Studies using balanced materials have found that both feature-based comparison and thematic integration play a role inconcept organization (e.g. Mirman & Graziano, 2012; Murphy, 2001), a proposal backed up by neurological findings.This experiment crossed taxonomic and thematic relatedness of abstract vs concrete pairs to examine how these processesaffect perceived similarity. Participants rated similarity of 96 normed word pairs and explained ratings in writing. Linearmixed effect modeling revealed a 3-way interaction on ratings, with taxonomic relatedness affecting ratings more forconcrete than abstract pairs only when a thematic relation was absent. No other abstractness effects were observed. Forcoded explanations, a difference emerged only for pairs related both taxonomically and thematically: concrete pairs wereprocessed more frequently thematically than taxonomically, with the reverse pattern for abstract pairs. Further, qualitativeanalyses of the explanations and Bayesian analyses of the relation between explanations and similarity ratings will bepresented.

Fenker, Waldmann, & Holyoak (2005) found that participants are faster to recognize a causal relation between wordspresented in predictive (cause first) than in diagnostic order (effect first). We extended these findings to a comparison of abstractand concrete word pairs. Causality may play a more prominent role in abstract concept relations. Given that causally relatedabstract concepts are not always observable, they may often involve diagnostic reasoning (e.g. inferring motives). Across twoexperiments, participants made timed judgments of whether abstract and concrete word pairs of equal bidirectional associativestrength were causally related. Items were presented in blocks comprising pairs in either predictive or diagnostic order. Reactiontimes were significantly lower for predictive order compared to diagnostic. This was not moderated by abstractness, but therewas a slightly greater effect for concrete pairs. These data indicate that causal relations are likely stored in memory for bothabstract and concrete concepts.