Studies in mathematics education often point to the necessity for students to engage in more cognitively demanding activities than just solving tasks by applying given solution methods. Lithner’s (2008) framework on mathematical reasoning address this by studying which reasoning a task promotes. Previous studies have shown that students that create their own solution methods, denoted creative reasoning (CMR), perform significantly better in follow up tests than students that are given the solution method and engage in algorithmic reasoning (AR). However, teachers and textbooks at least occasionally provide explanations together with a solution method (XAR) and this could possibly be more efficient than creative reasoning. In this study three matched groups practiced with either AR-, XAR- or CMR-tasks. The study showed that students that practiced with AR- and XAR-tasks performed similarly during both practice and test. The CMR-group did, although a worse practice score, outperform the XAR- group in the test. Additionally, there were differences in which variables predicted the test result with similar results between the XAR- and AR-group where cognitive proficiency and mathematics grade where significant. For the CMR-group the test score was predicted by the practice score alone. This would indicate that students that practice with CMR-tasks are not as dependent on their cognitive abilities for test performance as students that practice with AR- tasks.