This paper reports on a number of inconsistencies that appear in space syntax—a well-known technique of urban analysis—when dealing with certain geometrical configurations. At a simple level, the analysis of regularly gridded urban textures (such as Manhattan's) reveals the difficulty of accepting the claim that space syntax allows the modelling of pedestrian choice making. In more complex cases, the distortion of two ideal textures produces a topological discontinuity, leading to the unacceptable situation where one single urban configuration produces two conflicting outcomes when analysed with space syntax tools. Several other points are also discussed, such as the difficulty of space syntax to take into account building height and land use, and its sensitivity to boundary conditions. Conclusions seem to suggest that the topological representation of cities, on which space syntax is based, discards precious metric information and is rather limiting. It is envisaged that with current increases in computational power new algorithms might allow a deeper understanding of urban texture, based on the full exploration of its metric and topological properties. This would contribute to answer the fascinating question which space syntax has helped to frame: what is the influence of urban configuration on social life?