Potency determination of porcine, bovine and human insulins relative to the International Standard in the pharmacopoeial rabbit bioassay system requires that the log-dose response curves are parallel. Furthermore, the same relative potency should be obtained independent of how the hypoglycaemic response is defined. The results of 508 rabbit blood glucose assays have been analyzed by new multivariate statistical methods. No deviations from parallelism of the log-dose response curves were detected. However, the potencies showed significant variation depending on the blood sampling times. Pure porcine and human (semisynthetic and biosynthetic) insulin potencies decreased by 12% and 18%, respectively, from the 30-min to the 2.5-h response, whereas bovine insulin potencies increased by 9%. Since the standard is a 52:48 mixture of bovine and porcine insulins, these results could be due to porcine and human insulins having a quicker onset and shorter duration of hypoglycaemic effect than bovine insulin. This was confirmed in assays of bovine relative to porcine insulin and by direct comparison of mean blood glucose curves. It is concluded that there is a response time-dependent variation in potency when the test and standard insulin have a different species composition. Hence, pure species insulin standards - a porcine, a bovine and a human standard - are needed for assay of the three insulins.