The first paper develops a simple rational expectations model incorporating storage and trade to explain the high frequency behaviour of prices in two different centres. The model, which consists of a set of inequality constrained stochastic integral equations describing the actions of rational, risk neutral arbitrageurs, is solved numerically. The solution provides several new insights into the manner in which physical arbitrage and storage determine prices in different centres. In addition, the solution suggests that models of spatial price determination estimated using high frequency data should be formulated using non-linear conditional moments rather than the currently dominant cointegration framework. In Part II of the paper, the empirical framework is used to analyse the operation of the New York dollar-sterling market during the gold standard era, 1890 to 1901. The model provides considerable insight into the workings of the dollar sterling market, and in turn the data support many of the model's predictions about the manner in which storage and physical arbitrage jointly determine commodity prices in different centres.