Studies based on data from the US have reported that the birthweight distribution at gestational age 28-31 weeks is bimodal with a second peak occurring at approximately 3300 g, suggesting that there is misclassification of term infants. In these studies, gestational ages were estimated from the date of the last menstrual period (LMP), and it has been suggested that ultrasound-based estimates of gestational age would eliminate the problem with bimodal birthweight distributions. Swedish data include both measures, thus offering an opportunity for comparison. All singleton births in Sweden from 1993 to 2002 with information on birthweight were included in the study (n = 917 901). Both LMP- and ultrasound-based estimates of gestational age were available for 75.1% of the births. Two possible sources of misclassification were considered: measurement error, assuming that ultrasound-based estimates are better, and data entry errors. An algorithm for assessment of data entry errors was developed; 67.4% of the births were left for the analyses of data 'cleaned' from data entry errors. Based on the entire study population, the LMP-based birthweight curves for lower-gestational-age preterm births were bimodal, with a second peak around 3500 g. The bimodal distribution was greatly attenuated when using ultrasound-based gestational age categories, but did not disappear. After cleaning the data, the LMP-based birthweight distributions for infants at gestational ages <32 weeks were no longer bimodal, and were very similar to the ultrasound-based curves. In conclusion, data entry errors are more likely to cause the bimodality in the birthweight distribution among preterm infants than measurement errors in the LMP-based gestational age estimate.