The function of visual memory in scene perception
The research reviewed thus far has examined the capabilities of visual memory and the means by which memory is used to construct visual representations of scenes. I turn now to the question of the function of visual memory in scene perception. Given that participants can generate robust internal representations of a scene, how and to what purpose is this information used? I first consider the functional role of VSTM and then considerVLTM function.
The function of VSTM in scene perception
The literature on VSTM has seen a remarkable surge in research over the last decade (for a review, see Luck, 2008). Most of this research has sought to understand the capacity ofVSTM and the format ofVSTM representations, but the functional purpose of the VSTM system has received relatively little attention. After discussing two common accounts ofVSTM function, I argue that VSTM supports perceptual comparison operations that are required almost constantly during real-world perception and behavior.
VSTM and conscious awareness
A common proposal regardingVSTM function is thatVSTM forms the substrate of visual awareness (Becker & Pashler, 2002; Rensink, 2000; Rensink et al., 1997). In particular,VSTM is thought to reflect activation of the currently attended portion of a visual scene, with constraints on attentional capacity and VSTM capacity reflecting two sides of the same coin (Cowan, 1995; Rensink, 2000). However, it is highly unlikely thatVSTM plays any direct role in visual awareness.VSTM representations are not visible and thus are unlikely to be the substrate of visual awareness; one does not continue to see the items held in VSTM once they have been removed. For example, one does not see remembered items as persisting during an ISI between study and test images in a change detection task (as in Luck & Vogel, 1997). It is this very property ofVSTM—that it is not visible—that distinguishes VSTM from visible persistence (iconic memory), which is visible (Coltheart, 1980).
IfVSTM does not form the substrate of visual experience, then the fact that we can only hold three to four objects in VSTM does not necessarily mean that our visual awareness of a scene is limited to three to four objects. Indeed, Sperling (1960) showed that we see a great deal more than we can hold in VSTM. When participants were shown arrays of 12 letters in Sperling’s task, they saw 12 letters in the brief moment that they were visible, but they could only transfer three to four letter identities into STM for subsequent report. A quick demonstration proves this point. One tells a naive participant to view a briefly presented visual display. Then one presents an array of 12 letters for 50 ms (as in Sperling, 1960). What observers report is that there were 12 letters, but they can only report the identity of three to four of them. Because it is easy to report that there were 12 letters (and not six letters or three letters), participants must have seen 12 letters when they were visible. If visual awareness was limited to the capacity ofVSTM, then participants should have reported that there were only three to four letters present. The issue here is that, because the report of what one saw requires memory, limitations on memory can easily be confused with limitations on perceptual experience (Chun & Potter, 1995; Moore & Egeth, 1997;Vogel & Luck, 2002; for a full discussion, see Wolfe, 1999).