Task cues are quickly updated into working memory as part of their processing: The multiple-cue task-switching paradigm.

Goal-directed behavior requires maintaining the relevant goal in working memory (WM) and using it to guide behavior. The contents of WM should be regulated, so only relevant goals, but not irrelevant ones, are maintained. Computational models suggest that a gate, which is closed by default, separates WM from perceptual input. Transient opening of the gate enables WM updating. Indeed, previous studies show that updating WM with relevant information is controlled, effortful, and slow. In contrast to the above, here we show that WM updating with goal information is faster and more accurate than not updating. A multiple-cue task-switching paradigm is introduced. Participants were presented with a sequence of task cues, followed by a single probe. They needed to respond to each cue using its corresponding key. The cues were presented in red or blue. When the probe appeared, participants had to judge it using the task cued by the most recent red (but not blue) cue. Accordingly, they had to update their WM when the cue appeared in red, but not when it was blue (the color mapping was counterbalanced in Experiment 2). In two experiments, we show that performance in update trials was faster and more accurate than in no-update trials, suggesting that updating, rather than not-updating, is the default mode of operation.