Scarf (1960) proposed a market environment and a model of dynamic adjustment in which the standard tatonnement price adjustment process orbits around, rather than converges to, the competitive equilibrium. Hirota (1981) characterized the price paths by the configuration of endowments. We explore the predictions of Scarf's model in a nontatonnement experimental double auction. We find that the average transaction prices in each period do follow the path predicted by the Scarf and Hirota models. When the model predicts prices will converge to the competitive equilibrium, our data converge; when the model predicts prices will orbit our data orbit the equilibrium, and in the direction predicted by the model. Moreover, we observe a weak tendency for prices within a period to follow the path predicted by the model.