An essential feature of a modern patenting system is a classification schema for organizing, indexing and coding the technical information contained in a patent. Patent classification systems make it possible for patent examiners and prospective inventors to search through existing patents in order to find information pertinent to evaluating a patent application’s purported novelty. Patent classification systems also support the construction of a taxonomy for the various sources of inventive novelty embodied in patented inventions. Until 2013 the U.S. Patent Office utilized the United States Patent Classification system and since then it has used the Cooperative Patent Classification (CPC) system; these two systems implement very different classification logics with the CPC aiming at greater granularity. Here we examine the extent to which the two patent classification systems generate similar historical narratives as to the sources of inventive novelty. Despite the differences in classification principles, common patterns are revealed regardless of which classification system is used to identify technologies. Invention is primarily a cumulative process where new inventions are developed from combining existing technologies. Refinements (the re-use of existing technologies) and combinations of previously existing technological functionalities predominate in the patent record, while inventions embodying previously unseen technologies are very rare. The rate at which inventions representing non-refinements have been introduced into the stock of inventions has kept pace with the generation of inventions representing refinements, thereby feeding the combinatorial process.