Abstract
A fundamental process in object recognition involves matching ‘new’ or recently encoded objects with ‘old’ or previously encoded objects. Recently encoded representations of old objects can be supported by current sensory input (i.e., the original image remains present in the visual field), or they can exist in a fully mnemonic manner, i.e., without current sensory support. Here, we asked whether matching processes were more efficient when an ‘old’ object's representation was (1) supported by sensory data or (2) was fully mnemonic (stored in visual working memory, WM). If matching requires representing the test (new) object at a high level (say, in visual WM), then matching might be faster when the ‘old’ object is maintained in visual WM without continuous (and potentially interfering) input from sensory sources. To test this, we presented observers with two faces (upright or inverted in different experiments) and required them to report “same” or “different” as quickly as possible. In Experiment 1, on half of trials both faces were presented simultaneously (sensory condition), and on remaining trials, one face was briefly (400 ms) presented and then followed by a 1500 ms retention interval prior to the presentation of the second, to-be-matched face (mnemonic condition). Response times were fastest for the mnemonic condition. In a second experiment, controlling for sensory load at the matching stage, we added another (sustained sensory condition) in which the first face remained visible during the retention interval and when the to-be matched face was presented at the end of the trial. We found that responses in the mnemonic condition remained fastest but only when faces were upright; face inversion gave the sustained sensory condition an advantage. The results of these experiments shed light on the processes contributing to high-level object representation and suggest important differences between sensory and mnemonic data.