Leonard J. Savage’s sure-thing principle (1954) is a key assumption of the consequentialist conception of decision making under uncertainty, which states that if some option x is preferred to y given some other event A occurs, and if option x is preferred to y given this event A does not occur, then x should be preferred to y even when the outcome of A is unknown. Much theoretical and experimental research has examined whether the sure-thing principle was violated in a variety of situations. But, not as much focus has been placed on examining the reasons why it was violated. Two experiments were conducted to test the so called “reasons-based” account and “reluctance-to-think” account for the violation of the sure-thing principle in the present study. In Experiment 1, 60 participants in Singapore were recruited, who were presented a scenario similar to Tversky and Shafir’s (1992) vacation situation. The results showed that the mean reported choices for not knowing whether you passed or failed the exam (M = 4.13) was well between the mean reported choices of knowing that you passed the exam (M = 5.18) and that of knowing that you failed the exam (M = 3.13). A test of within-participants contrast (Helmert contrast) showed that there was no significant difference between the effect for not knowing whether you passed or failed the exam and the mean effect of knowing that you passed the exam and knowing that you failed the exam (F (1, 59) = .02, n.s.). Thus, no violation of STP was found. In Experiment 2, a 3 (decision condition: knowing that Event A occurred vs. knowing that Event A did not occur vs. not knowing whether A occurred, nested within participants)×2(scenario: product promotion vs. job performance)×2 (reason posed: one reason vs. two reasons) between-participants repeated factorial design were used to further test the “reasons-based” account and “reluctance-to-think” account. The results showed that there was a significant effect of decision condition (F (2,158) = 7.40, p < 0.01) in the designed two-reason scenario. A test of within-participants contrast (Helmert contrast) showed that there was a significant difference between the effect for not knowing whether event A occurred and the mean effect of knowing that event A occurred and knowing that event A not occurred (F(1,79) = 9.82, p < 0.01). Thus, a violation of STP was found. In the modified scenario with one-reason posed, participants’ the mean reported choices for not knowing whether event A occurred was well between the mean effect of knowing that event A occurred and knowing that event A not occurred. The main effect of the decision condition was not significant, F(1,79) = 0.13, p > 0.1. Thus, no violation of STP was found. These results showed that participants in the two-reason conditions violated the sure thing principle but satisfied the sure thing principle in the one-reason conditions as predicted by the “reasons-based” account. It indicated that the sure-thing principle was generally satisfied when decisions were based on a unique reason where it was known that Event A occurred as well as that A did not occur, but was sometime violated when decisions were based on two incompatible reasons. These results support the “reasons-based” account rather than the “reluctance-to-think” account. The empirical distinction between the “reasons-based” account and “reluctance-to-think” account, therefore, would deepen our understanding of the reasons for the violation of the sure thing principle. Theoretical and normative implications are discussed.