Attention Effects during Visual Short-Term Memory Maintenance: Protection or Prioritization?

Article excerpt

Interactions between visual attention and visual short-term memory (VSTM) play a central role in cognitive processing. For example, attention can assist in selectively encoding items into visual memory. Attention appears to be able to influence items already stored in visual memory, as well; cues that appear long after the presentation of an array of objects can affect memory for those objects (Griffin & Nobre, 2003). In five experiments, we distinguished two possible mechanisms for the effects of cues on items currently stored in VSTM. A protection account proposes that attention protects the cued item from becoming degraded during the retention interval. By contrast, a prioritization account suggests that attention increases a cued item's priority during the comparison process that occurs when memory is tested. The results of the experiments were consistent with the first of these possibilities, suggesting that attention can serve to protect VSTM representations while they are being maintained.

The goal of the present study was to examine the relationship between attention and visual short-term memory (VSTM). Specifically, we were interested in whetiier attention can influence representations already stored in VSTM. If attention affects representation maintained in VSTM, how does attention select these representations after the iconic image of stimuli has already faded? We examined two possible mechanisms underlying attentional selection from VSTM.

In the area of visual memory research, several studies have demonstrated that VSTM can hold or maintain a small number of items across the delay produced by an eye movement (Irwin, 1992; Irwin & Andrews, 1996) or across a temporal delay without an eye movement (Luck & Vogel, 1997; Vogel, Woodrow, & Luck, 2001). In the latter case, a change-detection task is commonly used to assess the information stored in VSTM. In a typical changedetection task, observers are shown a memory array and a test array, separated by a brief delay interval, and observers are asked to report whether the two arrays were identical or different along one or more feature dimensions. When color is the relevant dimension, for example, the memory array contains a set of colored squares, and the test array is either identical to the memory array or differs in the color of one item. Accuracy is typically high when there are only one to three objects on the memory array and lower when the number of objects in the memory array increases, suggesting that VSTM has a capacity of approximately three or four objects.

Attention is widely believed to play an important role in VSTM (Bundesen, 1990; Cowan, 1997; Duncan & Humphreys, 1989). Many studies have shown that attentiondirecting cues can influence which items are encoded in VSTM if the cues appear before stimulus offset, or before the iconic image of the stimulus has faded (Averbach & Coriell, 1961; Becker, Pashler, & Anstis, 2000; Griffin & Nobre, 2003; Landman, Sperkreijse, & Lamme, 2003; Schmidt, Vogel, Woodman, & Luck, 2002; Sperling, 1960; Woodman, Vecera, & Luck, 2003). In these cases, it is thought that attention selects a perceptual representation for entry into VSTM (see Schmidt et al., 2002; Vogel, Woodrow, & Luck, 2006).

Although attention can influence the storage of items in VSTM when attention is directed to an object before it appears (or shortly after it disappears), many findings suggest that attention does not influence VSTM when attention is directed long after the presentation of objects (i.e., beyond the range of iconic memory; Averbach & Coriell, 1961; Becker et al., 2000; Sperling, 1960). In Sperling's partial report procedure, for example, three rows of alphanumeric characters were presented briefly, followed by a delay and a report tone that indicated which row of letters to report. The observers' task was to report the characters from the cued row. Sperling found that the iconic trace of the memory array had largely faded after 500 msec, and no cuing effect was observed when auditory cues were presented 1 sec after th array offset. …