Efficient memory for targets in visuomotor sequential search task

We examined retrospective and prospective memory effects in a multiple-target, visuomotor search. Before each trial, subjects were given a cue sequence consisting of 4 target letters in immediate alphabetical order. They were then shown a search display of 8 letters: the 4 cued targets and 4 distractors. The letters were centered in 8 disks which were randomly placed in a 10 x 10 degree viewing square. The task was to use the mouse to move the cursor quickly to each target disk and click on it in the correct order. There were two conditions in Experiment One. In the Vanish condition, clicking on a target removed it from the screen. In the Remain condition, targets were unaffected by the click and always remained visible. In this first experiment, reaction times (RT) were identical in the two conditions. RTs to the second target were substantially (700 ms) faster than RTs to the first target, with RTs to targets 3 and 4 declining by a further 80 ms per target. This pattern suggests that multiple targets may have been located before the first response was made. In Experiment Two, after subjects clicked on a disk, the locations of letters ahead of that target in the alphabetical sequence were shuffled. This manipulation prevented subjects from planning ahead. This eliminated the RT gap between the first target and subsequent targets. RTs were overall elevated compared to Experiment One, and RTs in the Remain condition were consistently 100 ms slower than the corresponding RTs in the Vanish condition. RT declined linearly with each target by 80 ms per target in both conditions. This may indicate that subjects have a memory for prior target locations (Gibson, Skow, Salvagni, Cooke, 2000; Horowitz Wolfe, 2001). Together, these results suggest a spatial memory mechanism which efficiently indexes up to 4 prior target locations. Data from longer sequence lengths and larger set sizes indicates that this memory is of limited capacity