Problem solving is a complex process important both in itself and as a tool for learning physics. Currently there is no standard way to measure problem solving that is independent of physics topic, pedagogy, and problem characteristics. At Minnesota we have been developing a rubric to evaluate students' written solutions to physics problems that is easy to use and reasonably valid and reliable. The rubric identifies five general problem-solving processes and defines the criteria to attain a score in each: useful description, physics approach, specific application of physics, math procedures, and logical progression. An important test of the instrument is to check whether these categories as represented in students' written solutions correspond to processes students engage in during problem solving. Eight problem-solving interviews were conducted with students enrolled in an introductory university physics course to compare what students write down during problem solving with what they say they were thinking about as determined by their interview statements.