Similarity, Not Complexity, Determines Visual Working Memory Performance

Research output: Contribution to journalArticlepeer-review

Electronic versions

  • M.C. Jackson
  • D.E. Linden
  • M.V. Roberts
  • N. Kriegeskorte
  • C. Haenschel
A number of studies have shown that visual working memory (WM) is poorer for complex versus simple items, traditionally accounted for by higher information load placing greater demands on encoding and storage capacity limits. Other research suggests that it may not be complexity that determines WM performance per se, but rather increased perceptual similarity between complex items as a result of a large amount of overlapping information. Increased similarity is thought to lead to greater comparison errors between items encoded into WM and the test item(s) presented at retrieval. However, previous studies have used different object categories to manipulate complexity and similarity, raising questions as to whether these effects are simply due to cross-category differences. For the first time, here the relationship between complexity and similarity in WM using the same stimulus category (abstract polygons) are investigated. The authors used a delayed discrimination task to measure WM for 1–4 complex versus simple simultaneously presented items and manipulated the similarity between the single test item at retrieval and the sample items at encoding. WM was poorer for complex than simple items only when the test item was similar to 1 of the encoding items, and not when it was dissimilar or identical. The results provide clear support for reinterpretation of the complexity effect in WM as a similarity effect and highlight the importance of the retrieval stage in governing WM performance. The authors discuss how these findings can be reconciled with current models of WM capacity limits.
Original languageEnglish
Pages (from-to)1884-1892
JournalJournal of Experimental Psychology: Learning, Memory, and Cognition
Volume41
Issue number6
DOIs
Publication statusPublished - 25 May 2015
View graph of relations