|Abstract:||Patterns of inversion, complementizers and adjuncts are different in main and subordinate clauses. Inversion occurs in matrix interrogatives in English, for example, and not in subordinates. The hypothesis proposed is that families of markedness constraints evaluate syntactic structures in particular locations: at the top of matrix clauses, at the top of subordinate clauses and in all projections. Interaction among these universal constraints in language particular rankings generates main versus subordinate syntactic properties. Main and subordinate clauses are governed by the same constraints, relativized to different locations.