Differential effects of probabilistic retro-cues on search guidance and memory maintenance

<p>While some debate has recently focused on whether attentional templates are stored in discrete, binary states, a question remains: to what extent can their priority be flexibly adjusted according to probabilistic information and current task demands? We investigated this flexibility by manipulating retro-cue reliability across five conditions (absent, 50%, 70%, 80% and 100%) in a paradigm combining visual search and continuous memory recall. We tested whether priority allocation operates as an all-or-none switch or varies with cue reliability, and whether a common prioritization function governs attentional guidance and memory precision. Unexpectedly, although cue reliability modulated performance in both tasks, the task analyses revealed different nonlinear profiles. Search performance improved when cue reliability increased from 50% to 70%, with no further significant gains at higher reliability levels. Memory precision benefited from cue presence relative to the absent-cue condition and improved further only when the cue became fully reliable. Analyses including invalid trials showed that cue reliability altered the balance between cued and uncued representations, with high reliability increasing the cost of invalid cueing in memory. These findings are not readily explained by either a strictly all-or-none change in template status or a smooth linear increase in priority, and are instead consistent with flexible prioritization. Probabilistic information appears to influence both the benefits of prioritizing the cued representation and the costs of deprioritizing the uncued representation. These effects may differ depending on whether the representation is used to guide visual search or to support memory recall.</p>