The role of attention in challenging the constraints of spatial statistical learning

<p><strong>Context and relevance.</strong> Statistical learning is ability to extract and learn regularities from the sensory environment. If these regularities concern with the spatial arrangement of objects in relation to each other we say about visual spatial statistical learning. When we have some hierarchical structure as a pattern, information about the whole set or about some of its subsets can be learned. For example, only pairwise connections between elements can be learned. In a row of experiments on statistical learning it has been demonstrated that when global information about a set of elements is learned, information about subsets is blocked, and vice versa - learning of embedded structure is accompanied by blocking of information about the whole set. <strong>Objective.</strong> We were interested in the situation where attention is directed to some elements of a complex stimuli more often than to others. The experiment examined whether, in this case, global information about the set of elements would be retained, or only about the subset to which attention was more frequently directed. <strong>Methods and materials.</strong> Subjects (N=104) performed a search task for the target element of complex stimuli. All the stimuli were composed by the same rule. The frequency of directing attention to one or another element of the stimuli was varied by special instructions. Just after the learning phase subjects completed a series of two-alternative forced choice tests with new correct and incorrect complete and incomplete stimuli. <strong>Results.</strong> We found that correct complete stimuli and correct subsets of more attended elements were appeared more familiar than stimuli with disrupted patterns. <strong>Conclusions. </strong>We demonstrate in our experiment the possibility of learning spatial information about both the global pattern and its substructure. The results are discussed in terms of two types of statistical learning, attention-dependent (explicit) and attention-independent (implicit).</p>