The reliability and, to a lesser extent, the validity of the newly created Number Sense Test was evaluated with a group of 40 3- to 5-year-old children attending Head Start. Six number sense skills (e.g., counting, number identification, addition-subtraction) and children's feelings about school were assessed both before and after instruction either in 2 of these skills (i.e., counting and number identification) or in insects (i.e., an attention control condition). The Number Sense Test was found to have high internal consistency and to be a valid means of assessing pre- to post-test improvements in instructed skills as a function of condition. Except for addition-subtraction, training effects did not generalize to non-instructed number sense skills. In neither condition did instruction have a harmful effect on children's largely positive feelings about school.