The amount of testosterone required for quantitative maintenance of spermatogenesis has been re-evaluated using techniques aimed at minimizing the synthesis of testosterone after removal of the testis. Adult male rats were treated with ethane dimethane-sulphonate (EDS) to destroy the Leydig cells, and were supplemented with 25, 5 or 1 mg testosterone esters by injection every 3 days for 21 days. Serum hormone levels, testicular morphology and spermatogenesis were assessed and the intratesticular levels of testosterone compared in testes either removed under ether anaesthesia and placed in liquid nitrogen (right testis) or removed after collection of blood and placed in ice (left testis). Data for testosterone-supplemented rats were compared with those for control rats and rats treated with EDS alone. All doses of testosterone suppressed LH and FSH levels in serum to within the hypophysectomized range, and Leydig cell regeneration in EDS-treated rats was prevented completely. Treatment of EDS-injected rats with 25 or 5 mg testosterone maintained testicular weight, the number of germ cells and the diameter of seminiferous tubules at stage VII within or above the control range, although there was a significant increase in the number of degenerating pachytene spermatocytes at stage VII with 5 but not 25 mg testosterone; none of these parameters was maintained at control levels by a dose of 1 mg testosterone. Levels of testosterone in testosterone-supplemented rats differed little between testes collected in ice and liquid nitrogen, but in controls and rats treated with EDS alone, testosterone levels were overestimated by 75 and 27% respectively when comparing testes collected in ice with those collected in liquid nitrogen. This suggests that synthesis of testosterone does occur after removal and cooling of the testes. Using the data for testes collected in liquid nitrogen, the present results suggest that intratesticular levels of testosterone need to be maintained at 24–46% of control values (i.e. higher than previously shown) for quantitative maintenance of spermatogenesis, although it is argued that even these values probably still represent underestimates. J. Endocr. (1988) 117, 19–26