Objective: To assess the ability of a food-frequency questionnaire (FFQ) to rank Australians according to their intake of total carbohydrate, sugar, starch, fibre, glycaemic index (GI) and glycaemic load (GL). Design: Cross-sectional sample from a population cohort. Setting: Two postcode areas west of Sydney, Australia. Subjects: From 1992 to 1994, a total of 2868 older Australians provided dietary data using a 145-item Willett-derived FFQ. A representative sub-sample of 78 subjects completed three 4-day weighed food records (WFRs). Pearson and Spearman correlations, Bland–Altman plots and weighted kappa values were calculated. Results: Compared with the WFR, the FFQ provided higher mean estimates of all nutrients except starch and GI. All Pearson and/or Spearman correlations were greater than 0.5, except for GL. For GI, sugar, starch and fibre, the regression lines from the Bland–Altman analysis indicated a non-significant linear trend (P50.07, P50.36, P50.28 and P50.10, respectively). For GL and total carbohydrate, however, there was a significant linear trend (P50.006 and P,0.0001, respectively), indicating that as the GL and carbohydrate intake of individuals increased, so did the magnitude of the error between the FFQ and WFR. Weighted kappa values all indicated moderate to good agreement, with the exception of GL which was only fair. The proportions of subjects correctly classified within one quintile for all of the nutrients were over 50% and gross misclassification was low (,10%). Conclusion: This FFQ was able to rank individuals according to their intakes of total carbohydrate, sugar, starch, fibre and GI, but not as well for GL.