Hande Melis Altunay, Laura Ziemann, Nora Swaboda and Azzurra Ruggeri
Affiliation: Technical University of Munich, Max Planck Institute for Human Development, Technical University of Munich, University of Potsdam, Technical University of Munich, Central European University
Category: Psychology
Keywords: Adaptiveness, Ecological active learning, Exploration, Information-search, Modal reasoning, Preschoolers
Date: Wednesday 3rd of September
Time: 17:30
Location: Gen. Henryk Dąbrowski Hall (006)
View the full session: Modal Cognition & Problem Solving
Humans are not passive learners—they actively shape their exploration to match the demands of their environment, a skill known as ecological active learning (Ruggeri, 2022). While preschoolers and toddlers often struggle with efficient information search due to cognitive limitations (Leahy, 2020) or task demands (Ruggeri, 2022), research suggests that by age three, they can adjust their question-asking and exploratory actions based on task characteristics (Ruggeri et al., 2019). Recent findings indicate that even by 18 months, toddlers can adapt their strategies in implicit tasks (Poli et al., 2024). This raises the question of whether these early adaptive abilities can be explicitly measured, building on research demonstrating rudimentary information-search competence in infants (Téglás et al., 2007; Denison & Xu, 2010; Stahl & Feigenson, 2015}. To investigate, we conducted two experiments adapting the paradigm from Ruggeri et al., 2019) to examine ecological active learning in 2- and 3-year-olds. Experiment 1 Participants were 156 24- to 42-month-old children (M=33.7 months; SD=5.25 months; 74 female) recruited from kindergartens, family centers, and museums from [blind for review]. We constructed two marble-run apparatuses with transparent tubes, creating four exits. In the Uniform condition, balls could exit from any point, while in the Skewed condition, cylindrical blocks left only one open path. Children completed Search and Catch tasks. In the Search task, four small boxes were aligned with the exits, each with a rear opening, front window, and transparent covering. After learning two actions—opening a box or lifting a lid—children observed five balls landing in boxes under either condition. At test, they searched for a red ball performing Hypothesis-Scanning (i.e., opening one box) or Constraint-Seeking (i.e., lifting the lid) option. In the Catch task, children first observed five balls exiting the marble run and we marked the exits. They were then introduced to small and large boxes and chose one to catch the red ball. A Chi-squared test revealed that, overall, children in the Uniform condition were more likely to perform the CS action (61%) compared to the Skewed condition (45%), χ²(1) = 5.62, p=0.018. A mixed-effects logistic regression showed a significant three-way interaction among condition (Uniform vs. Skewed), age group (younger: 24–33 months vs. older: 33.1–42.9 months, age median split), and task (Search vs. Catch), β=-2.27, SE=1.02, z=-2.21, p=0.027. Specifically, in the Search task but not in the Catch task, a higher proportion of older children performed the CS action in the Uniform (64%) compared to the Skewed condition (37%), χ²(1, 77) = 4.68, p=0.030. We found the opposite pattern for younger children: a higher proportion of them performed the CS action in the Uniform (76%) compared to the Skewed condition (49%), χ²(1, 68) =4.23, p=0.04) in the Catch task, but not in the Search task. Interim Discussion Although the Search and Catch tasks are structurally similar, they differ in one key aspect: the Catch task requires prediction (anticipating where the ball will fall), whereas the Search task requires postdiction (inferring where the ball has already fallen). Older children’s success in the Search task is unsurprising, as by age three, they adapt their search strategies (Ruggeri et al., 2019). Similarly, it is expected that two-year-olds struggled with the explicit search measure, despite evidence suggesting that they seek information adaptively in implicit tasks (Poli et al., 2024). Three-year-olds’ failure in the Catch task is consistent with research on probabilistic and modal reasoning (Mody & Carey, 2016; Redshaw & Suddendorf, 2016), which shows that young children struggle to consider multiple possible outcomes simultaneously. Surprisingly, younger -but not older- children succeeded in the Catch task One possibility is that older children were less motivated by the task scenarios. To test this, we introduced slight modifications to the design in Experiment 2. Experiment 2 Participants were 22 2-year-old children (M=31.68 months, SD=2.82 months, 10 females) and 19 3-year-old children (M=43.63 months, SD=3.00 months, 10 females) recruited from kindergartens. Experiment 2 focused exclusively on the Catch task and was nearly identical to Experiment 1, with two key modifications. First, we placed the boxes farther apart, requiring children to actively move toward their choice. Second, we changed the goal from catching a red ball to saving a fragile glass ball, reinforcing this by using a transparent ball, cotton-lined boxes, and an elevated marble run during the test. A Chi-squared test revealed that, overall, children in the Uniform condition were more likely to perform the CS action (68%) compared to the Skewed condition (44%), χ²(1)=3.66, p=0.05. A logistic mixed-effects model predicting children's box choices (small vs large) with condition (Skewed vs Uniform) and age in months as fixed effects and participants' ID as a random effect revealed main effects of condition (OR=0.34, 95\% CI [0.13, 0.92], β=-1.07$, SE=0.49, z=-2.18, p=0.03) and age (OR=0.90, 95\% CI [0.84, 0.97], β=-0.10, SE=0.04, z=-2.67, p=0.008, AIC=107.37). Children were more likely to perform the HS action in the Skewed condition compared to the Uniform condition, and 3-year-olds were more likely than 2-year-olds to perform the CS action. Adding an interaction effect between condition and age did not improve model fit (AIC=109.52). Discussion This study examined whether younger preschoolers adapt their exploratory actions to the task at hand. In Experiment 1, even 2-year-olds adapted their strategies when given an age-appropriate task. Experiment 2 replicated this result. While 3-year-olds did not adjust their choices based on statistical structure, they consistently preferred the big box, which was equally effective. One can speculate that unlike younger children, for whom box choices may have differed in cost, 3-year-olds faced no such trade-off and opted for the CS option. Choosing the large box in the Uniform condition and the small box in the Skewed condition suggests they can consider multiple possibilities simultaneously. Otherwise, they would have mostly chosen the small box regardless of the condition, as covering one exit aka accounting for one possible outcome would seem sufficient. This result aligns with recent studies suggesting earlier development of modal reasoning abilities (Alderete & Xu, 2023; Turan-Küçük & Kibbe, 2024; Turan-Küçük & Kibbe, 2025}, challenging previous research on modal reasoning.