This eye tracking study examines the learning benefits of two common active learning approaches — generating predictions and retrieval — for young children. Both generating predictions and retrieval practice are active learning approaches that involve generating responses and then being provided with the correct information or retrieving previously provided correct information. Participants included 90 children (mean age: 7 years; Female = 46, Male = 42). Parents reported children’s race and ethnicity as follows: 2% Asian/Pacific Islander, 5% African American, 74% Caucasian, 3% other, and 6% identified as two or more categories; demographics largely reflective of the county where the data was collected, but nevertheless the generalizability of these findings to more diverse populations may be limited. In this study, young children learned facts about insects (e.g., “insects are hard on the outside”) while we measured their attention to the lesson using eye tracking technology. Then their knowledge was assessed on an immediate test. All children were presented with the same materials but the presentation order was modified based on condition assignment. In the generating predictions condition, children saw examples of animals and were asked if an animal was an insect or they saw animals and were asked to identify which one was the insect, followed by the correct response. In the retrieve condition, the presentation order was reversed such that children first saw the correct response and then were asked if the animal was an insect or which of two examples was an insect. Results suggest that although retrieval practice results in overall better learning outcomes, generating predictions increased children's attention to the materials (d = 1.92), and among children who were able to maintain attention, learning outcomes were equal among the two conditions.