In many environments a monopolist can price discriminate even though consumers can select any option offered by the firm. Though consumers seem indistinguishable to the firm, the monopolist can discriminate by exploiting its knowledge of the joint distribution of agents' characteristics (in this paper, locations and reservation prices) to offer the most profitable menu of alternatives given the constraints imposed by the maximizing behavior of each type of consumer (see Chiang and Spatt [1982] for a partial bibliography of this research). In many of these models consumers agree typically on the definition of the best available good (exceptions include Adams and Yellin [1976] and Telser [1979]) but disagree on their incremental valuation of quality. The consumer's selection of a price-quality contract from the set offered by the monopolist can elicit information that the monopolist endogenously exploits. For example, the price-discriminating monopolist may charge a steeper price-quality gradient than predicted by cost considerations (see Mussa and Rosen 1978). In a spatial market the monopolist can discriminate on the basis of differences in cost of travel or location of consumers. I analyze a model I analyze by geometric techniques a spatial model in which a monopolist uses location differences across consumers to price discriminate. The model is used to examine price discrimination in a model of variety in a setting in which consumers disagree on the definition of quality. The full burden of separation of types is not on binding self-selection conditions. Examples are presented in the paper to show the ambiguous output and welfare comparisons between imperfect discrimination and singleprice monopoly. At various extremes discrimination can either (relative to single-price monopoly) make some consumers better off and none worse off, while raising firm profit, or, alternatively, reduce firm output. The effect of an increase in the cost of travel on output, variety, and consumer surplus is demonstrated. The implications of the model are also shown to be exhaustive under various assumptions about the set of observables. X An earlier version of this paper was presented at the NBER-NSF Conference on "Hierarchies, Monitoring, and the Nature of the Firm," at the University of Minnesota, Minneapolis, December 1980. 1 gratefully acknowledge the helpful comments of the referee and support from National Science Foundation grant SES-8015086.