I will consider Armstrong's problems in trying to account for structural universals,i.e., a kind of complex universal whose instantiation by particulars involves differentparts of those particulars instantiating several basic properties and relations, such asthe property of being a molecule of methane. I present and criticise Armstrong's mostrecent attempt to explain structural properties by means of the identification of universals with types of states of affairs and I state my own solution to the problem by appealing to formal relations holding between particulars.