The Modifiable Areal Unit Problem (MAUP) denotes the fact that the results of many statistical analyses of data for geographic areas are sensitive to the particular areas used in the analysis, especially their number and boundaries. This paper shows that the MAUP is due to the failure to incorporate area effects into the statistical model underpinning the statistical analysis. Once this is done the MAUP can be explained in terms of various factors that lead to area level effects. It is shown how scale effects can be explained in terms of the influence of unobserved area level variables and other factors inducing correlations between individuals within the same geographic area. Inclusion of these effects in the statistical model enables researchers to clearly specify what they are trying to estimate, that is, the targets of inference, and develop appropriate methods of analysis. We believe that efforts to explain and resolve the MAUP should concentrate on developing statistical models that incorporate individual and area effects and the associated analysis and estimation methods. It is argued that geographic analysis should concentrate on the relationships between variables at the area level, after removing the effects of individual level relationships. © 1996 OPA (Overseas Publishers Association) Amsterdam B. V. Published in The Netherlands under license by Gordon and Breach Science Publishers SA.