The role of working memory in probabilistic cuing of visual search.

Abstract

In a typical probabilistic cuing experiment, participants are asked to find a visual target among a series of distractors. Although participants are not informed about this, the target appears more frequently in one region of the display, resulting in faster search times for targets located in this region. This bias is thought to depend on a habit-like attentional control mechanism, unconstrained by the availability of working memory (WM) resources. However, the only study that has explored this feature in the past suffers from methodological shortcomings that leave the results open to alternative explanations. In three experiments, we aimed to confirm whether probabilistic cuing is unaffected by visual, spatial, and spatiotemporal WM load. For each experiment, one group of participants performed the visual search task during the retention interval of a WM task (high-load group), whereas another group of participants (no-load group) carried out the visual search task after the WM task. We hypothesized that the probabilistic cuing effect would be larger for the no-load group compared to the high-load group. This hypothesis was confirmed in one experiment, but exploratory analyses suggest that the results can be highly dependent on the analytic approach, casting doubts on its robustness. Overall, our results provide partial support for the hypothesis that probabilistic cuing is not affected by a secondary task. However, given that some analyses reveal an effect of WM load, we conclude that it might be premature to rule out the possibility that the expression of this attentional bias requires WM resources.

Rethinking Attentional Habits

“…In the case of contextual cuing, we have not found any decrease in attentional habit under conditions that overload WM (Vicente-Conesa et al, 2022). In contrast, we did find a slight decrease in probabilistic cuing in one of three experiments (Giménez-Fernández et al, 2023).…”

“…In this regard, Chen et al (2019) have shown that taxing the executive component of WM by asking participants to manipulate information abolishes the acquisition and the expression of contextual cuing. Regarding the modality of the information, visuospatial information can partially disrupt contextual and probabilistic cuing (Giménez-Fernández et al, 2023; Travis et al, 2013). Thus, some of the latest results suggest that it might be premature to infer that these biases are completely independent of WM resources.…”

The Effects of Load on Spatial Statistical Learning

“…Importantly, the current results are clear that this type of SL is less resistant to parallel demands than previously suggested, challenging the aspect of automaticity (Kim et al, 2009;Sherman et al, 2020;Turk-Browne et al, 2005;see, Vadillo et al, 2016) Implicit (statistical) learning in the visual domain is a broad term, encompassing various types of tasks with different operational definitions. Several studies found that some implicit learning tasks are robust to load or dual-task situations (Garrido et al, 2016;Giménez-Fernández et al, 2022;Vickery et al, 2010;Won & Jiang, 2015). Our results demonstrated that load easily disrupted spatial SL and in that they resemble the deteriorating effects of load on explicit spatial learning tasks (Barrouillet et al, 2007;Vergauwe et al, 2009;Won & Jiang, 2015).…”

“…Importantly, in their search task, the target location was either cued explicitly (by spatial or verbal cues), or implicitly (by a high probability that the target will appear at a certain quadrant of the display). They found that regardless of the type of memory task -if it created a memory load over the search task -it impaired only explicit cues, but not implicit cues acquired through incidental learning (see Giménez-Fernández et al, 2022, for similar results).…”

Mapping the reliability multiverse of contextual cuing

“…All these approaches are entirely valid, a priori. In addition, because search times in the task usually improve with practice, some researchers estimate the decay rate in each condition (e.g., fitting a power function) and then analyze this parameter instead of raw or transformed search times (e.g., Chun & Jiang, 2003;Giménez-Fernández et al, 2023). Likewise, most experiments take some measure to detect and remove potential outliers, but the methods used to select valid trials can be more or less stringent (Morís-Fernández & Vadillo, 2020;Ratcliff, 1993).…”