BACKGROUND: previous studies of the accuracy of self-reported weight have been criticised for using inadequate methods and have included only young or middle aged adults. Self-report is more likely to be relied upon in both clinical and research practice in older age groups. The cultural pressures that may cause the tendency among younger women to underestimate their weight, particularly when they are overweight, may operate differently in older women. OBJECTIVE: to determine the accuracy of self-reported weight among older women. METHODS: national cross sectional survey of women aged 60-79 from 9 towns across England, Scotland and Wales. Self-reported weight from a participant questionnaire was compared to measured weight at examination. RESULTS: of 2729 women who were invited, 1636 (60%) returned the questionnaire (of whom 1549 gave a self-reported weight) and 1384 (51%) attended for examination (of whom 1381 were weighed). In total there were 1310 (48% of the total invited sample) with complete self-report and measured weight. Self-reported and measured weight were highly correlated (Pearson's correlation coefficient, 0.982 95% confidence interval, 0.979-0.983) and self-reported weight differed from measured weight by only 1.0 kg (95% confidence interval 0.8 kg, 1.1 kg) on average. However, a difference plot, with limits of agreement at -4.0 kg to +6.0 kg (95% confidence intervals: lower limit -4.3 kg, -3.8 kg; upper limit +5.7 kg, +6.2 kg) revealed poor agreement between methods. Obese individuals, in particular, were more likely to underestimate their weight. CONCLUSIONS: though self-report of weight by women in their 60s and 70s is highly correlated, at an individual level differences between self-report and measured weight are frequently large. Obese individuals, in particular, tend to underestimate their weight. Self-report of weight should not be relied upon in prospective epidemiological studies or clinical practice when accuracy at the level of the individual is required.