The feature positive effect in the face of variability: Novelty as a feature

Past literature has shown, across various methods and species, that feature positive (FP) tasks (AB+/B-) are learned more easily than are feature negative (FN) tasks (AB-/B+), giving rise to what is known as the feature positive effect (FPE). Experiment 1 was intended to assess the role of expected absence in the FPE through manipulation of the context (constant vs. varied common elements) that was paired with the feature. The results indicate that novel contextual elements (varied common elements) may be salient enough to eliminate the FPE, creating a feature negative effect (FNE). Experiment 2 replicated the FNE. The addition of a testing phase confirmed that participants in the FN task judged novel stimuli to be strong positive predictors whereas participants in the FP task did not, thus producing a novelty FPE. These results are problematic for contemporary associative learning models.