Two thromboplastin reagents ("Thrombotest" and "Normotest Automated") were used in evaluation of an automated method for determination of prothrombin time based on turbidimetric measurement of clot formation in a centrifugal analyzer. We used 60 plasma samples from patients with various diseases or being treated with oral anticoagulant and 16 normal plasma samples. Prothrombin times were calculated by a computer connected to the analyzer, a reading being made at either a certain per cent increase in total absorbance or a fixed absorbance increase. Both correlated well with the manual method (r = 0.98-0.99). The reading points best fitting the manually obtained data were estimated by minimizing the residual sum of squares in regression analyses performed at various absorbance increases. The per cent reading was better in this respect. Normotest Automated could be nearly perfectly related to the manual method, whereas Thrombotest showed a (negligibly) small deviation. Reproducibility was good within run (CV less than or equal to 3.2%) as well as between batch of the reagents, as assessed from variation in INR (CV less than or equal to 4.9%). We conclude that turbidimetry of clot formation may be validly used in automation of the prothrombin-time test. The equipment needed and the total time per analysis are about as for chromogenic substrate methods, but reagent cost is considerably lower.