Pre-college students were randomly assigned to learn about electrical circuit analysis with an instructional program that included two problem solving practice conditions. In the first condition, students learned to solve parallel circuit problems that were contextualized around real electrical devices and represented with realistic diagrams. In the second condition, students learned to solve the same problems except that they were de-contextualized and represented with abstract diagrams. To measure learning, students were given near and far transfer problem solving tests. In addition, students' learning perceptions were measured using a program-rating survey that included three subscales: overall program usefulness, problem representation usefulness, and perceived cognitive load. Students who learned with abstract problems produced higher scores on the near transfer test and made better problem representations during problem solving than those who learned with contextualized problems. The contextualized group gave marginally higher ratings for the program representation usefulness. The findings suggest that abstract electrical circuit representations promote better learning because they facilitate thinking about a variety of problem contexts.