I study an example of a coordination game, and examine the robustness of equilibrium predictions with respect to changes in the information structure. I find two main results: First, the critique of Morris and Shin (1998) is not robust in the sense that if perfect common knowledge is viewed as the limit of imperfect information structures, multiple equilibria are maintained, as long as there exists some valuable public information. I also find that in general, the possibility of coordination is more likely to arise when the overall level of noise is low and when the public information is relatively informative. These results can be related to the structure of higher-order uncertainty: With a public signal, higher-order uncertainty vanishes, as the noise in the signals disappears.
