To increase the connection between educational research and educational practice, a process called "curriculum reformulation" was used to incorporate recent advances in research on learning and instruction into science classroom experiences. The cognitive demands of a thermodynamics curriculum were successively refined while maintaining the same microcomputer based laboratory (MBL) software and the same basic experiments. The 13-week curriculum was reformulated four times, and each version was evaluated using the same criteria. Overall, a four- to ten-fold increase in student learning (depending on the criteria applied) was achieved as a result of reformulations based on cognitive research. The results showed that some suggestions from research offer promise for use in realistic settings and that other suggestions such as "offer multiple representations" are wrong or incomplete when applied in realistic settings. Includes a bibliography of 75 references. (Author/MVL)