Reducing derivation in arbitrary relations produces natural-like responding patterns, but nothing shakes the happiness superiority effect

Using the Implicit Relational Assessment Procedure (IRAP), we established (training-IRAP) and assessed (testing-IRAP) arbitrary derived relations in a total of 139 participants across the following conditions: relations in mutual entailment (A–B) (n = 44), relations in combinatorial entailment (A–B–C) and high derivation (no opportunities to derive the novel relations) (n = 49), and relations in combinatorial entailment with low derivation (128 trials to derive the novel relations) (n = 46). Stimuli in set A were pictures of happy and fearful faces, in set B were two nonsense words (“dreate” and “wippel”), and in set C were two abstract figures. The verbal cues for relational control were “similar” or “different”. The trained relations between two given sets of stimuli were fourfold (e.g., happy is similar to dreate, happy is different from wippel, fear is different from dreate, and fear is similar to wippel). Different combinations of stimuli to be related as “similar” or “different” were counterbalanced across participants. The assessed relations were between the set C symbols with happy and fearful words (e.g., “cheerful” and “fearful”), using “true” and “false” as relational coherence indicators participants could pick as their response per trial. Results showed that relating happy symbol with happy word was always quicker than relating fearful symbol with fearful word. Participants in low derivation showed the most similar pattern to that of natural relations. Starting the testing-IRAP with the inconsistent block always produced quicker responses in coherent blocks than starting with the consistent block.