Young Children’s Performance Decreases in a Probability Learning Task When Tested Online

With online data collection becoming increasingly common in developmental research, it is crucial to validate the comparability of virtual online and offline in-person methods in assessing children’s cognitive processes and behavior. We examined differences in 3- and 4-year-olds’ choice behavior and strategy use in a probability learning task when tested either online via video chat (n = 60) or offline in person (n = 39). Consistent with previous research, children’s performance across the two testing modalities was similar when averaged across trials. However, we found an interaction between testing modality and trial block: Children tested online were less likely to choose the more frequently rewarded option toward the end of the task than were children tested offline. Computational modeling analyses revealed differences in value-based learning across testing modalities: Children who were tested offline chose the higher-valued option more deterministically toward the end of the task than children tested online. In sum, our findings highlight the importance of considering task length and the characteristics of the online testing environment to ensure valid comparisons between online and offline studies in developmental research and add to a growing body of literature highlighting the context-dependency of reinforcement learning parameters.