A large number of curling shots using a wide range of rotational and translational velocities on two different ice surfaces have been recorded and analyzed. The observed curling rock trajectories are described in terms of a semi-phenomenological model. The data are found to rule out `dry friction' models for the observed motion, and to support the idea that the curling rock rides upon a thin liquid film created at the ice surface (i.e. `wet friction'). Evidence is found to support the hypothesis that the frictional force acting upon each segment of the curling rock is directed opposite to the motion relative to this thin liquid film and not relative to the underlying fixed ice surface.