Risk source plays an important role in risky decision making. Unlike “lottery risk” that arises from a stochastic process, “person-based risk” is originated from others’ decisions. Bohnet and Zeckhauser (2004) first studied the effect of risk source on risk preference using “trust game” task in which participants were asked to assign their “minimum acceptable probabilities (MAPs)” for securing the higher payoff, so as to make themselves just incline to the risky option from the sure one. It was found that the MAPs were generally higher when the occurrence of a certain outcome was determined by other person rather than a stochastic process. This phenomenon was named “betrayal aversion”. Nevertheless, Fetchenhauer and Dunning (2012) obtained quite different findings: participants were significantly more willing to choose the risky option in trust game than in lottery game when they were informed that their chance of drawing a white ball or interacting with a trustworthy person was about 50%. These inconsistent findings hint the necessity of undermining the causes of “betrayal aversion”. Hence the present study is to investigate how risk source and need for affiliation affect risky decision-making. In Study 1, participants face a binary-choice trust game with two options: a certain option in which both the participant and the anonymous partner would earn ￥10 for sure, and a risky one in which the participant may have a “good” outcome (with probability of 50%, participant and partner would earn ￥15 respectively) or a “bad” outcome (50% probability, participant earns￥8 and the partner earns ￥22). A between-subject design was used to explore the effects of risk sources. In “lottery risk” condition, whether the risky outcome would be (15, 15) or (8, 22) is depended on a lottery mechanism. In “person-based risk” condition, the outcome distribution depends on the partner’s choice. Participants in Experiment 1 were asked to calibrate their MAPs which could represent their risk preferences. In Experiment 2, participants were asked to choose between the certain option and the risky one directly. After that, their needs for affiliation were measured. The results of Experiment 1 showed that participants are more risk-averse when the outcome is determined by an anonymous partner rather than a lottery mechanism. Experiment 2 found that participants who have low-need for affiliation are betrayal averse, while people with high-need for affiliation prefer lottery risk and person-based risk the same. In Study 2, 2 (sources of risk: lottery risk / person-based risk) × 2 (emotion：no fear/ fear) between-subject experiments were designed to investigate further the effect of need for affiliation on decision-making by arousing the emotion of fear that was induced through watching a horror video in Experiment 3 or describing a frightening scenario in detail in Experiment 4. When emotion of fear aroused, individuals were found more risk-averse under lottery risk, but more risk-seeking under person-based risk. In other words, the betrayal aversion phenomenon even reversed when under fear. It was also found that the need for affiliation acts as a mediator between fear and risky decision making. The findings of “betrayal aversion” and “reversed betrayal aversion” suggested that risk sources are critical to risky decision making. Not only the probabilities but also the risk sources should be considered in evaluation of utility. Compared with the decision making under lottery risk, decision under person-based risk is more complicated. When trust is violated, people may experience betrayal costs. On the contrary, if the partner is trustworthy, people may reap extra psychological benefits such as satisfying the need for affiliation. This study deepened our understanding of trust and betrayal. To encourage trust, it is important to diminish the material costs of betrayal and to emphasize the warmth of interpersonal interaction as well.