possibility of bias but also the plausible direction in which it might operate.We were at great pains, even in the title of our paper, to show that our findings supported a familial aetiology and that this might be genetic. On the other hand, the segregation analysis makes the genetic aetiology more plausible; we made it quite clear that this was not proof. We also pointed out that voluntary childlessness might not have been exactly the same in both groups, and therefore we performed a sensitivity analysis to see what effects this bias might have had on our conclusions -a point that Gregson and colleagues overlook.The high incidence of subfertility would not seem to explain the striking similarity in the sperm counts across the three families in which extended sampling was permitted.Our paper shows that we were aware of all the biases identified by Gregson and colleagues, and more besides. We believe that our findings should be corroborated and that they are an intriguing clue. If Gregson and colleagues can suggest a method to overcome bias arising from what must inevitably be verbal evidence then we would love to discuss it with them. We would, for example, like to know whether they really think that observers could be blinded in a study of genetic histories and whether this has been accomplished successfully for other topics for which genetic histories are required from cases and controls.With regard to the letter from F J Stewart and A J Hill, absence of the vas deferens causes azoospermia, whereas we were concerned with oligozoospermia. There is no particular reason for thinking that oligozoospermia might be associated with mutations in the cystic fibrosis gene.

No abstract