Learning and memory are so obviously related that it is hard to see how the understanding of one could proceed without an understanding of the other. Yet, in psychology, they are studied by two different research communities. The concept of association, which is central both to the field of conditioning and to that of retrieval and forgetting, could be used to bridge the gap between the two concepts. However, the concept is quite different in the fields of learning and memory, a situation for which this article argues that the Rescorla-Wagner model is mainly to blame. By viewing Pavlovian conditioning as the outcome of a predictive process but using the traditional associative language developed in memory studies to describe this process, it has introduced an unnecessary confusion between memory and prediction within the field of learning. This confusion needs to be acknowledged so that the concepts of associations and predictions can again be differentiated. This would allow for better integration of the fields of learning and memory.