Not all perceptual difficulties lower memory predictions: Testing the perceptual fluency hypothesis with rotated and...

Studies typically show that perceptual difficulties at the time of encoding lower memory predictions. One potential exception to this is the inverted-word manipulation, in which participants produce equivalent memory predictions for upright and inverted words, despite higher free-recall performance for the inverted words (Sungkhasettee, Friedman, & Castel in Psychonomic Bulletin & Review, 18, 973-978, 2011). In the present set of experiments, we aimed to investigate the contributions of online perceptual difficulties versus a priori beliefs through two disfluency manipulations conceptually similar to the inverted-word manipulation: inversion and canonicity. The inversion manipulation involved presentation of upright and inverted object images, whereas the canonicity manipulation involved presentation of objects to participants from frequent (canonical) or infrequent (noncanonical) viewing perspectives. Memory predictions were made either on an item-by-item basis or aggregately. In all studies, the perceptual identification latencies for inverted and noncanonical items were slower than those for upright and canonical items, respectively. In experiments conducted with item-by-item memory predictions, predictions were not significantly different from each other across encoding conditions. In contrast, in experiments using aggregate memory predictions, fluent items produced higher memory predictions than did disfluent items. These results show that in certain cases, participants may not consider online objective perceptual difficulties. Moreover, item-by-item and aggregate memory predictions produce different patterns, evidence of a dissociation between the two types of predictions. The results are discussed in light of theories that rely on objective perceptual fluency differences across encoding conditions versus theories that rely on participants' a priori beliefs about fluency.