Abstract
A new test determined whether a task's temporal limits are set by processes before or after selection. We use a moving window of attention to change rapid local alternation into a constant signal at the post selection level. Our two tasks were letter masking and feature pairing. Results: speed of letter masking was limited pre-selection whereas feature binding was limited post- selection. Here are 4 frames for the letter masking task (“≡” is the mask). Reading down a column shows that each location alternates in time between letter and mask.
Frame 1 A ≡ A ≡ A Reading across a row (across space in one frame) shows that adjacent Frame 2 ≡ A ≡ A ≡ locations alternate out of phase. Therefore, if we select an adjacent item on Frame 3 A ≡ A ≡ A each subsequent frame (follow along a diagonal), we always retrieve the Frame 4 ≡ A ≡ A ≡ letter. To guide attention, a circle marked one letter in a frame and moved to the adjacent letter on the next frame. In the nontracking condition without the guide circle, Ss monitored individual locations to report letter identity. Result: tracking the letter with moving attention did not increase the maximum rate at which Ss could report the letter. Therefore the mask had irretrievably degraded the letter prior to selection. On the other hand, tracking greatly improved the maximum speed in a feature binding task. For example, Ss reported letter pairs when shown AB alternating with CD, as on left below, or AD with CB, as on the right. Monitoring individual locations, Ss could only manage accurate report at 3Hz whereas with tracking (follow the diagonals), they could achieve rates of 5 Hz or higher. Here, the rate of feature pairing or binding is limited by processes following selection.
Frame 1 AB CD AB CD Frame 1 AD CB AD CB
Frame 2 CD AB CD AB Frame 2 CB AD CB AD
Frame 3 AB CD AB CD Frame 3 AD CB AD CB