Well-specified problems of the type presented boxed in the introduction to this article are extremely common in science courses. Unfortunately, this does not mean that students find them easy to solve, even when a teacher provides model answers to problems which differ only marginally (in the teacher's eyes) from those put before the students. The central difficulty with such courses is that they do not embody instructional principles that reflect students' need for “direction” in problem solving. In this article, we describe how the necessary heuristics and strategic knowledge were built into the remake of a conventional thermodynamics course. In contrast to mainstream American work on learning problem solving we chose to direct our curriculum reconstruction using the Gal'perin theory of stage-by-stage formation of mental actions and Landa's description of the “through” systematization of knowledge. As indicated by both, we first developed an integrated system of instructional objectives: a programme of actions and methods (PAM) to solve problems in thermodynamics. Then the plan of instruction was designed. This plan indicates which instructional procedures and materials should be used to realize the instructional functions, derived from the learning theory. The evaluation design contained two control and three experimental courses. In discussing our main findings, we consider the generalizability of the procedures we followed in constructing the PAM and the instructional plan.