A domain-specific information processing model of problem solving behavior in elementary electrostatics was developed. Instruction based on this model was evaluated in a large university-level physics course. The experimental design utilized treatment and instructor as crossed factors. Dependent variables included a test of problem solving strategy comprehension, a test of problem solving skill, and a problem solving protocol score. Significant treatment group differences were found only for the comprehension test. Parallels between the content of the experimental treatment and (1) an instructor effect in the skill measure, and (2) an ability effect in the protocol measure, support the validity of the model developed. (Author)