Judgment Patterns in Children’s Non-Symbolic Probabilistic Choices: A Hierarchical Drift-Diffusion Modeling Study

Six- to 10-year-old children’s probabilistic reasoning yields seemingly conflicting results across task conditions. While children often succeed when heuristic cues match probabilistic structure (congruent), they perform at chance when these cues conflict with proportional information (incongruent). This apparent inconsistency reflects a tension between probabilistic and heuristic accounts of reasoning under uncertainty, raising questions about whether judgment patterns remain consistent across conditions and whether they reflect distinct underlying decision-making mechanisms. We studied probabilistic reasoning in 6- to 10-year-old children (N = 92) in Montevideo, Uruguay, using a two-alternative forced-choice task with congruent and incongruent trials, with small or large probability differences. We identified three clusters in the incongruent-large condition: a proportion-based group, a heuristic group relying on target numerosity, and a chance-level group. Distinct sensitivities to probability differences and numerosity indicated consistent judgment patterns across conditions. Diffusion modeling showed that the heuristic group had higher decision efficiency in congruent trials and a pre-decisional bias toward the absolute number of targets, whereas the proportion-based group accumulated evidence more efficiently in incongruent trials and showed unbiased pre-decisional states in this condition. These findings suggest that 6- to 10-year-old children may rely on distinct, internally consistent judgment patterns supported by different decision mechanisms. This pattern suggests that individual differences in decision dynamics may play a role in children’s probabilistic choices. By combining the analysis of choices in probabilistic tasks with the diffusion modeling framework, this study offers a new perspective on probabilistic thinking in children, highlighting individual differences in evidence accumulation and pre-decisional biases.