For 8 to 12 weeks, 180 students in Grades 3, 5, 7, and 8 used calculators (C) or no calculators (N) for problem solving. A parallel study in which all 116 students, Grades 5, 6, and 7, used calculators for problem solving was conducted simultaneously. Posttesting with and without calculators produced few differences on the number of problems attempted and the number with correct operations. When using calculators on posttests, C groups obtained significantly more correct answers to problems than N groups in 7 out of 12 comparisons. On paper-and-pencil tests of computation and problem solving, C groups performed at least as well as N groups.