Memory-driven capture occurs for individual features of an object

The human visual system is constantly bombarded with information, requiring selective attention to prioritize relevant information for processing and storage. Selective attention can operate on objects or features, improving performance and enhancing brain responses related to task-relevant elements while suppressing irrelevant ones. Similarly, internal attention (selection within working memory, WM) allows us to prioritize or remove information based on relevance. Studies have shown that retro-cuing specific features improves memory for those features. This study aimed to investigate the interaction between internal feature-specific prioritization in WM and memory-driven capture – the automatic capture of attention by distractors matching memorized information. The hypothesis was that prioritizing specific features in WM would modulate memory-driven capture, reducing or eliminating capture for no-longer-relevant features.

Existing research extensively demonstrates the interplay between perceptual attention and selective attention operating on stored information in WM. Memory-driven capture, where memory contents automatically capture attention during search, has been well-documented. Studies have shown that this automatic capture is reduced or eliminated when internal representations are retro-cued as irrelevant, suggesting that WM prioritization modulates the attention-memory interaction. However, this interaction has not been examined at the feature level, despite evidence that WM representations can be unbound into separate features, and focusing attention on task-relevant features leads to feature-specific memory benefits. This gap motivated the current research.

The study comprised two experiments. Both experiments involved memorizing the color and orientation of an isosceles triangle. A retro-cue then indicated which feature (color, orientation, or both) was relevant for a subsequent continuous report task. Participants performed a visual search task, and performance was compared across conditions with and without memory-matching features in the display. Experiment 1 used a central retro-cue (color, orientation, or both). Eighteen participants performed 396 trials. Stimuli were isosceles triangles with varying colors and orientations. Search reaction times (RTs) and memory errors were measured. Experiment 2 manipulated the location of the retro-cue (small letter C or O for color or orientation) – it appeared at the target location (40% of trials) or at the distractor location (60%). Eighteen different participants performed trials under this condition. The “cue both” condition was removed to ensure a larger number of trials per condition to test for the impact of cue location. Stimuli and apparatus were identical to Experiment 1 except for the cue form and location. Search accuracy and RTs were measured. Data analysis involved repeated-measures ANOVAs and planned comparisons to examine the effects of cues and feature matches on search performance and memory accuracy.

Experiment 1 showed better memory for retro-cued features. Crucially, color-matching distractors captured attention only when color remained relevant. Orientation showed no capture effect. Experiment 2, with spatially cued retro-cues, revealed color capture in all cue conditions and orientation capture when orientation was relevant, showing a greater capture effect when it was cued as relevant. The location of the cue (target vs. distractor) did not significantly affect capture, suggesting that focused attention broadly, rather than attentional location, is crucial. However, the manipulation of cue location in Experiment 2 enabled the detection of capture by orientation.

The findings provide the first direct evidence of dynamic interaction between internal and external attention at the feature level, supporting the idea that WM representations can be decomposed into individual features. The lack of orientation capture in Experiment 1, and its appearance in Experiment 2, suggests orientation capture depends on focused attention, unlike color. The results suggest that internal feature-specific prioritization modulates memory-driven capture, but also highlight that feature salience influences capture effects. One possible mechanism is slower disengagement of attention from memory-matching features.

This study demonstrates that internal feature-specific prioritization within objects modulates memory-driven capture, showing a dynamic interplay between WM and attention at the feature level. Future research should investigate the underlying mechanisms responsible for this modulation, considering factors like information protection, resource limitations, and time-based decay. Further research could also explore different feature types and their relative salience in memory-driven capture.

The study's use of a specific stimulus set (isosceles triangles) and task (comparative judgment) limits the generalizability of the findings to other stimuli and tasks. The limited sample sizes in both experiments reduce statistical power, potentially influencing conclusions regarding some comparisons. The focus on visual features may limit the conclusions about auditory or other sensory modalities. Finally, the lack of complete measurement of the effectiveness of the retro-cue in Experiment 2 could be a limitation in interpreting the results.