Several studies have used factor analysis to identify food intake patterns in epidemiological studies as an alternative to nutrient-based analyses, but few have validated the factors in a larger population. Our present objectives were: to compare the factor scores based on a food-frequency questionnaire (FFQ) with scores based on a 7 d diet record; to examine the consistency of the factor score correlations across strata of age, BMI, energy intake, education, physical activity and smoking and to compare factors identified in two sub-populations. In 879 men and 927 women, of the total population sample of 3785, scores on food intake factors, three for men (‘green’, sweet’ and ‘traditional’) and two for women (‘green’ and ‘sweet–traditional’), identified in data from the FFQ and the diet record, were compared. The loadings of foods on the factors were very similar and the correlations between the corresponding factor scores, based on the two dietary assessment methods, were: for men ‘green’ 0·61, ‘sweet’ 0·55, ‘traditional’ 0·34; for women, ‘green’ 0·61, ‘sweet–traditional’ 0·57. Stratification did not significantly modify the correlations, with a few inconsistent exceptions. Factors obtained in a different subsample of the population, for which there was only data from the FFQ, were almost identical to the factors found in the subsample, who provided both FFQ and diet record information with regard to food loadings and model fit. In conclusion, the food intake factors identified were reproducible using two different dietary assessment methods and, furthermore, independent of stratification.