BACKGROUND: Reports of a secular decrease in semen quality remain controversial, particularly due to the possibility of selection bias. We aimed to describe the potential bias due to self-selection of volunteers in semen studies involving fecund men. METHODS: Using data from the French multicentre study REPRHOM, we compared the characteristics of the partners of pregnant women for three levels of participation: completion of a refusal questionnaire (n = 698), agreement to complete the study questionnaires only (n = 676) and agreement to complete the study questionnaires and give a semen sample (n = 331, 13% of the subjects approached). RESULTS: Poorly educated men refused more often to participate than highly educated men. Semen providers were more likely to have experienced unfavourable pregnancy outcomes (odds ratio 1.68, 95% confidence interval 1.14-2.49) compared with participants completing the questionnaires only. Time to pregnancy was similar for all participants. CONCLUSIONS: This study demonstrates the existence of selection bias in semen studies associated with fertility and socio-demographic characteristics of men. The results of semen analysis for this population sample cannot be extrapolated to the whole population from which the volunteers originate. More information is required on who participates, and participation rates should be reported in semen studies to make it possible to interpret the results correctly.