Fig. 1: Input gating manipulation and strategy use.

a Task schematic. Subjects completed a single-trial WM task comprised of three sequentially presented and reorderable stimuli: two ‘item’ stimuli (a letter and a symbol) and a ‘context’ stimulus (a number) that specified which item was relevant. When the context appeared first (CF; top), it could be used to drive input gating of only the relevant item into WM. When the context appeared last (CL; bottom), it could only be used for selectively output gating the relevant item. Each stimulus was on screen for 500 ms, followed by a randomly jittered inter-stimulus fixation between 250 and 800 ms. The trial concluded with response mappings, to which subjects had to indicate (by left or right button press) where the relevant item appeared. Inset: Correct responses by context. Numbers acted as higher-order context, specifying which of the lower-level items (two possible letters and two possible symbols) was relevant on each trial. CF, context first; CL, context last. b Superior performance accuracy (i.e., fewer errors) on CF compared to CL trials demonstrates that subjects tended to use an input gating strategy when possible (F(1,20) = 6.83, p = 0.017). Data are represented as individual datapoints, and condition probability densities and medians calculated across subjects (n = 11 biologically independent samples). Boxplots present the medians and interquartile ranges, and whiskers the 1.5*IQR from the quartile. Source data are provided as a Source Data file. *p < 0.05. c RT did not differ significantly between CF and CL trials (n = 11 biologically independent samples), same conventions as b.