The Self-Consistency Model (SCM) was initially proposed for the confidence in two-alternative forced-choice general knowledge questions. Since then, several studies have suggested that the model may also apply to other fields, such as belief and attitude, but more researches were needed for the verification and the complement. The aim of this study was to test the SCM of subjective confidence in Chinese cultural context as it applied to social belief. The experiment was conducted individually on the computer. Forty-four paid undergraduate or postgraduate students (22 females and 22 males from 18 to 24 years old) participated in this study. The task, which repeated six times, was divided into two parts. Each part included three blocks in which the entire set of 60 propositions (Social Axioms Survey) depicting a social belief were presented. Participants pressed the key to decide whether the proposition was “True” or “False”, and indicated their confidence in their choices with a scale of 0 to 10. Their response latency was measured by computer. For each participant, the choices were classified as “frequent” when they were made more than 3 times, and as “rare” if they were made less than 3 times across the 6 blocks ; “item consensus” means the proportion of participates choose the majority, consensual response on a particular occasion. We found that, (1) The confidence was inversely related to the response latency (r=-.32, p<.001). It could reflect the likelihood that the same choice would be made in the same item (r=.94, p<.05). So could the choice latency (r=-1.0, p<.001). Moreover, males was lower than females in confidence (t=-5.48, p<.001), and higher than females in response latency (t=4.62, p<.001). (2) The confidence was higher for the frequent responses (6.22) than for the rare responses (5.19), t=8.67, p<.001. In addition, the confidence in the frequent responses tended to increase with item consistency, while the confidence in the rare responses decreased with item consistency. The confidence was higher for the consensual responses (7.35) than for the nonconsensual responses (6.21), t=11.58, p<.001. Besides, the confidence in the consensual responses tended to increase with item consensus whereas the nonconsensual responses tended to decrease with item consensus. A similar pattern was observed for the response latency. (3) The responses that were consistently chosen by the same person (high consistency) were also more likely to be chosen by others (high consensus), r=.31, p<.001. The result showed that a commonly shared population of representation associated with each proposition. Comparing with the cross-person consensus, the within-person consistency was a better diagnostic of the self-consistency. As expected, the within-person consistency and the cross-person consensus analysis showed that the decision to accept or refuse a social belief was based on the sampling of a pool of representations associated with the belief. The confidence depended on the consistency with which the belief was supported across the sampled representations, and reflected the likelihood that a new sample would yield the same decision. The choices that based on the representative samples were associated with relatively higher confidence and took relatively shorter time to form. The choices with high consistency were associated with the high consensus. The results verified the existence of the Self-Consistency Model in Chinese cultural context.