An account in which classification depends on stored examples and their similarity to a new case. An unseen central case can resemble many stored examples and receive strong aggregate support, so prototype-like behavior need not imply an explicitly stored average.