Animal studies show clear evidence for a causal link between testosterone and aggression. This review assesses studies involving androgens, principally testosterone, and human aggression. Evidence for a possible effect of prenatal androgens is inconclusive. In adults, higher testosterone levels are found in groups selected for high levels of aggressiveness. Correlations between testosterone and aggression were low when hostility inventories were used, but higher (r = .38) when aggressiveness was rated by others. Regression analysis data and studies of boys at puberty were inconclusive. Other studies show that the outcome of aggressive and competitive encounters can alter testosterone levels, thus confounding interpretation of the correlational evidence. The design of future studies to reveal evidence of a causal link is considered. Suggestions concerning two important methodological problems, the experimental manipulation of hormone levels and the nature of the dependent variable, are made.Studies of a wide variety of vertebrate species have indicated that androgens, principally testosterone, facilitate aggressiveness between males, thus providing a mechanism for the increase in male competitive aggression at a time when the animals are reproductively active (Archer, 1988). This ethological evidence raises the question of whether testosterone also facilitates aggression in human males. Some reviews of the human evidence have concluded that it does (Donovan, 1985; Rubin, Reinisch.8~ Haskett, 1981) whereas others have urged caution in drawing this conclusion (Benton, 1983;Carlson, 1986;Rada, Kellner & Winslow, 1976).The purpose of the present paper is to provide a systematic review of current evidence on this issue, to include a limited meta-analysis where possible, and to highlight methodological issues raised by existing studies. The review begins with a brief survey of animal studies to provide a background for the human evidence. The possible involvement of neonatal androgens is then considered briefly, before examining the following types of evidence from adults : comparisons of testosterone levels in selected groups which show high and low levels of aggressiveness; correlations in both selected samples and unselected samples ; and regression analyses. Studies carried out on boys at puberty -when testosterone levels rise dramatically -are then considered. The limitations of correlational evidence are discussed, particularly in relation to the influence of the social environment on testosterone levels. Existing attempts to provide causal analyses are described, before considering how best to design future studies to reveal more clearly whether testosterone facilitates human aggression.