Abstract / Summary
Matrix-reasoning assessments commonly use abstract geometric figures, although familiar content may change both performance and processing. This within-participants study compared accuracy and gaze-defined patterns of visual exploration for abstract and contextual items requiring corresponding relational rules. Participants were 56 children and adolescents recruited from a public school in São Paulo, Brazil, aged 11 to 15 years. Each participant completed six abstract and six contextual 3 × 3 matrices while gaze was recorded. Trial-level transition matrices were constructed from fixation sequences across nine matrix cells and the response alternatives, and k-means clustering identified two recurring patterns. Contextual items were answered more accurately than abstract items, with 2.44 times higher odds of a correct response. Completion time, fixation count, and fixation duration were similar between conditions. Cluster 1 was characterized by localized transitions involving the final row and response alternatives, whereas Cluster 2 involved broader, more sequential exploration across the matrix. Contextual items had lower odds of Cluster 1 classification, indicating that representational format modestly altered the distribution of gaze patterns. A significant condition-by-cluster interaction further showed that Cluster 2 was associated with higher accuracy for abstract items, but no cluster difference was detected for contextual items. These exploratory findings suggest that familiar content changes how matrices are explored, indicating that abstract and familiar matrix items should not be treated as interchangeable.