A model of expertise in physics was tested on a sample of 374 college students in 2 different level physics courses. Structural equation modeling was used to test hypothesized relationships among variables linked to expert performance in physics including strategy use, pictorial representation, categorization skills, and motivation, and these variables were examined for their influence on physics achievement. Gender was included in the model to examine how it influenced achievement indirectly through its influence on the other variables in the model. Two levels of expertise were examined by testing the model on trigonometry-based physics students and on more advanced, calculus-based physics students. Results were similar across both levels of expertise: For both courses, student motivation had a significant influence on students' strategy use and categorization skills. Categorization skills, in turn, influenced student achievement directly, and indirectly, through strategy use. Strategy use had a significant influence on achievement. Pictorial representation played little role in the model. Gender contributed primarily through motivation, but for the more advanced level course it also directly predicted strategy use. (Contains 1 figure and 5 tables.)