An estimate for the source of coagulase-negative staphylococci in blood cultures is presented and its use illustrated in two populations. The method estimates the probability that a positive blood culture is positive due to contamination and the probability of it being positive due to bacteraemia. The effect of changes in efficiency of the blood culture technique on these estimates is demonstrated. One of the illustrative populations consists of 537 sample pairs from 273 patients, taken after open-heart surgery in 1978–80. The other consists of 500 consecutive sample pairs taken in one of two hospitals in 1983–4.