SUMMARY Rete testis fluid (RTF) was collected from six conscious Clun Forest rams by cannulation of the efferent ducts, and from anaesthetized rats by cannulation of the rete testis 16–24 h after efferent duct ligation. Testosterone concentration in this fluid was measured by radioimmunoassay and compared with levels in blood collected from the subcapsular testicular artery and veins and spermatic and jugular veins. The testosterone concentration in ram RTF fluctuated from nondetectable (< 100 pg/ml) levels to approximately 20 ng/ml over 1–25 h collection periods. The mean concentration was 8·5 ± 1·2 (s.e.m.). This value was approximately one fifth the mean concentration of testosterone in samples from the testicular veins at the dorsal pole of the testis. Testosterone concentration (ng/ml) in rat RTF collected under sodium pentobarbitone anaesthesia was 26·4 ± 2·7 (18 measurements, 18 rats). A lower value was obtained when ether was the anaesthetic used during efferent duct ligation (ether, 20·3; halothane, 29·9; sodium pentobarbitone, 29·1). The mean concentration of testosterone in the RTF was similar to that in testicular venous blood (29·8 ± 3·9; 27 measurements, 27 rats), although the highest values were obtained after ether anaesthesia (ether, 38·3; halothane, 29·0; sodium pentobarbitone, 21·7). Testosterone concentration in the testicular arterial inflow was similar to that in jugular venous blood in the ram but exceeded carotid arterial levels in the rat. The values reported are used to discuss: the origin and significance of androgens in rete testis fluid, species differences, the effect of anaesthetics and season, and vascular exchange of testosterone in the spermatic cord.