A common method for studying derived relations is the matching‐to‐sample (MTS) preparation. However, certain aspects of its training format potentially hinder the emergence of new relations. The training version of the implicit relational assessment procedure (training‐IRAP) may present an alternative. Our primary objective involved comparing the effectiveness of delayed MTS(2s) and training‐IRAP procedures on participant yield. The secondary objective involved comparing mean number of training blocks per procedure. Given additional components in the standard training‐IRAP not found in MTS, changes were made to the former, producing the modified training‐IRAP and delayed modified training‐IRAP(2s). Sixty‐eight typically developing students participated in a between‐subjects design. Two classes comprising five abstract stimuli were employed. Yield was analyzed at three levels, 91.67, 83.33, and 79% correct responses, with at least 87.5% correct responses at baseline mixed‐block maintenance. All participants maintained baseline criterion during tests. At the three levels of analyses, the modified versions of training‐IRAP produced higher yield, followed by DMTS(2s) and then the standard training‐IRAP. Mean number of blocks to complete training phases was lowest for the delayed MTS(2s) and delayed modified training‐IRAP(2s) groups. Limitations and implications of the findings toward greater precision, scope, and depth in conceptual, experimental, and applied settings are discussed.