This collaborative study was designed to assess the performance of commercial methods for protein S (PS) antigen measurement. Twenty-five different samples were distributed deep-frozen (24 plasmas) or lyophilized (one plasma) to five laboratories. They were analyzed blind in each laboratory by the method used locally and by three commercial methods which included two electroimmunoassays (EIA), Asseraplate-PS (Diagnostica Stago), Rellplate-S (American Diagnostica) and an ELISA system, Asserachrom-PS (Diagnostica Stago). 1. Reproducibility. Average between-laboratory coefficients of variation were 15.4%, 17.6% and 25.3% for Asserachrom-PS, Asseraplate-PS and Rellplate-S. 2. Specificity. Results of all methods showed that PS is underestimated when C4b binding protein is high. This influence was particularly evident for the ELISA Asserachrom-PS and disappeared when the antibody-antigen incubation period was prolonged to overnight. 3. Sensitivity. In all laboratories ELISA detected even the lowest PS concentration (4 U/dl), whereas the two EIAs were less sensitive (lower detection limit 14 U/dl). All methods and laboratories correctly diagnosed a plasma sample from a PS congenitally deficient patient. Conclusions. This study shows that better standardization of PS immunoassays is necessary to improve accuracy and reduce interlaboratory variability before a candidate plasma standard can be successfully calibrated in an international collaborative study.