Abstract
This essay argues that monetary theories should not contain an undefined object labeled money. Among existing theories that do not satisfy that dictum are models which assume that real balances are arguments of utility or production functions and models which assume cash-in-advance constraints. A main weakness of theories that do not satisfy the dictum is that they cannot address questions about which objects constitute money. Theories that do satisfy the dictum are those which specify assets by their physical properties and which permit the assets’ role in exchange to be endogenous. The essay briefly describes one such theory, a random matching model with assets that differ according to whether they throw off real dividends.