Abstract: Collaborative learning, valued as a 21st-century skill, enables individuals to learn from each other and solve complicated problems together. However, it can impose challenges on children who are still developing skills for coordinating shared representations with others, leading to inefficient or ineffective collaborative learning. The current study is a secondary analysis of data from Young’s (2016) study comparing younger (5-6 years) and older (7-8 years) children’s category learning across three betweensubjects conditions (individual, unconstrained-collaborative, constrained-collaborative). Children played a novel board game in which they sampled exemplars from a continuous twodimensional category space in order to learn the categorical preferences of a puppet (e.g., liking big, not small toys). Analyses of children’s category boundary learning showed that collaboration was beneficial only for older children. We are currently modeling additional data on children’s sampling and verbal communication over time to differentiate dyads that achieved effective collaboration from those who did not.