The sexual behavior of 42 stallions from French national and private studs was examined in two contexts: semen collection for artificial insemination (AI) and in-hand natural service (NS). Each stallion was observed twice in the same context. Erection and ejaculation latencies, the number of mounts leading to ejaculation, dismount latency and total breeding time were measured and compared between AI and NS. Mount without erection was rare (6/83 observations). Erection latency was 89+/-11s, and was not different between NS (62+/-22s) and AI (100+/-13s, P=0.128). Stallions ejaculated after either one mount (62/83 observations), or two (11/83 observations) or three mounts (10/83 observations). Ejaculation latency was 85+/-15s (84+/-19 in AI and 86+/-28 in NS). If 1st mount did not lead to ejaculation, then ejaculation latency increased several fold following the 2nd mount during both AI and NS. The results provide reference measures for semen collection in French studs. Difference in erection latency between AI and NS, although not statistically significant, may reflect different contributions of excitatory inputs from the brain and the genital area to the activation of spinal networks controlling erection. In contrast, lack of difference in ejaculation latency between AI and NS suggests that the spinal network that controls ejaculation follows a more rigid motor pattern.