Conclusion of Experiment 6:
Researchers used three LMERs to find out which type of similarity best predicted possibility judgments. People were more likely to believe in events, if they saw them as similar to things that already happened (b = 0.51, SE = 0.03, p < .001), if they thought the world would appear similar if the event occurred (b = 0.24, SE = 0.02, p < .001), and if they viewed the events as similar to fictional occurrences (b = 0.12, SE = 0.02, p < .001). They compared all three models: past similarity was the strongest predictor (AIC = 12,991), followed by counterfactual similarity (AIC = 14,438, ΔAIC = 1,446), and lastly, fictional similarity (AIC = 14,746, ΔAIC = 1,755). Neither the counterfactual nor fictional similarity models were substantially better than the best model based on past similarity.
