Recent research on early word learning suggests that children's behavior when-generalizing novel nouns integrates their prior vocabulary knowledge with the specifics of the task. This study examines how these factors interact on the moment-to-moment time scale of the training children receive and the sequence of stimuli they are shown. In 1 condition, we used a combination of training and stimulus factors predicted to produce a bias to generalize nouns by shape similarity. We then reduced this shape bias and amplified a bias to generalize nouns by material similarity via manipulations of training and stimuli across 3 other conditions, Additional analyses suggest that children's generalizations on individual trials are influenced by what they have seen and done on previous trials. These results highlight the importance of the task and stimuli in bringing children's prior knowledge to bear in early word learning.