Recent models of brain function have been successful in showing that the emergent properties to higher-level mental processes such as learning and memory can be obtained from the collective behavior of an enormous number of simple processing units. These models, in general called parallel distributed (PDP), or "connectionist" models, are appealing because they seem to capture certain basic features of what kinds of computations brains are good at and these features are obtained by architectures that are at least somewhat consistent with biological reality. This chapter will summarize the appealing aspects of PDP models and will then describe a complex application: the modeling of the characteristics of dream and waking fantasy mentation.