In a population study of injuries in two samples from the general population of Göteborg, Sweden, of altogether about 1200 50- and 60-year-old men, the non-participation rates were 25% and 19% respectively. The aim of this report was to estimate the size of the bias caused by non-participation. Besides comparing official register data, e.g., marital status, dwelling conditions and mortality between non-participants and participants, it was also possible to measure certain morbidity variables by using the register of an emergency department over a 7-year period. The non-participants had less stable dwelling conditions and were more often unmarried, divorced or widowed than the participants. They also had more accidents per person and more head injuries, were more often transported to the emergency department by ambulance, tended to have more serious injuries and had a higher mortality rate during follow-up. The non-participants were about six times more often inebriated at the attendance and the differences seemed to be associated mainly with the alcohol factor and less with the participation factor. In conclusion, the bias in incidence estimates caused by non-participation appears to be small to moderate in this type of study as long as the non-participation rate can be kept on the same level as in this study or lower.