Belief bias in syllogistic reasoning refers to the finding thatindividuals are more likely to accept believable than unbeliev-able conclusions independent of their logical validity. Mosttheories argue that belief bias is driven by differences in rea-soning processes between believable and unbelievable syllo-gisms. In contrast, Dube, Rotello, and Heit (2010) proposedthat belief bias is solely an effect of response processes. Weinvestigated belief bias without having to rely on response biasmanipulations (Klauer, Musch, and Naumer, 2000) or confi-dence ratings (Dube et al., 2010). Instead, we added a thirdresponse (“I don’t know”) to the usual binary response set(“Yes”/“No”). This allowed us to test belief bias with a fullyidentified multinomial processing tree model, in a hierarchicalBayesian framework. We found evidence that the belief biasis driven by differences in response processes. Evidence for adifference in reasoning processes was inconclusive.

Trials in a two-alternative forced-choice (2AFC) recognition-memory task require individuals to choose the stimulus in apair that they deem as having been previously studied. Be-cause of the relative nature of the judgments made, 2AFC tri-als are typically considered to be free from response biasesconcerning the old/new status of stimuli. Recent studies havesuggested that this assumption is incorrect, and individuals of-ten resort to single-stimulus old-new (ON) judgments instead.The present study tests this claim by joint modeling 2AFCand ON judgments using extended SDT models that includethe possibility of ON contamination. Results show that therelative-judgment assumption provides an excellent account ofthe data, providing no support for the notion of ON contami-nation in typical experimental designs.