Abstract
Suppose you are making decisions about whether a greenish-yellow spot is green or not. If truly green spots are rare (low prevalence) and if you receive trial-by-trial feedback, you will be less likely to label ambiguous spots as green (classic low prevalence effect - LPE). If you do not receive feedback, the effect can reverse. You will be more likely to call the same spot green (prevalence induced concept change – PICC, see Levari et al, 2018). Now suppose that each item is defined by more than one feature as might be the case in complex decisions about suspicious spots on lung CT, for example. We used two-dimensional stimuli that varied in shape from “bouba” (rounded) to kiki (spikier) and in color (yellow-green). We could change the prevalence along one or both feature dimensions and we could provide feedback or not. Observers made 2AFC, “green kiki” versus “not green kiki” responses. When prevalence was reduced for one dimension (e.g. fewer green items, overall), effects were seen on the other dimension (i.e. the bouba/kiki boundary moved). Feedback produced LPE. Without feedback, PICC effect was seen, but more weakly. In Experiment 2, stimuli were drawn from separate green and yellow bouba-kiki continua. Observers made 3AFC responses (“green kiki”, “yellow kiki”, or “not kiki”). Prevalence of green kiki was decreased while prevalence of yellow kiki stayed the same. Changing the prevalence of green stimuli showed more effect on green than yellow stimuli, but prevalence effects, especially LPE, spread to yellow stimuli. PICC effects were again weaker in this experiment, implying that LPE and PICC effect, though coexist at low prevalence, may be mediated by distinct underlying processes. These results suggest that in complex real-world decisions, prevalence of one type of target may influence decisions about related, but distinct targets.