How Visual is Visual Prediction?

Abstract

Infants are readily able to use their recent experience to shape their future behavior. Recent work has confirmed that infants generate neural predictions based on their recent experience (Emberson, Richards, & Aslin, 2015) and that neural predictions trigger visual system activity similar to that elicited by visual stimulation. This study uses behavioral methods to ask, how visual is visual prediction? In Experiment 1, we confirmed that when additional trials provide additional visual experience with the experimental shape, infants exhibit a robust novelty preference. In Experiment 2, we removed the visual stimulus from some trials and presented the predictive auditory cue alone, allowing the effects of neural prediction to be assessed. We found no evidence of looking preferences at test, suggesting that visual prediction does not contribute to the computation of visual familiarity. In Experiment 3, we provided infants with a degraded visual stimulus to test whether visual prediction could bias visual perception under ambiguous conditions. Again, we found no evidence of looking preferences at test, suggesting that visual prediction is not biasing perception of an uncertain stimulus. Overall, our results suggest that visual prediction is not visual, in the strictest sense, despite the presence of visual system activation.