFQ as the 7DD (Spearman rank correlations, men 0.32 v. 0.30, women 0.27 v. 0.22, all P< or =0.0001, energy-adjusted data). Ranking of participants by other nutrient intakes tended to be of the same order according to the two dietary methods, e.g. rank correlations for CEFA linoleic acid against FFQ and 7DD estimates respectively, men 0.38 v. 0.41, women 0.53 v. 0.62, all P< or =0.0001, energy-adjusted % fat). For alpha-tocopherol there were no correlations between plasma level and estimated intakes by either dietary method. Quartile agreement for energy-adjusted nutrient intakes between the two self-report methods was in the range 37-50 % for men and 32-44 % for women, and for alcohol, 57 % in both sexes. Disagreement (misclassification into extreme quartiles of intake) was in the range 0-6 % for both sexes. The dietary methods yielded similar prevalences (about 34 %) of low energy reporters. The two methods show satisfactory agreement, together with an expected level of systematic differences, in their estimates of nutrient intake. Against the available biomarkers, the machine-readable FFQ performed well in comparison with the manually coded 7DD in this study population. For both methods, regression-based adjustment of nutrient intake to mean dietary energy intake by gender appears on balance to be the optimal approach to data presentation and analysis, in view of the complex problem of low energy reporting.