This study investigated three factors associated with third grade children's ability to solve open sentences of the types N + a = b, a + N = b, a - N = b, and N - a = b, where a and b are given whole numbers and N is to be found. Two factors were: (1) the size of the numbers a and b; and (2) the context of the sentence, whether presented alone or as a step in the solution of a verbal problem. A 16-item test was constructed by completely crossing sentence type with these two factors, and administered to a random sample of 16 boys and 16 girls drawn from three representative elementary schools. A latin square procedure was used to control any item sequence effects, and every child received the same 16 items in a different order. Response scores, analyzed by MANOVA, showed that: (1) there was no significant sex effect; (2) the sentence type a - N = b was significantly more difficult than the other types, which did not differ significantly; (3) sentences with one-digit numbers were significantly easier than those with two digit numbers; and (4) the verbal problems used in the test were not more difficult than the corresponding abstract number sentences. (MM)