Purchase this article with an account.
Edwin Dalmaijer, Claire Poullias, Rosyl Somai, Masud Husain; When is reward-associated information prioritised in visual working memory?. Journal of Vision 2017;17(10):869. doi: https://doi.org/10.1167/17.10.869.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Visual Working Memory (VWM) is a highly limited resource that can be flexibly distributed over a limited number of items. Storing more items comes at the cost of a lower fidelity for individual items. Given this pressure on VWM resources, we hypothesised that they are biased to be distributed to behaviourally relevant information. In several experiments, we presented participants with a variable number of coloured squares (set size 2 or 4; colours sampled from CIE LAB space), which were all equally likely to be probed. One third of all trials was neutral (no rewards). In the rest of the trials, half the items were associated with a potential reward (indicated by a black or white stimulus outline). If a reward-associated item was probed, participants received a reward that was either incidental or contingent on recollection accuracy (results are similar between these conditions). Responses were collected by the method of adjustment (mouse-click on a continuous colour wheel), and were fitted with a mixture model. Our results indicate that if reward-associated and non-rewarded stimuli are remembered, recall precision is equal for both. However, reward-associated stimuli have a higher chance of being recollected. In addition, recall precision for a set of two neutral stimuli was better than recall precision for two reward-associated stimuli in a set of four. Thus participants' recollection of potentially rewarded information was affected by the presence of non-rewarded information. At the very least, non-rewarded items acted as distractors during VWM encoding. An alternative possibility is that all items are initially encoded, but non-rewarded items are subsequently discarded during maintenance. This would result in the observed equal recall precision for reward-associated and non-rewarded items, as well as a higher guessing rate for non-rewarded items.
Meeting abstract presented at VSS 2017
This PDF is available to Subscribers Only