In daily life, it’ s common for people to engage in Bayesian reasoning which means updating already existed beliefs according to new evidence. Bayesian reasoning has been widely applied into various fields such as psychology, medical treatment, and business. However, despite the wide usage of Bayesian reasoning, individual performance was surprisingly poor, as manifested by previous laboratory studies. In order to improve the Bayesian inference capacity, many researchers have investigated solutions from various aspects (e.g., data format, graphics assistance, and explicit nested set structure). One of the potential factors is causal information. Krynski and Tenenbaum proposed the causal Bayesian framework according to which reasoning consisted of three stages: model construction, parameter assignment and mathematical calculation. As to the “base-rate neglect”, a common bias observed in Bayesian reasoning tasks, they further argued that, the false positive rate in conventional study materials couldn't be completely understood, resulting in incorrect model construction; in contrast, problems with causal information might promote constructing proper model. However, subsequent studies couldn't reach an agreement on the validity of this assertion.This controversial result questions the ability of causal information to help people inference rationally. According to the "triple processing mind" model, the activation of introspective mind could suppress the intuitive beliefs (e.g., the transpositional representation bias in Bayesian reasoning), and then help subjects think more rationally. Hence, if instructed to reason introspectively (through instruction), individuals might be more likely to solve Bayesian problems. In consideration of the unstable effect of causal information and the triple processing mind model, causal information might only help people construct correct model, but only intuitively, while instruction could essentially facilitate rational process (including the correct modeling as a byproduct). Hence, instruction could be one of the potential factors; and this study assumes that, when participants receive both causal information and instruction, compared with other situations (with the absence of causal information and/or instruction), their performance would reach the highest. To test this hypotheses, 138 undergraduates were recruited to took part in the 2 (with or without causal information) × 2 (with or without instruction) between-subjects experiment. The dependent variable was Bayesian reasoning performance, with 2 criteria: the correction ratio and the accuracy. The results show that, under the accuracy criterion, the main effects of causal information (F(1,134)=6.06, p<.05, η2=.04) and instruction (F(1,134)=3.68, p<.05, η2=.03) are both significant, the interaction is also existed (F(1,134)=4.13, p<.05, η2=.03). Based on these results, we come to 3 conclusions: (1) causal information does promote the ability of Bayesian reasoning, but only without the presence of instruction ; (2) instruction has more robust facilitating effect, even if there is no causal information; (3) one the basis of instruction, causal information cannot bring more improvement . It can be seen that providing causal information can promote the accuracy of the Bayesian reasoning of the subject. And the second conclusion could provide support for our assumption about the mechanism of instruction effect. While the last one might indicate that, causal information has limited superiority for individuals at relative high metacognitive level.