We have assessed the tumour markers placental alkaline phosphatase (PLAP), lactate dehydrogenase (LD), and human chorionic gonadotrophin (beta HCG) using 2,000 serum samples from 286 patients with seminoma. The ROC curves show that no one marker performs adequately for the detection of disease either at initial staging or during follow-up. We used a Markov model heuristically to devise strategies, in which marker results were assessed in combination, which might be useful in clinical practice. We found that the best strategy was to consider a test result abnormal only if either the beta HCG was greater than 6 Ul-1 or the LD was greater than 400 U l-1 and the PLAP level was greater than 60 U l-1. This will detect about 50% of patients with disease and the false-positive rate is 2%. In practical terms this means that PLAP need only be estimated in patients whose beta HCG is less than 6 IU l-1 and whose LD is greater than 400 U l-1.