ABSTRACT
Intradimensional discrimination training may cause peak shift, in which participants respond more frequently to a novel stimulus presented during a generalization test than the positive exemplar used in training. Previous research has shown that peak shift is most likely to occur in participants who have achieved an intermediate level of proficiency with the discrimination. We sought to examine whether discrimination learning and peak shift could be altered through variations in stimulus discriminability. An international sample of 117 adults were trained to discriminate one visual representation of risk from COVID-19 (S+) from a second level of risk (S-) that was either lesser or greater. In a high discriminability condition, a single indicator of risk (a bar length) was presented on each trial. In a moderate discriminability condition, participants were required to estimate a person’s risk from multiple risk indicators. In a low discriminability condition, participants were additionally required to consider risk mitigation factors when estimating a person’s risk. Peak shift was absent in the high discriminability condition but present in the moderate condition. The low discriminability condition produced either flat or monotonic generalization gradients. The results additionally demonstrate how presenting health risk information to people in relatively simple or relatively complex ways affects their ability to judge that information correctly.