Item-based and feature-based selection in working memory

Abstract

The capacity of our brain to maintain visual information in working memory (WM) is limited. One of the ways in which the brain compensates for these limited resources is by employing selective attention. Selective attention serves to prioritise information during perception and also acts on mental representations maintained in WM. So far, studies have shown that retrospectively attending the spatial location of one of the items in WM improves performance on the subsequent recall when asked about a feature of that item. One recent study has shown a performance benefit by selectively attending feature dimensions of items in WM (Niklaus et al., 2017, Scientific reports). However, no study to date has directly compared item-based and feature-based attention nor their neural correlates, and much remains unknown about feature-based attentional cueing. We tested and compared the benefits of retrospective cues (retro-cues) orienting attention to an item location or feature dimension of the memoranda in WM in 30 participants. To chart neural effects, we recorded electroencephalography (EEG). Behavioural results show attentional benefits in both recall accuracy and response onset time following both item and feature cues. We used the EEG data to decode the 4 presented features on every trial and trace sensory and mnemonic representations of each feature following the item-based and feature-based retro-cues. Preliminary data analysis suggests that both item-based and feature-based attention upregulate representations of the task-relevant features compared to task-irrelevant features. The results challenge the strict notion of object binding, where upregulating one feature of an item necessarily upregulates the other feature of the same item in tandem. We argue that WM representations are stored in a goal-oriented format where relevant information – at both the item or the feature-level – can be prioritised in a format that optimises ensuing/anticipated task performance.

Acknowledgement: Wellcome Trust