Children vary widely in their number knowledge by the time they enter kindergarten, and this variation is related to their future academic success. Although talk about number predicts children's early understanding of foundational number concepts, we know little about whether interventions can increase this talk nor about the types of number talk that are most beneficial to children's number understanding. The current project examines whether embedding number talk in goal-based stories leads to more robust number learning than providing the same numeric input outside of this context. Parent-child dyads (N = 71; child age = 3 years) were randomly assigned within their preintervention cardinal number understanding levels to one of three conditions: rich counting books (narrative and pictures involving a number goal), sparse counting books (text and pictures that provide opportunities to count and label the same cardinalities as in the rich condition but with no narrative), or nonnumerical control books that involve labeling colors of objects. Children's number knowledge was measured at pretest and at 2 and 4 weeks into the intervention. Findings showed that children randomized to rich counting book condition showed significantly greater cardinal number knowledge and counting skill by the final testing session compared to children in the sparse counting book condition and the color control book condition, which did not significantly differ from each other. Results have implications for the types of number talk that most effectively support children's number learning in the home environment and for the design of interventions. [This is the online first version of an article published in "Developmental Psychology" (ISSN 0012-1649).]