Unravelling the interplay of Statistical Learning, Top-Down, and Bottom-Up Mechanisms during target selection: Insights from Behavioural and EEG Experiments

Abstract

The natural environment exhibits consistent patterns, rendering it repetitive and partially predictable. Statistical learning (SL) enables us to discern these regularities from past experiences to then direct attention toward relevant elements for our objectives. Yet, it remains unclear whether SL collaborates or acts independently of other experience-independent attentional control mechanisms, specifically top-down and bottom-up control. In a series of interconnected experiments, we recorded behaviour and EEG activity during a visual search task to critically examine their interaction during target selection. In Experiment 1, we assessed the combined influence of top-down control (modulated via endogenous cueing - neutral/valid), bottom-up control (introduced by a salient item), and SL (induced by an imbalance in target probability - high/low - across locations) on behaviour. Additionally, Experiments 2 and 3 investigated how the N2pc, an EEG marker related to target selection, was impacted by the interplay of SL with top-down and bottom-up control, respectively. Simultaneously, SL and bottom-up control enhanced behavioural performance for targets at high (vs. low) frequency locations and for salient (vs. non-salient) targets. Crucially, an interaction revealed that the benefit of top-down control for validly cued targets could override the SL effect (Experiments 1-2). Moreover, EEG results indicated a greater N2pc for validly cued and salient targets, but only when at a low-frequency location. This finding suggests compensation for the lower attentional resources allocated to that location due to SL (Experiment 3). In summary, while top-down control and SL closely interact, the latter appears to be mostly independent from bottom-up signals.