关键词: 工作记忆, 空间位置, 工作记忆移除

Abstract: Working memory (WM) is a limited-capacity system, its functioning relies heavily on the ability to remove no-longer-relevant information. While previous research has primarily focused on the removal of entire objects, it remains unclear whether all features within an object share a common fate during removal. In particular, given the privileged status of spatial location across various cognitive processes, it is an open question whether location information is treated differently during removal. To answer this question, we directly compared the removal of spatial and non-spatial features from WM. We proposed two competing hypotheses: The location-specific hypothesis assumed that spatial location was more difficult to remove from WM than non-spatial features, due to its central role in object representation. In contrast, the location-non-specific hypothesis assumed that spatial and non-spatial features share similar removal difficulties. To test these hypotheses, we conducted two experiments with 55 participants (N = 26 in Experiment 1; N = 29 in Experiment 2). We employed a newly developed paradigm combining retro-cue with a change-detection task. On each trial, participants memorized a single object defined by multiple features (e.g., color and location). After a 500-ms delay, a retro-cue indicated the task-relevant feature, and participants were explicitly instructed to remove the uncued feature from WM. Following a 1000-ms delay, a probe appeared, and participants judged whether the cued feature had changed. Crucially, the uncued (task-irrelevant) feature changed in 50% of the trials, allowing us to quantify residual interference from information that should have been removed. The critical measure was the interference effect, defined as the difference in reaction time (RTs) between trials with and without a change in the task-irrelevant feature. If removal is successful, changes in the task-irrelevant feature should produce little or no interference. In Experiment 1, we observed greater interference when spatial location was the task-irrelevant feature compared to when a non-spatial feature was irrelevant. This finding supports the location-specific hypothesis. However, because different feature types were tested under separate task conditions, it remained possible that the observed difference reflected task-related confounds (e.g., differing task difficulty). To address this concern, we conducted experiment 2 which tested spatial and non-spatial features within a unified task structure, eliminating between-condition differences. The results replicated those of Experiment 1: changes to task-irrelevant locations produced reliable interference, whereas changes to task-irrelevant non-spatial features did not. Across both experiments, we found consistent evidence that spatial location is more difficult to remove from WM than non-spatial features. This asymmetry was robust to changes in task design and supports the view of spatial location. Taken together, our study supported the privileged status of spatial location in WM removal.

Key words: working memory, spatial location, working memory removal