Purchase this article with an account.
Emma Wu Dowd, Eren Gunseli, Martijn Meeter, Christian Olivers, Tobias Egner; Changes in task-irrelevant context invoke updating of task-relevant representations in working memory. Journal of Vision 2016;16(12):351. doi: https://doi.org/10.1167/16.12.351.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Working memory (WM) involves the active maintenance of information necessary for ongoing tasks, such that task-relevant changes in the environment trigger the dynamic updating of representations within WM. Long-term memory is strongly affected by context (i.e., aspects of the environment that are not directly task-relevant), but how is WM impacted by context changes? We hypothesized that changes in task-irrelevant context are automatically encoded into WM, resulting in WM updating for task-relevant memory representations. To examine behavioral and neural effects of task-irrelevant changes on WM updating, we collected fMRI data during a delayed matching task. Participants remembered a tilted Gabor patch across a 10s delay for a two-alternative-forced-choice memory test. These target patches were superimposed on a face or house image, serving as task-irrelevant context (only during cue display, not at test). Critically, we independently varied repetitions and changes of both targets and context cues, resulting in factors of task-relevant (target-new, target-old) and task-irrelevant (context-new, context-old) changes. Multivariate pattern analyses were used to track active neural representations of targets, to test whether context changes triggered WM updating for the target representations. Neural representations of both targets and context cues were successfully decoded from ventral visual and temporal cortex during the delay. When context did not change, behavioral memory performance improved with increasing target repetitions, whereas neural decoding accuracy decreased, suggesting that learning of the memory target resulted in less active maintenance. Context repetition did not affect memory performance, but classification accuracy for the target increased for context changes. Importantly, when context changes occurred late in the target repetition sequence (i.e., target-old/context-new), memory performance dropped while decoding strength peaked, consistent with re-activation of the repeated target representation in WM. These findings indicate that not only are context changes automatically encoded, but this task-irrelevant updating also re-invokes WM updating for the task-relevant target.
Meeting abstract presented at VSS 2016
This PDF is available to Subscribers Only