Most proteins can interact with other proteins to form complexes. Is the dynamic assembly of a protein complex connected to the intrinsic flexibility of its component subunits? Can the order of this assembly process be related to evolutionary changes in quaternary structure? Two methodological advances have allowed us to address these questions. First, we introduced a simple method for predicting the free-state flexibility of protein subunits from the structures of protein complexes based on accessible surface area calculations. Second, we showed that we can predict assembly pathways through analysis of intersubunit interface areas, and validated this method by experimentally characterizing the assembly of several heteromers with macromolecular mass spectrometry. using these new methods, we demonstrate that the intrinsic flexibility of protein subunits is crucial for assembly, as it facilitates the conformational changes that enable both the packing together of multiple distinct subunits and the formation of asymmetric interfaces required for cyclic symmetries. We also find that protein complex subunits have a strong tendency to assemble in an order corresponding to their flexibility and magnitude of conformational change, thus suggesting a possible mechanism for control of ordered assembly. Moreover, we reveal that changes in quaternary structure driven by evolutionary gene fusion events tend to occur in such a way so that they mimic the existing order of subunit assembly. This shows that there is selective pressure in the evolution of protein complexes to conserve assembly pathways. Finally, we observe that evolutionarily more recent subunits tend to be more flexible than more ancient subunits within a complex. Together, these results reveal the influence of both evolution and intrinsic structural dynamics on protein assembly and demonstrate, for the first time on a genome-wide scale, the biological importance of ordered assembly pathways.