An earlier presentation in this series discussed feedforward connectionist networks. The current talk extends the discussion to recurrent networks, that is, networks which have feedback connections. In part, recurrent networks extend the computational power of feedforward networks by allowing the network to encode previous context. We will give some examples of recurrent networks and then describe a method for training a recurrent network by `unrolling' it into a feedforward network. We will present simple equations for training feedforward networks and show how they apply to recurrent networks.