Operator experience is considered to in¯uence the safety and success of medical procedures. We performed a retrospective survey to assess learning curves in chorionic villus sampling (CVS). Data of 2081 consecutive women, in whom CVS was carried out in a tertiary care university hospital for prenatal diagnosis, were available for analysis. Endpoints of the analysis were fetal loss, failure of the procedure and need for several needle insertions. Frequencies of each endpoint were calculated and plotted for consecutive series of 50 samplings per operator. Logistic regression analysis was used to quantify the effect of operator experience. We observed a statistically signi®cant learning effect for the unit as a whole as assessed by fetal loss ®gures and the same tendency for need to do several insertions. We did not observe a (collective) learning effect for failure of the procedure. Individual operators showed a signi®cant learning pro®le for some endpoints. The individual learning pro®le predominantly depended on previous experience in amniocentesis. Our study substantiates the presence of a learning curve for CVS. This supports the view that centralization and restriction of CVS to a limited number of experienced operators within centres, is likely to have a positive in¯uence on safety and success of the procedure.