An assumption in case-control studies is that forces of selection are the same for cases and controls. This may not be true for studies of male infertility among infertility clients. Earlier reproductive outcomes may introduce modification of risk behavior or differential referral. Selection bias might also occur when infertile males are compared with fertile males. Partners of sterile men are more likely to have "normal" fertility, while partners of a reference group of normozoospermic men tend to have a lower fertility potential. The latter may lead to overrepresentation of causes of reduced female fertility and introduce bias into estimates of risk factors shared by couples. The relation between cigarette smoking and semen quality was studied in a population of infertility clients from the Netherlands during 1995-1996. To reduce the potential for bias, this relation was studied first in a restricted population less aware of the type of infertility involved. The odds ratio of infertility with smoking was elevated in the restricted population as compared with the total population. Adjustment for smoking by the female partner increased the odds ratios for male smoking as well. These results indicate that bias may occur in clinic-based fertility studies because of different forms of selection.