Abstract
The perceptual wink model of the Attentional Blink (AB) assumes that the AB is a perceptual deficit, reflecting a failure to perceive that the second target belongs to the target category. Providing a unified account of the AB and Repetition Blindness (RB), we augmented the perceptual wink model with a Bayesian decision process that compares the observed evidence in short-term memory against evidence priors to determine how many times a particular identity appeared. This unified explanation of the AB and RB does not require type-token binding; in lieu of tokenization, performance is based on the magnitude of evidence for each type. Chun (1997) examined RB in a letter-number attentional blink AB task, finding that some manipulations selectively reduced the AB while others selectively reduced the RB. The perceptual wink model is a dynamic neural network with perceptual habituation, and explained these dissociations as reflecting perceptual habituation for a character’s appearance/identity (i.e., which letter or number?) in the case of the RB versus perceptual habituation for alphanumeric category (i.e., is it a number or a letter?) in the case of the AB. We assessed the unified model with Chun’s AB/RB paradigm by manipulating the category mapping; one group of subjects received consistent mapping, with a set of characters consistently assigned to the target category (e.g., always letters or always numbers), while another group received varied mapping, with variation across trials for the target category (e.g., letters on some trials and numbers on other trials). As predicted, the category mapping manipulation affected RB and the AB in a similar manner. Multiple-choice testing confirmed the prediction that in the midst of both the AB and RB, participants would claim that the trial only contained one target, as expected from a failure to perceive that the second target belonged to the target category.