Revisiting Age-Related Changes in Statistical Learning: The Importance of Longitudinal Evidence

In a recent article in Cognitive Science, Rogachev et al. (2025) presented a cross-sectional investigation of visual statistical learning (SL) in children aged 3-9 years and concluded that implicit SL remains stable across early childhood. They cited our longitudinal study (Tóth-Fáber et al., 2024) as supporting this conclusion. Here, we clarify that this interpretation is incorrect. Using a longitudinal design tracking the same individuals from ages 7 to 14, we demonstrated a reliable developmental decline in implicit SL, along with substantial interindividual variability. We further showed that executive functions measured at age 14 predict individual developmental trajectories of SL, indicating a dynamic reorganization of learning systems with maturation. Importantly, tasks used to measure SL inevitably recruit multiple cognitive processes, and differences in these task demands can substantially influence observed developmental trajectories. We argue that longitudinal and cross-sectional designs yield qualitatively different evidence about developmental change. Longitudinal evidence and relatively process-pure measures are, therefore, essential for accurately characterizing developmental dynamics in SL.