Working memory can be impervious to value-based attentional bias and trial reward value.

Working memory (WM) encoding is shaped by attention. We attend to rewarding stimuli, and we focus during important moments. Is attentional allocation to high-value stimuli fully under strategic control? Does this change when a trial is important? To explore the contributions of stimulus and trial value, we used the retro-cue paradigm. A valid retro-cue shifted attention after WM encoding. The neutral retro-cue allowed us to assess how intrinsic and reward values guide attention during WM encoding. The valid retro-cue tested whether post-encoding shifts of attention interact with value-dependent encoding. Exp. 1 probed different task-specific reward values, and participants earned more points when a high-value item was probed. To disentangle task-specific from intrinsic value effects, we leveraged a priori value of coins (Exp. 2A-E) and tested whether higher-value items attracted attention, leading to superior performance when those items were probed. Stimulus value was task irrelevant. Finally, we mimicked moment-to-moment fluctuations in importance across trials or blocks (Exp. 3A-B). There was no WM benefit for high-value stimuli or trials. Together, our findings indicate that effects of rewards may be less robust than previously thought; working memory is not automatically attracted to highly valued items and does not easily improve under high stakes.