To date, few large web-based epidemiological studies have been performed in a population-based setting. Sweden has optimal prerequisites for web-based studies with more than 80% of the general population having access to the Internet. Our aim was to investigate (I) response rates in an epidemiological study using primarily the web as a tool for data collection and (II) whether socio-demographic patterns vary between responders to a web and a paper questionnaire. In 2003, we invited 47,859 women to complete a web questionnaire. Two reminders were sent to non-responders; in the first a random sample received a paper questionnaire and in the second the majority received a paper questionnaire. All other non-responders received web questionnaires. Differences in response rates between responders to web and paper questionnaires with regard to socio-demographic and other variables were analyzed, and estimates of the bias introduced by these differences were estimated. In total, 41% of the women responded to the web questionnaire and 31% to the paper questionnaire (overall response rate 72%). The web-, paper- and non-responders respectively did not differ significantly in age, physical activity levels, and body mass index. Women answering web or paper questionnaires had a higher level of education and income and a lower level of smoking than non-responders. The bias associated with collecting information using web questionnaires was not greater than that caused by paper questionnaires. We conclude that web-based questionnaires are a feasible tool for data collection in large population based epidemiological studies in Sweden.