Expected utility theory, the normative benchmark for evaluation of risky prospect, implies that the evaluation is linear in probability, that it is concave in the outcome, and that probabilities and outcomes are multiplied. The present study examines how the evaluations of risky prospects are affected by the availability of comparative anchors. We report an experiment comparing a within-subjects design (WSD), in which 20 participants evaluated 36 prospects, with a single-subject design (SSD), in which each of 720 participants evaluated a single prospect. The results of fitting cumulative prospect theory (Tversky & Kahneman, 1992) to data showed that in the WSD, there was a roughly linear probability weighting function and a concave value function, as suggested by expected utility theory. In the SSD, on the other hand, there was a linear value function and a severely nonlinear weighting function for probability. The participants thus found it difficult to maintain the linear use of single-event probability disclosed in the WSD when they made evaluations in an SSD without access to comparative anchors. We argue that people may have much of the normative competence captured by expected utility theory, but this competence can only be manifested as normative performance given the availability of relevant comparative anchors. We discuss the possibility that this could explain why some economic markets are deemed rational, whereas others are not. (PsycINFO Database Record (c) 2018 APA, all rights reserved)