The relation between S and P station anomalies is classically assumed to be linear. We reconsider this relation in the light of recent surface and body wave studies concerning the U.S., which suggest lateral variations of more than one independent physical parameter, such as thickness of a layer, in the first 200 to 300 km of the mantle. This implies breaking up the travel time data into provinces within which only one parameter can be assumed to vary. We apply this to U.S. data from ISC bulletins for the period 1964--73, and find that the new correlations are more easily interpreted in terms of known physical processes in the upper mantle.