Perhaps the two most fundamental findings of associative learning are that behavior changes when a relation is arranged between two entities and that this behavior returns toward its prior state when this relation is rescinded. This is true both in Pavlovian conditioning, in which one arranges for some stimulus to signal some outcome, and in instrumental learning, in which an organism’s response produces an outcome. Over the years, our understanding of the nature of the associative changes that occur when such relations are arranged has become very detailed and our theories extremely sophisticated. What is surprising is that our understanding of the loss of behavior when these relations are removed is much more primitive. Even more surprising is the fact that the study of the phenomenon of extinction has decreased, rather than increased, over the years. Whereas learning textbooks used to devote whole chapters to it in the 50s, 60s, and 70s, contemporary texts often give it only a few pages of discussion, sometimes confined exclusively to the partial reinforcement extinction effect. Nevertheless, if we are to understand learning processes, it is obvious that we must understand the nature of extinction.