It is conjectured that some skill entering into problem solving in physics courses might be responsible for making these courses so difficult. This might be the case if one or more basic problem-solving skills could not be mastered in the time and practice available. In order to test this hypothesis, a model of the way in which students solve problems is proposed. The model has six observable steps based on the structure of the solution; each step is assumed to correspond to a basic skill. For a fairly typical group of students in an introductory calculus-based physics course, it was found that all but two of these skills are learned as well as might be expected. The skills not learned at mastery levels are the recognition of laws that apply and the substitution of all definitions into laws. The later skill is learned rapidly and might be improved upon by additional materials, but the former skill is learned too gradually for substantial improvement to come about through practice. Thus the skill of recognizing which laws apply is likely to be a source of great difficulty presenting an unattainable goal of instruction in its mastery in problem solving. (Author/MK)