Abstract
As an individual's goals change, they must flexibly shift the focus of attention. In real-world scenarios, multiple stimuli, each with different likelihoods, can signal that it is appropriate to shift or to hold attention on a moment-by-moment basis. In the current study, we independently manipulated the likelihood of shifting attention and the likelihood of receiving particular cue stimuli by presenting multiple shift- and hold-attention cues to dissociate the behavioral costs associated with violations of shift readiness from stimulus identity predictions. After excluding trials with exact cue stimulus repetitions-where response times were significantly shorter than for other trials-we observed additive updating costs for shift and stimulus identity likelihood prediction errors across two experiments. Together, our results suggest that when cue stimuli do not consecutively repeat, the processes of updating attention-shifting readiness and stimulus identity are best explained by a serial processing architecture. The data, materials, and analysis code for this experiment are available online ( https://osf.io/zdtvc/ ).