This article has a dual purpose. First, it describes the main econometric specifications which can be used to represent spatial heterogeneity, reflected inan instability ofparameters inspace and/or a heteroscedasticity oferror terms. Only the specifications valid in cross-section are examined. Second, it explains the links between spatial heterogeneity and autocorrelation, the other major feature of localised data, defined by the absence of independence between geographical observations. In particular, we look at the extent to which traditional tests of heteroscedasticity or instability need to be amended to take account of spatial autocorrelation. An application designed to illustrate the main models and the different tests is also proposed.