Résumé
There is a long-standing debate as to whether violations of rational choice axioms reflect normative deviations from the theory or simply mistakes. We contribute to this debate by reproducing and replicating Nielsen and Rehbeck's (2022) experimental study, with a new focus on heterogeneity across axioms. We conduct a three-part analysis comprising a direct computational reproduction, a robustness reproduction, and a high-powered preregistered replication (N = 451) focusing on the Independence of Irrelevant Alternatives (IIA). We find robust evidence that individuals express a desire to follow canonical axioms, but perceive violations as mistakes only for a subset of them-specifically, IIA and Transitivity. In contrast, we find no evidence that violations of Independence, First-Order Stochastic Dominance, Branching, or Consistency are perceived as mistakes. We discuss these findings through the lens of cognitive complexity, suggesting that individuals may fail to recognize violations of more demanding axioms even when prompted.