It is easy to describe the components of connectionist models; they are made of nodes (or units) and connections between them. The nodes are the basic processing elements, and each is associated with an activation level, represented by a number. The connections between nodes allow them to inﬂuence each other, so that their activations interact. Each connection is associated with a weight, also represented by a number. These two sets of numbers, activation levels and weights, are found in all connectionist networks, and in many cases they are the only numbers we need to be concerned with. This fundamental simplicity and uniformity of network models is of some importance to modellers, and is intended to correspond, to a ﬁrst approximation, to the basic principles of neural architecture and processing outlined above.