How does one overcome the limitations of a distributed associative memory? One approach would be to string chains of such systems together -- making some of the units hidden. However, this won't work -- a chain of these linear memories is no more powerful than a single link in the chain. To deal with this, we must introduce the ability to make decisions -- to incorporate nonlinearity into our processing units. This second building block is the topic of this week's lecture. For our activity after the lecture we will install the Rosenblatt software which brings a perceptron to life and we will investigate using this software on a simple musical problem.