To learn a verb, children must attend to objects and relations, often within a dynamic scene. Several studies show that comparing varied events linked to a verb helps children learn verbs, but there is also controversy in this area. This study asks whether children benefit from seeing variation across events as they learn a new verb, and uses an eye tracker to test whether children adjust their visual attention to specific objects to better understand how they may be comparing events to each other. Children saw events in which the tool varied, the affected object varied, or there was no variation (control). No prior verb study has tested children's visual attention to specific objects under different variability conditions. We found 2½- and 3½-year-olds could extend verbs, and they were more successful with age. Analyses of the looking patterns in the learning phase show that children's attention to specific objects in events varied by condition, and that reduced looking to the tool was linked to less success at test. Eye tracking can provide a more detailed view of what children attend to while learning a new verb, which should help us better understand how children are learning from variation across examples.