Background: Despite doubts voiced on their efficacy, a series of studies has been carried out on the capacity of training programmes to improve academic and reasoning skills by focusing on underlying cognitive abilities and working memory in particular. No systematic efforts have been made, however, to test training programmes that involve both general and specific underlying abilities. If effective, these programmes could help to increase students' motivation and competence. Aims: This study examined the feasibility of improving problem-solving skills in school children by means of a training programme that addresses general and specific abilities involved in problem solving, focusing on metacognition and working memory. Participants: The project involved a sample of 135 primary school children attending eight classes in the third, fourth, and fifth grades (age range 8-10 years). Method: The classes were assigned to two groups, one attending the training programme in the first 3 months of the study (Training Group 1) and the other serving as a waiting-list control group (Training Group 2). In the second phase of the study, the role of the two groups was reversed, with Training Group 2 attending the training instead of Training Group 1. Results: The training programme led to improvements in both metacognitive and working memory tasks, with positive-related effects on the ability to solve problems. The gains seen in Training Group 1 were also maintained at the second post-test (after 3 months). Conclusions: Specific activities focusing on metacognition and working memory may contribute to modifying arithmetical problem-solving performance in primary school children.