Analysis Plan: Critical DV: We tested the basic responsibility difference by comparing good vs. bad in only the scenarios that participants rated first. For responsibility, this was done with a 1-way ANOVA on "responsible" with group as the IV. We tested whether including Extraversion:meek as a covariate changed the results. We predicted that people would hold the person more responsible for good, rather than bad behaviors. As an exploratory investigation, we also tested whether these results were moderated by cynicism (factor-analysis extracted scores using Maximum Likelihood estimation on a single factor). Further Investigations Responsibility: We then tested whether the asymmetry between good and bad was seen for the second judgement. We first tested whether being asked the ‘widgets’ question alters the second judgment (ANOVA: responsibility = DV, IVs = good/bad [binary], widgets in between [binary], interaction of widgets-in-between and good/bad). Providing no interaction, we collapsed across the groups of widgets in-between or after). Then, we just tested whether the asymmetry was seen in the second scenario. We predicted that it would not. To better understand why the asymmetry would not be seen, we tested our two possible mechanisms. The first was likelihood of getting the widgets question correct. This was done using a probit regression on getting the answer right (coded 1 = 5, 0 = all other answers), with the order (in-between, after, binary coding), whether the first scenario was the good one (binary coding) and their interaction. It was predicted that, if seeing the good scenario induces process 2 thinking, then we would see lower probability of getting the widgets question correct only for those who saw the bad scenario first and only in the in-between question (as all participants had seen the good scenario at the end of the 2 scenarios). We also tested whether participants submitted their answers faster in the bad-first compared to the other 4 groups. This was done using an ANOVA on time to submission with good/bad (binary coding), first or second scenario (binary coding) and their interaction. We predicted that, if seeing the good scenarios induce type 2 processing, then participants should respond fastest in the bad scenario but only for the first scenario they saw (as all participants would have seen the good scenario at the end of both scenarios). For identity: we used a multinomial logistic regression with good/bad as the only predictor variable. We also included a simple chi^2 analysis to view the pattern of results. We predicted that people would be more likely to say that the "real" Alex/Mark was the one who was both the person before and after the tumor in the good conditions, but in the bad conditions they would more likely be the one who "no longer..." (the one without the tumor). As an exploratory investigation, we tested whether these results are moderated by cynicism (composite from the cynicism subscale) Then, we tested whether people’s first judgments were different from their second judgement on responsibility for good vs. bad behaviors. If given two scenarios back to back, we predicted that this difference only occurs in the first, not the second scenario. This was done by testing the interaction of display order and whether the scenario was good with responsibility ratings as the DV, display order as the within-subjects variable, good/bad and the interaction of good/bad and display order as IVs.