Summary. The possibility of seasonal variation in the feedback effect of testosterone or oestradiol was investigated by giving replacement treatment to geldings for 2–3 weeks during breeding and non-breeding seasons. In the non-breeding season, testosterone suppressed LH values (mean ± s.e.m., ng/ml) in all geldings (before treatment, 7·5 ± 2·3; final treatment week, 1·8 ± 0·2; P <0·05), whereas early in the breeding season, testosterone caused a prolonged rise in LH (before, 6·8 ± 2·3; final week, 18·9 ± 6·4; P <0·05). In all testosterone experiments, LH returned to pretreatment levels within 2 weeks after treatment. Oestradiol treatment caused a prolonged increase (P <0·05) in LH concentrations (mean ± s.e.m., ng/ml) in both seasons (breeding: before 5·2 ± 1·1; final week, 16·2 ± 4·8; non-breeding: before, 10·9 ± 1·9; final week, 20·1 ± 5·2). We conclude that in geldings the feedback effect of testosterone varies with season and, further, that testosterone replacement may be able to restore to geldings the stallion's seasonal pattern of LH secretion. The results suggest that, in male horses, testosterone and possibly oestradiol, are important components in the neuroendocrine pathway controlling seasonal breeding and, moreover, are essential for the generation of a positive signal for LH secretion in the breeding season.