Data from an ongoing prospective population study of women in Göteborg, Sweden, were used to assess agreement between self-reported birth weight and birth weight obtained from original delivery records of women aged 44-60 years. Of the eligible population with traced delivery records (n = 693), only 28% (n = 192) could report their own birth weight. Spearman correlation between self-reported birth weight and birth weight from original records was r = 0.76. However, a difference plot, with limits of agreement at -1,028 to 1,038 g (95% confidence limits: lower limit, -1,157 to -901 g, upper limit, 910 to 1,166 g) revealed poor agreement between methods. Of the self-reported birth weights, 53% were in error by 250 g or more, and 31% were positively or negatively discordant by 500 g or more. Application in an analysis of cardiovascular risk factors in adulthood found conflicting results between self-reported and recorded birth weights. Low reporting rate, poor reporting accuracy, and misleading findings in application led to the conclusion that self-reported birth weights from middle-aged women would not be a satisfactory replacement for birth weights from original records.