The study of relational categories has emphasized the importance of within-category comparison for learning and transfer – guided by predictions from structure mapping theory (Gentner, 1983). Recent research has yielded the predicted comparison advantage under the supervised observational learning mode (over sequential exposures) but, puzzlingly, not under the supervised classification mode. In the present study we evaluate performance pressure as a possible contributor to the ineffectuality of comparison under classification by crossing performance pressure (elevated and standard) with two classification learning formats (single-item and within-category pairs). We found: (1) that pressure hindered single-item learning, but not comparison learning; and (2) a novel comparison advantage under standard classification. We conclude: (1) that performance pressure exerts a deleterious effect on relational category learning that comparison may compensate for; and (2) that pressure does not seem to underlie lackluster classification + comparison performance (relative to observational learning). Implications and new evidence are discussed.