In this study, we randomly assigned 69 Grade 4 students to a writing-to-learn treatment (n = 23), comparison (n = 23), or no-treatment control (n = 23). Treatment and comparison students completed a science experiment involving balance. During the experiment, treatment students wrote four short responses and an extended response to document their learning. To control for writing time, comparison students wrote four short responses and an extended response about topics other than balance. On a 20-item balance knowledge posttest, control outperformed treatment (d = 0.89) and comparison (d = 1.05) on the lowest level balance questions (Level 1). At the highest level questions, Levels 3 and 4, treatment (ds = 1.42 and 0.94) and comparison (ds = 1.62 and 1.37) outperformed control. There were no significant differences in total words written or level of balance understanding on a posttest written response. The performance of individual responders to treatment is also discussed.