Effects of guided (GI) and direct instruction (DI) in solving subtraction problems for mathematically low performers in regular schools were compared. In the GI condition, self-development of solution procedures was encouraged whereas in the DI condition one prescribed strategy was to be used. Forty children (M age =9.3 years) were trained during 34 lessons. Only girls performed better and showed higher comfort in the GI than in the DI condition. Results are explained in terms of girls' increased motivation in the GI condition, which could be explained by assuming that a more constructivist setting is more conducive to girls' learning.