Since 1990 a new genre of research, often described as the 'new economic geography', has emerged. It differs from traditional work in economic geography mainly in adopting a modelling strategy that exploits the same technical tricks that have played such a large role in the 'new trade' and 'new growth' theories; these modelling tricks, while they preclude any claims of generality, do allow the construction of models that - unlike most traditional spatial analysis - are fully general-equilibrium and clearly derive aggregate behaviour from individual maximization. The new work is highly suggestive, particularly in indicating how historical accident can shape economic geography, and how gradual changes in underlying parameters can produce discontinuous change in spatial structure. It also serves the important purpose of placing geographical analysis squarely in the economic mainstream.