Human memory, like all memory systems, has both a means of information storage and, its counterpart, a means of information retrieval. Human memory is content addressable. An input-signal prompt begins a search of the entire memory system. Such a prompt might be an incoming visual image of an unknown object. Searching involves looking for a memory folder or folders containing information related to that prompt. If the search does make a hit in a particular folder, that folder’s contents are then delivered to the memory output as retrieved information. A folder's contents might typically be sensory images such as visual, auditory, tactile, vestibular, olfactory, etc. all recorded simultaneously. Information is stored long-term at a molecular level, and is not stored in the memory’s neural networks. The neural networks’ function is to facilitate retrieval, not storage. A cognitive memory performs pattern recognition as well as data storage. Presence of an unknown object prompts the memory to connect the unknown object to previously encountered objects that have been identified. As we are all living with a cognitive memory, learning how to model that memory’s functions can suggest how to design an artificial cognitive memory for computers that is a separate memory, complementary to a conventional computer memory.