Starting from research insights related to inefficient rote learning, the Learning by Imitative or Creative Reasoning design research project studies an approach to teaching based on student’s own creation of knowledge and compares this to the common imitative model of teaching. Mathematical tasks were designed based upon a mathematical reasoning framework (Lithner, 2008) utilizing a specific adidactical practice situation (Brousseau, 1997). Thus, the participating students were engaged either in algorithmic reasoning (AR), or in creative mathematical founded reasoning (CMR), during practice. The research questions were: What are the outcomes when comparing imitative and creative designs? How can these be understood? The sample consisted of 102 students at the Natural science program in Swedish upper secondary school (16-17 year olds). The participants were matched into two groups (CMR & AR) based on a cognitive composite score (Raven matrices & operation span), gender, and mathematics grade. The intervention comprises training tasks related to the target knowledge, i.e. 14 formulas. The training session was computer based and took about half an hour to complete. Except for the presented algorithmic representation in the AR-tasks the training tasks were identical for the two groups. The retention interval was 6-8 days, and the same test was administered to all students. The test tasks were also computer based, the first task evaluated knowledge of a specific formula and then, if needed, a possibility to reconstruct the formula was given in two consecutive tasks. The CMR group significantly outperformed the AR group on all three test tasks. The cognitive composite score was found to be a significant predictor for performance in the AR group but not in the CMR group. The result indicates that CMR performances as a function of CMR practice is unrelated to individual’s cognitive resources. Instead, successful training was found to be the predictive factor for performances in the CMR group. The opposite pattern was seen for the AR group. I addition when removing participants with the highest 33% cognitive composite score the patterned remained. Hence, the differential effects between CMR and AR were not driven by the cognitively stronger students. Our suggestion is that CMR model of teaching is more effective and neutral in terms of basic cognitive resources. References