Abstract
How does the prevalence of the target influence perceptual decisions? Levari et al (2018) made 2AFC decisions as to whether a dot on a blue-purple continuum was blue. When blue items were less prevalent, observers were more likely to call ambiguous stimuli “blue”. They
expanded their concept of blue (more liberal criterion). In contrast, Wolfe et al. find that observers become more conservative at low prevalence and find fewer targets in visual search experiments. What determines whether observers’ decision criteria become more liberal or more conservative when prevalence is low? In a series of replications of Levari’s study, we tested the effect of the response type (2AFC vs. Go/No-Go), color (blue-purple vs. red-green continua), stimuli type (solid color vs texture), and trial-by-trial feedback. Feedback appears to be the critical variable. In the presence of feedback, observers become more conservative at low prevalence. In the absence of feedback, they become more liberal, broadening the definition of less common target categories. Similar results were obtained with a shape continuum from rounded (“Bouba”) to spiky (“Kiki”). When Boubas are rare, ambiguous stimuli are called “Bouba” when there is no feedback. They are more likely to be called “Kiki” when there is feedback. There appear to be two, prevalence-based forces pushing and pulling decision criteria. These effects of prevalence have practical implications for low prevalence tasks like cancer screening. Do the pressures of prevalence cause experts to incorrectly categorize ambiguous stimuli as abnormal? In other work, multiple labs have shown that low prevalence causes experts to miss more targets. We will discuss whether category broadening and elevated miss errors might (unfortunately) coexist at low prevalence.