People make conjunction errors, rating a conjunction as morelikely than one of its constituents, across many different typesof problems. They commit the conjunction fallacy in problemsof social judgment, in physical reasoning tasks, and in gam-bles of pure chance. Doctors commit the fallacy when mak-ing judgments about hypothetical patients. Do all these errorsshare an underlying cause? Or does the fallacy arise indepen-dently in different types of reasoning? In a series of studies, welook for structure in conjunction errors across various types ofproblems. We find that error magnitudes are related for someclusters of items, but there does not appear to be a universalrelationship between all cases of this fallacy.

Ludwin-Peery, Bramley, Davis, and Gureckis (2019) reported finding evidence of conjunction fallacy errors in an intuitivephysics reasoning task. However, this finding was limited to a single paradigm involving the behavior of only two objects,interacting in a consistent manner, in a highly regular setting. In this project, we provide an important generalizationtest of this result, and examine several new paradigms under which conjunction errors might be observed. We find somecases that produce the expected errors, representing an important generalization of the original finding, as well as someparadigms which do not appear to produce conjunction errors.

In this paper, we describe three experiments involving simplephysical judgments and predictions, and argue their results aregenerally inconsistent with three core commitments of proba-bilistic mental simulation theory (PMST). The first experimentshows that people routinely fail to track the spatio-temporalidentity of objects. The second experiment shows that peopleoften incorrectly reverse the order of consequential physicalevents when making physical predictions. Finally, we demon-strate a physical version of the conjunction fallacy where par-ticipants rate the probability of two joint events as more likelyto occur than a constituent event of that set. These results high-light the limitations or boundary conditions of simulation the-ory.