Many previous studies of learned predictiveness effects in animal and human learning indicate an advantage for cues that have previously been experienced as good predictors of outcomes over those that have been poorer predictors. These studies do not, however, reveal whether learned predictiveness exerts its effects at the level of learning or performance (or both). An experiment using human participants and a novel "mutant scientist" paradigm was used to investigate this issue. Results indicated that altering the learned predictiveness of cues after a stage of critical learning influenced responding to those cues, demonstrating that learned predictiveness must exert an influence on performance, in terms of responding to cues.