Fit analysis is widely performed in item response theory (IRT) based test development to assess the fit of individual items to the IRT model being used. The paper explores a step fit analysis procedure that is an extension of IRT-based item fit diagnostics applied to the response categories present in popular performance-based tasks. The step fit procedure involves computing category fit statistics and constructing category fit plots. Category fit statistics are used to flag possible misfits at the response category level, and category fit plots are used to facilitate investigation of potential fit problems by displaying the magnitude and pattern of deviations for each step of a polytomous item. A step-by-step description of the procedure is presented, with findings from a simulation study and from a real data application involving 8 to 11 tasks for 5 content areas at 11 grade levels, with 400 to 1,000 examinees (students) at each grade level. Results suggest that the procedure is promising for examining the statistical characteristics at the response category level and for diagnosing potential problems in item content at the category level, while clarifying and improving the scoring rubric. Three tables and six figures present study results. (Contains 4 references.) (SLD)