Research suggests that the relational completion procedure (RCP) is effective for studying derived relations of same and opposite. Previously, procedural parameters, such as the presence or absence of a confirmatory response requirement, were found to have a facilitative effect on the number of training trials to criterion and overall arbitrary test pass rate (i.e., yield). These experiments report on the manipulation of additional potentially important parameters--the presence or absence of selection-based feedback, 4 vs. 8 trial types and of a linked nonarbitrary and arbitrary training phase--across both the RCP and matching-to-sample (MTS) protocols with either 3 or 5 comparisons. During selection-based feedback, after the production of the confirmatory response, the sample, contextual cue, and selected comparison were presented along with corrective feedback. During the linked nonarbitrary-arbitrary phase, in the presence of the sample stimulus and the same and opposite contextual cues, respectively, selections of a nonarbitrary comparison stimulus at either end of a specified physical dimension were reinforced. Findings indicated a trend for improved performance with the inclusion of selection-based feedback and the linked nonarbitrary-arbitrary phase. There was a significant difference in yield between the RCP and MTS conditions (84% vs. 56%, respectively). The implications of the findings for contemporary research on derived relational responding are discussed. (Contains 5 tables, 3 figures, and 1 footnote.)