The current research focuses on the effects of two fundamental dimensions, agency and communion, on realistic threat perception. Agency and communion play important roles in social perception and judgment. Previous researches found out that communion is priority over agency in individual perception. In group perception, studies also showed that communion trait has influence on the global evaluation of out-group rather than agency trait. However, there are inconsistent results showing that agency has main effect on out-group perception either. Furthermore, the perception process of individual and group are different. Group perception bases on memory and makes judgment depending on the information from memory just when it is needed. Realistic threat is a significant part of out-group perception, and people would perceive it when their out-group may occupy more resources or hurt their bodies. Realistic threat perception is related to communion trait. The Stereotype Content Model supposes that judging whether one’s intension is good is crucial to avoid harm. To some degree, out-group’s agency trait is also important to realistic threat. High agency trait can predict higher career success and social status, and high status groups are more likely to trigger the realistic threat of out-group. In light of these considerations, our main hypothesis is that communion and agency are both important in the perception of realistic threat. Research 1 recruited 355 participants ranged in different occupations and ages. Participants completed a survey in which they evaluated agency, communion and realistic threat of 19 natural groups differing in social hierarchy. Group and individual level analysis get the same result that communion and agency can both predict realistic threat (individual: β communion=--.250, SE communion=.032, t communion=-7.737, p<.001; β agency= .141, SE agency=.032, t agency=5.493, p< .001, group: b communion=-.455, t communion(16)= -3.283, p< .01; b agency=.285, t agency(16)=2.613, p< .05). Communion negatively predicted realistic threat while agency positively predicted it. Research 2 used Minimail-Group paradigm to avoid the impact of social background and a total number of 87 undergraduates and graduates took part in it. We manipulated the agency and communion information to explore the influence of out-group traits on realistic threat perception in a 2(agency: high\low)×2(communion: high\low) between-subject design. First, participants were assigned to two different groups randomly and completed a personality questionnaire. Then they were informed the results would be calculated and during this time they should finish a diet memory task. After showing each group the result of the other group (4 conditions), we let them ascribe out-group’s traits as manipulation check and measured their realistic threat perception. The results also showed that both agency and communion have significant main effect (Fagency (1，83) = 8.646，p< .01，η2=.094；Fcommunion (1，83) = 8.765，p< .01，η2=.096), when agency of out-group was high or its communion traits was low, in-group members perceived more realistic threat. Our present two researches demonstrate that the communion and agency traits of out-group both have important effects on realistic threat perception no matter in social groups or experimental groups condition. Previous researches mostly focused on the effect of communion trait on intergroup perception yet lost sight of agency trait. We emphasize the influence of both traits on realistic threat perception can make the research of two fundamental dimensions of group perception more integrity. However, how these two fundamental dimensions of social cognition affect different facets of group perception and evaluation still needs further exploration.