Exploiting exogenous natural variation to study the impact of hard-to-randomize policies based on instrumental variables (IVs) is a widespread research design in economics. The key identification assumption underlying this design is the exclusion restriction, requiring the IV to affect the outcome variable only through the instrumented treatment variable. We review the literature using topography as an IV to show that systematic violations of this assumption are likely because of topography's ubiquity in socio-economic relationships. Furthermore, as topography often lacks first-stage strength, even subtle violations of the exclusion restriction undermine any causal inference. Instead of the vindication that often accompanies IV applications, we advocate a falsificationist approach to the use of topographic IVs, grounded in precaution and skepticism. We apply this approach to a seminal example of a topographic IV, Dinkelman (2011).