Males achieve markedly higher scores than females on mental rotation tests (MRTs). Therefore, it might be hypothesized that, within groups of males, testosterone levels modulate MRT performance. However, studies of this relationship have yielded inconsistent results. Notably, a recent study of 28 American men, using the computerized Shepard and Metzler MRT (SM), found significant associations between salivary testosterone levels and the intercepts of the functions relating response time and error rate to the angular disparity between comparison objects. Conversely, a study of 35 British men, using the same methodology, found no such associations. We attempted a cross-cultural replication of these studies, in which we obtained salivary testosterone levels, together with performance measures on the SM, from 92 heterosexual right-handed men, aged 21-38, in Beijing, China. We hypothesized that Chinese men might perform more slowly and carefully than Western men on this test (which imposes no time limitations), but that associations of testosterone levels with performance, if real, should nevertheless be detectable across cultures. We found that the Chinese men indeed displayed significantly longer response times than the American men, although the Chinese men were equally accurate. Interestingly, testosterone was significantly associated with the slope of the response time function in Chinese men, whereas the earlier American study had found that testosterone was associated with the intercept, but not the slope, of this function. These observations suggest that differing cultural values regarding speed and accuracy may influence MRT performance--and that these values must be considered in future studies of testosterone and MRT measures.