Using the blocked-translation paradigm with healthy participants, we examined Crutch and Warrington's hypothesis that concrete and abstract concepts are organized by distinct principles: concrete concepts by semantic similarities and abstract ones by associations. In three experiments we constructed two types of experimental blocking (similar vs. associative) for both abstract and concrete words. In Experiment 1, we first attempted to transfer previous findings observed in patients by Crutch and Warrington with semantic impairment to healthy participants. In Experiment 2 only noun stimuli were used, and we further included two semantically categorical conditions that differed by a degree of semantic similarity (close vs. distant). In Experiment 3, verbs were used exclusively. Consistent results were obtained across all three experiments: Significant interference effects were observed for abstract items that were blocked by an associative relationship and by a semantic similarity, and for concrete items that were blocked by a semantic similarity (category) but not when they were blocked by an association. The effect of similarity-close was greater than that of similarity-distant in the noun experiment. We argue that the results are in conflict with Crutch and Warrington's proposals, and can be accommodated by a theory of cooperating similarity and association connections for concrete and abstract concepts, with the association bearing more weight for abstract concepts.