Learning about a moving visual stimulus was examined in zebrafish larvae using an automated imaging system and a t1-t2 design. In three experiments, zebrafish larvae were exposed to one of two inputs at t1 (either a gray bouncing disk or an identical but stationary disk) followed by a common test at t2 (the gray bouncing disk). Using 7days post-fertilization (dpf) larvae and 12 stimulus exposures, Experiment 1 established that these different treatments produced differential responding to the moving disk during testing. Larvae familiar with the moving test stimulus were significantly less likely to be still in its presence than larvae that had been exposed to the identical but stationary stimulus. Experiment 2 confirmed this result in 7dpf larvae and extended the finding to 5 and 6dpf larvae. Experiment 3 found differential responding to the moving test stimulus with 4 or 8 stimulus exposures but not with just one exposure in 7dpf larvae. These results provide evidence for learning in very young zebrafish larvae. The merits and challenges of the t1-t2 framework to study learning are discussed.