Understanding and describing the physical capabilities of users with motor impairments is a significant challenge for accessibility researchers and system designers alike. Current practice is to use descriptors such as medical diagnoses to represent a person's physical capabilities. This solution is not adequate due to similarities in functional capabilities between diagnoses as well as differences in capabilities within a diagnosis. An alternative is user self-reporting or observation by another person, but these solutions can be problematic because they rely on individual interpretations of capabilities and may introduce unwanted bias. The current research focuses on defining an objective, quantifiable, repeatable, and efficient methodology for assessing an individual's physical capabilities in relation to use of information technologies. Thirty-one users with a range of physical capabilities participated in the evaluation of the proposed performance-based functional assessment methodology. Building on the current standard for such assessments, multiple observers provided independent assessments that served as the gold standard for comparison. Promising metrics produced through the performance-based assessment were identified through comparisons with these observer evaluations. Predictive models were then generated via regression and correlation analysis. The models were validated using a three-fold validation process. Results from this initial research are encouraging, with the resulting models explaining up to 92% of the variance in user capabilities. Directions for future research are discussed.