The orbital decay of a satellite moving within the dark halo of a larger galaxy is studied. Although the decay results from a complex resonant interaction between the satellite and halo 'stars', it is often modeled by a simple application of the classical dynamical friction formula. Unfortunately, detailed numerical studies of the decay have disagreed about the level of description required to model it correctly, and thus about the physical nature of the dominant interaction. Here, a series of experiments which pinpoint the origin of these disagreements are presented. The present results confirm previous claims that although the response of the halo to the satellite is global in nature, its effect on the orbit is well represented by the local formula. The form of the response does not appear to be sensitive to the inclusion of its self-gravity.