Protein domain movements are of critical importance for understanding macromolecular function, but little is understood about how they are controlled, their energetics, and how to characterize them into meaningful descriptions for the purpose of understanding their relation to function. Here we have developed new methods for this purpose based on changes in residue contacts between domains. The main tool used is the “Dynamic Contact Graph” which in one static graph depicts changes in contacts between residues from the domains. The power of this method is twofold: first the graphs allow one to use the algorithms of graph theory in the analysis of domain movements, and second they provide a visual metaphor for the movements they depict. Using this method it was possible to classify 1822 domain movements from the “Non-Redundant Database of Protein Domain Movements” into sixteen different classes by decomposing the graphs for each individual protein into four elemental graphs which represent the four types of elemental contact change. For each individual domain movement the output of this process provides the numbers of occurrences of each type of elemental contact change. These were used as input for logistic regression to create a predictor of hinge and shear using assignments for these two mechanisms at the "Database of Macromolecular Movements". This predictor was applied to the 1822 domain movements to give a tenfold increase in the number of examples classified as hinge and shear. Using this dataset it was shown that contrary to common interpretation there is no difference between hinge and shear domain movements. The new data is presented online with new websites which give visual depictions of the protein domain movements.