My goal in the present treatment is to sketch and compare two scientific approaches to understanding the mind. The first approach, that of classical cognitivism, depicts mind as a manipulator of chunky, quite high-level, symbols. The second approach, that of connectionism (Artificial Neural Networks, Parallel Distributed Processing) depicts mind as a product of the complex interactions between multiple so-called sub-symbolic elements. I shall try to clarify this contrast by associating classical cognitivism with the development of what I shall call semantically transparent systems, and connectionism with the deliberate eschewal of this strategy. Connectionism, I then argue, represents a subtle twist on the standard philosophical view of mental states as functional states. For it suggests a kind of microfunctionalism in which the inner roles do not map neatly onto roles determined by our everyday, contentful, purposive characterizations of the mental. (For the reader unfamiliar with some of these terms, such as functionalism, sub-symbolic, etc. don't worry: these will be explained as we go along).