A specific leptin RIA was developed to assess concentrations of leptin in ovine plasma, and was shown to be efficient with bovine and caprine plasma. A specific, high-affinity antibody was generated against recombinant ovine leptin which, when used in a competitive leptin RIA, provided valid estimates of linearity (r=+0·989-0·998), recovery (102%), repeatability (13%) and limit of sensitivity (0·83 ng/ml for 100 µl sample size). Serial dilutions of five ovine, bovine or caprine plasma samples showed good linearity and parallelism with the recombinant ovine leptin standard curve. A comparison of this RIA was made with a commercial 'multi-species' RIA kit using 56 ovine plasma samples. Major differences were found in assay sensitivity. Non-lactating, non-pregnant, ovariectomized ewes were fed a ration for 65 days which provided 90 9% (control; n=12) or 39 2% of maintenance energy requirements (underfed; n=16) in order to analyse the respective effects of body fatness (estimated by either an in vivo dilution technique or body condition scoring) and of nutritional status on plasma leptin concentration. There was a significant positive correlation between body fatness or body condition score and plasma leptin levels (r=+0·68, P<0·001 or r=+0·72, P<0·001 respectively). When concentrations of leptin were assessed over time, underfed ewes exhibited a dramatic reduction in plasma leptin values ( 56%, P<0·001). These data provide strong evidence that, in sheep, the variations in plasma concentrations of leptin are related to variations in body fatness (35%) and, to a lesser extent, in nutritional status (17%).