This paper presents experimental results of applying extended Kalman-Bucy filtering and Bayesian hypothesis selection to estimate motion, tire forces, and road coefficient of friction (μ) of vehicles on asphalt surfaces. The filter estimates motion and tire forces of an eight degree-of-freedom vehicle from vehicle-mounted sensors without requiring a priori knowledge of μ and without requiring a tire force model. Resulting force estimates are compared statistically with those that result from a nominal analytic tire model to select the most likely μ from a set of hypothesized values. The methods have application to off-line construction of tire models, verification of vehicle dynamic models, and development of vehicle control systems that require road friction.