approach. Our model is especially attractive because it starts from economic primitives: a specification of the set of possible individual preference patterns. It shows how specific properties of the distribution of preferences translate into properties of aggregate demand. This allows us to understand the relationship between the distribution of preferences and the degree to which the market introduces biases in product selection. This paper presents a general model of the demand for differentiated products which has as special cases two of the most popular approaches used to analyse welfare and competition in monopolistically competitive markets: the model of spatial competition on the circle and the model of symmetric aggregate demand. The circle model is attractive because both demand and costs have a definite (and easy to visualize) physical foundation and because it is tractable. Salop (1979) provides an exemplary modem analysis. With this model one can analyze the relationship between the optimal amount of product variety and the amount which a monopolistically competitive market will support. In many plausible cases competition produces too much variety. Such a strong conclusion raises questions of robustness. Almost simultaneously, Spence (1976) and Dixit and Stiglitz (1977) developed an alternative approach to monopolistic competition, based upon the Chamberlinian assumption of symmetry (see also Shubik (1959)). Rather than specifying a distribution of preferences and deriving demand through aggregation (as did Salop and his predecessors), these authors started by specifying a symmetric aggregate benefit function indicating how social welfare depends on the amounts of the various products supplied. Both Spence and Dixit-Stiglitz were able to show that the welfare conclusions from the circle model were not robust: depending on the functional form of the benefit function, equilibrium may entail either excessive or insufficient product variety.1 The assumption of symmetric demand has proven fruitful in the analysis of many economic problems. It is therefore natural to ask what distribution of tastes would give