Preschoolers made numerical comparisons between sets with varying degrees of shared surface similarity. When surface similarity was pitted against numerical equivalence (i.e., crossmapping), children made fewer number matches than when surface similarity was neutral (i.e, all sets contained the same objects). Only children who understood the number words for the target sets performed above chance in the crossmapping condition. These findings are consistent with previous research on children's non-numerical comparisons (e.g., [Rattermann, M. J., & Gentner, D. (1998). "The effect of language on similarity: The use of relational labels improves young children's performance in a mapping task." In K. Holyoak, D. Gentner, & B. Kokinov (Eds.), "Advances in analogy research: Integration of theory and data from cognitive, computational, and neural sciences" (pp. 274-282). Sofia: New Bulgarian University; Smith, L. B. (1993). "The concept of same." In H. W. Reese (Ed.), "Advances in child development and behavior, Vol. 24" (pp. 215-252). New York: Academic Press]) and suggest that the same mechanisms may underlie numerical development.