A central statement of the theory of natural trading partners is that preferential trading with regional trading partners is less likely to be trade diverting and therefore geographically proximate partners are to be considered "natural" partners for preferential arrangements. This paper examines this question empirically. The analytical framework involves a general equilibrium model of preferential trade and an econometric model with tight links to this theory. This framework is used to implement tests of the natural trading partners hypothesis using U.S. trade data for the years 196495: Welfare changes that would result from preferential tariff reductions by the United States against various trading partners are first estimated, and correlations with bilateral "distance" measures (with and without controls for income levels) are then examined. Since the argument for "natural" trading partners is based on the greater likelihood of geographically proximate countries to be more significant trading partners, correlations between the welfare change estimates and bilateral trade volume are examined as well. Both geographic proximity and trade volume are found to have no effect. Thus this paper is unable to find any support for the natural trading partners theory in U.S. data.