Inter-item distortions in visual working memory

Abstract

Current models of working memory (WM) assume that mnemonic representations of individual items are stored independently of one another. Recent studies, however, have indicated that WM representations interact, leading to systematic distortions along their feature dimension. In a series of experiments we found that several aspects of the encoding situation influence whether this interaction leads to attraction or repulsion between two retained dot motion directions. Items consistently repulsed each other when presented simultaneously during encoding. When presented sequentially, however, inter-item interactions depended strongly on an item's serial position within and between trials as well as on the spatial and temporal characteristics of the item presentation. First, similar to the serial dependence effect of perception we found that an item memorized on the preceding trial attracted its successor in the current trial. In contrast, when both items belonged to the same trial, the preceding item (S1) repulsed the successor (S2). Next, we investigated in more detail the distortion effect on S1 and S2, separately, by manipulating the their spatial position and inter-stimulus interval (ISI). Interestingly, we found opposing distortion patterns for the S1 and S2. Specifically, when items were presented at different instead of the same retinal position: S2 attracted S1, whereas no systematic bias in responses to S2 was observed. However, when both items were presented in immediate succession instead of being separated by different ISIs, S2 was again repulsed by S1, despite different positions. Our results provide new evidence for systematic inter-item distortions in WM. Interestingly, the observed distortions varied considerably between different encoding conditions and serial positions within and between trials, hinting at different origins of the distortion effects. These findings may demand an extension of the current WM models. Meeting abstract presented at VSS 2016