"Research Findings:" Early science skills predict later science achievement, and persistent achievement gaps in science appear as early as preschool. The current study compared the effectiveness of different instructional approaches for teaching preschoolers about sinking and floating and examined individual differences in learning. Typically developing 4--5-year-olds (N = 93; 47% female) were randomly assigned to an Explicit Instruction, Discovery Learning, or No-Instruction Control condition. A pre-posttest design was used to measure change in children's knowledge of sinking and floating using an assessment created for the current study. Participants also completed measures of executive function and non-verbal and verbal IQ. Children in the Explicit Instruction condition learned more than children in the other two groups. Additionally, individual differences in age, socioeconomic status, IQ, and executive function predicted learning, although only age and SES were uniquely predictive when all variables were included in the model. Lastly, children's sinking and floating pretest scores significantly interacted with instructional condition to predict learning. Among children in the Discovery Learning condition, those with more prior knowledge gained more after instruction than children with less prior knowledge. "Practice or Policy:" These findings highlight the importance of considering how learner characteristics might influence the effectiveness of different instructional approaches.