A study involving 17 kindergartners tested association- and schema-based views of simple mental addition. Six children appeared to use mechanical rules: Two each stated one of the addends, added one to an addend, and constructed a teen answer from one of the addends. Five other children appeared to use more genuine estimation strategies. Eight weeks of computational practice affected the errors of unpracticed combinations on a retest. Moreover, 7 of 10 children mastered previously unknown combinations involving zero. This resulted from learning a relationship (adding with zero leaves a number unchanged) rather than from the practice and memorization of individual facts. The results indicate that mental-arithmetic errors, changes in error patterns, and mastering some simple facts cannot be explained entirely as a function of practice.