**Study 5 Description** Throughout these studies, I have assumed that hindsight bias and belief updating are interrelated phenomena. Without preregistration, individuals can retroactively alter their predictions to be more in line with evidence. If individuals consider the gap between their predictions and evidence when updating their beliefs, then smaller gaps should lead to less updating. In Study 5, I seek to test this mechanism for the effect of preregistration on belief updating. *Exploratory Mediation Analyses - Studies 1a and 1b* My hypotheses for this mechanism came from exploratory analyses of Studies 1a and 1b. After discovering the differences in predicted vs. actual performance on the anagrams task between the conditions, I then explored the difference of predicted time and actual time spent on the anagrams as a mediator of the effect of preregistration. It may be that participants were using the difference between their predictions and the outcome of the test to inform their beliefs about their verbal ability. If participants in the hindsight condition were shifting their predictions to be more in line with their outcome (enabling them to match their prediction more often than those in the preregistration condition), this may lead them to update their beliefs about their verbal ability less. Indeed, the difference was lower in the hindsight condition between the participant’s predictions and the amount of time they spent on the task (M = -23.01, SD = 44.01) compared to participants in the preregistration condition (M = -44.39, SD = 44.83), d = .48, 95% CI [.41, .55]. Also, across conditions, the difference between predicted and actual time spent on the task was correlated with change in beliefs about verbal ability, r(3,127) = .21, 95% CI [.17, .24]. In support of the proposed mediator, there was an indirect effect of difference between predicted and actual time spent on the anagrams task on belief updating, (1000 iterations bootstrapped estimate of indirect effect = .18, 95% CI [-.22, -.14]). That is, the effect of preregistration on belief updating is accounted for, in part, by restricting participants’ ability to alter their predictions to be more in line with their actual performance. These results were generated by a measurement of mediation design (e.g., Baron & Kenny, 1986; Spencer, Zanna, & Fong, 2005). The causal claim of this design assumes that the mediator covaries with the outcome measure, that the mediator occurs before the outcome measure, and that it is unconfounded (Cook, Campbell, & Shadish, 2002). In this case, the difference between predicted and actual time spent on the anagrams task was related to belief updating. Also, participants were able to assess this difference, by making their predictions and completing the task, before re-rating their verbal ability. However, the design does not sufficiently rule out other possible explanations for the detection of this mediation effect. It could be that the preregistration manipulation independently leads to larger differences in predicted compared to actual performance on the anagrams task and greater belief updating with both effects being caused by an unmeasured mediator. To address this limitation, I will attempt to manipulate the proposed mediator in Study 6 (e.g., Pirlott, & MacKinnon, 2016). Half of participants will complete the same procedures as Studies 1a and 1b. The other half will complete a very similar procedure but not be told how long they have to complete the anagrams task and will not be shown a timer when completing the task. This should make it much more difficult for participants to know the difference between their predicted and actual performance on the task. If the effect of preregistration is mediated by the difference between predicted and actual performance, then mediation should only be detected when participants can readily know this difference. However, if the effect of preregistration on belief updating and the proposed mediator are independent of one another, both effects could be detected regardless of procedure.