Attention's Accelerator

Abstract

How do people get attention to operate at peak efficiency in high-pressure situations? We tested the hypothesis that the general mechanism that allows this is the maintenance of multiple target representations in working and long-term memory. We recorded subjects' event-related potentials (ERPs) indexing the working memory and long-term memory representations used to control attention while performing visual search. We found that subjects used both types of memories to control attention when they performed the visual search task with a large reward at stake, or when they were cued to respond as fast as possible. However, under normal circumstances, one type of target memory was sufficient for slower task performance. The use of multiple types of memory representations appears to provide converging top-down control of attention, allowing people to step on the attentional accelerator in a variety of high-pressure situations.

Keywords: attention; cognitive neuroscience; evoked potentials; open materials; visual memory; visual search.

Fig. 1.

Illustration of the task and stimuli used in Experiments 1 and 2. Each trial of Experiment 1 began with a cue indicating the reward level for that trial and then a cue identifying the target (a). After a blank screen, the search array was presented for 2,000 ms. A given target object was cued and searched for across seven trials, and then a new target was selected. The large- and medium-reward cues could appear anywhere across the target repetitions within a run, but were mostly clustered at Target Repetition 5. The trials and procedure of Experiment 2 were similar (b), except that the first cue indicated whether the trial was to be performed with a balance of speed and accuracy (normal baseline) or as fast as possible (go fast).

Fig. 2.

Results from Experiment 1. The graphs in (a) and (b) show reaction time (RT) and error rate in the small-reward condition as a function of target repetition, along with mean RT and error rate on Target Repetition 5 in the other two reward conditions. The grand-average waveforms in (c) are from the fronto-central electrode and show the anterior P1 amplitude (within the shaded area) on the critical Target Repetition 5 trials in each reward condition. The grand-average waveforms in (d) are from the lateral parietal, occipital, and temporal electrodes on Target Repetition 5 trials. The contralateral delay activity (CDA) is the difference (indicated by the gray shading) between waveforms from sites contralateral and ipsilateral to the target cue; separate waveforms are shown for each reward condition. The x-axes in (c) and (d) show the timing of stimulus presentation relative to the waveforms, with presentation of the target cue marked in yellow, from 0 to 100 ms, and presentation of the search array marked in blue. The graphs in (e) and (f) show the mean amplitudes of the anterior P1 and CDA as a function of target repetition in the small-reward condition, along with their mean amplitudes on Target Repetition 5 in the other two reward conditions. In (a), (b), (e), and (f), error bars represent ±1 SEM, and the shaded areas highlight results for Target Repetition 5.

Fig. 3.

Results from Experiment 2. The graphs in (a) and (b) show reaction time (RT) and error rate in the normal baseline condition as a function of target repetition, along with mean RT and error rate on Target Repetition 5 in the go-fast condition. The grand-average waveforms in (c) are from the fronto-central electrode and show the anterior P1 amplitude (within the shaded area) on the critical Target Repetition 5 trials in each speed condition. The grand-average waveforms in (d) are from the lateral parietal, occipital, and temporal electrodes on Target Repetition 5 trials. The contralateral delay activity (CDA) is the difference (indicated by the gray shading) between waveforms from sites contralateral and ipsilateral to the target cue; separate waveforms are shown for each speed condition. The x-axes in (c) and (d) show the timing of stimulus presentation relative to the waveforms, with presentation of the target cue marked in yellow, from 0 to 100 ms, and presentation of the search array marked in blue. The graphs in (e) and (f) show the mean amplitudes of the anterior P1 and CDA as a function of target repetition in the normal baseline condition, along with their mean amplitudes on Target Repetition 5 in the go-fast condition. In (a), (b), (e), and (f), error bars represent ±1 SEM, and the shaded areas highlight results for Target Repetition 5.