For complex rocks in which the structure of minerals, rather than their chemical composition, changes with progressive evolution of the system, it makes sense to try and monitor such an evolving system through the progressive change in the crystal structures of the constituent phases. In effect, the paragenetic sequences of minerals in such complex environments should be related to the crystal structures of the constituent minerals. In order to consider variations in structure topology, we need to organize crystal structures into hierarchical schemes, using the hypothesis that structures may be hierarchically ordered according to the polymerization of the coordination polyhedra with higher bond-strengths. Structural units are organized according to the mode of polymerization: unconnected polyhedra, clusters, chains, sheets and frameworks. The bond-valence structure of (OH) and (H2O) shows that on one side, (OH) and H2O are strong Lewis bases; on the other side, they are weak Lewis acids. As a result, a very important role of both (OH) and (H2O) is to prevent polymerization of the structural unit in specific directions. Thus, the dimensionality of the structural unit is controlled primarily by the amount and role of hydrogen in the structure. The way in which we have formulated these ideas also allows development of a predictive framework within which specific aspects of the chemistry and structure of phosphates can be considered. This approach to mineral structure, applied via the idea of a structural unit, can play a major role in developing structural hierarchies in order to bring about some sort of order to the plethora of hydroxyhydrated-phosphate structures. Furthermore, by combining the idea of binary structural representation with bond-valence theory, we see the eventual possibility of predicting stoichiometry and structural characteristics of these minerals, particularly those in complex low-temperature hydrothermal environments.