Choice-dilemmas items were administered with the customary risk-taking (R) format (minimal acceptable odds to warrant recommendation of the desirable risky alternative) and with a pessimism-optimism (P) format (estimation of the actual "real-world" probability of success of the desirable risky alternative). Following individual testing to establish baseline levels, subjects were randomly assigned to either an R-P or P-R treatment. In the former, discussion to consensus was focused on the R dimension, and subsequent P judgments were made privately; in the latter, discussion to consensus was focused on the P dimension, and subsequent R judgments were made privately. Results indicated that the link between R and P was asymmetrical. The risky shift was produced only when discussion was oriented along the R dimension. In contrast, pessimistic shifts were obtained following both R and P discussions. Group-induced pessimism was attributed to pooling or heightened realization of the "things that might go wrong" in pursuing a risky course of action.