Applying an attentional set to perceived and remembered features

Abstract

Previous research has examined our ability to attend selectively to particular features of perceptual objects, as well as our ability to switch from attending to one type of feature to another. This is usually done in the context of anticipatory attentional-set control, comparing the neural mechanisms involved as participants prepare to attend to the same stimulus feature as on the previous trial ("task-stay" trials) with those required as participants prepare to attend to a different stimulus feature to that previously attended ("task-switch" trials). We wanted to establish how participants maintain or switch attentional set retrospectively, as they attend to features of objects held in visual short-term memory (VSTM). We found that switching, relative to maintaining attentional set retrospectively, was associated with a performance cost, which can be reduced over time. This control process was mirrored by a large parietal and frontal amplitude difference in the event-related brain potentials (ERPs) and significant differences in global field power (GFP) between switch and stay trials. However, when taking into account the switch/stay GFP differences, thereby controlling for this difference in amplitude, we could not distinguish these trial types topographically. By contrast, we found clear topographic differences between preparing an anticipatory feature-based attentional set versus applying it retrospectively within VSTM. These complementary topographical and amplitude analyses suggested that anticipatory and retrospective set control recruited a qualitatively different configuration of underlying neural generators. In contrast, switch/stay differences were largely quantitative, with them differing primarily in terms of amplitude rather than topography.

Figure 1. Trial order schematic.

A trial order schematic, showing the timing of pro- and retro-cues, array and probe stimuli.

Figure 2. Topographical analyses.

A) The result of the group-level segmentation process. The different colors show the durations of the various clusters of topography, across the four conditions, for the cue-locked epoch. Each of these clusters can be summarised as a single topography – the mathematical average of all topographies within that cluster – shown below. B) The results of the within-subject fitting procedure: i) the mean duration for which the ‘red’ and ‘green’ topographies were the best fit across pro-cue and retro-cue trials, between 120 and 300 ms; ii) the mean duration for which the ‘yellow’, ‘orange’ and ‘blue’ maps were the best fit across pro-cue and retro-cue trials, between 300 and 650 ms; and iii) the mean duration for which the ‘yellow’ and ‘orange’ maps were the best fit between 300 and 650 ms. In all cases the error bars show the standard error of the mean. In all cases the ‘*’ denotes significant differences, the absence of this denotes non-significant differences.

Figure 3. Grand-average waveforms.

Grand-average waveforms, time-locked to the cue onset at 0 ms. These are shown separately for pro- (left-hand column) and retro-cue trials (right-hand column), for each recording site along the anteroposterior axis. Each waveform shown is the average of the left-, right-hemisphere and midline electrodes, with the solid lines representing task-switch trials and the dotted lines representing task-stay trials.