Bayesian inference, and Bayesian networks in particular, have become a popular conceptual framework for computational neuroscience. Bayesian networks are made of nodes representing probabilistic variables, and algorithms perform inference on them typically by passing messages between neighboring nodes. Their semantics and computations are thus more complex than MLP and Hopfield style neural networks, and Bayesian network nodes are perhaps more suitably implemented by microcircuits of neurons such as a cortical column rather than by individual cells. We will introduce Bayesian networks through simple inference problems, build and make inferences with a network in Matlab, and look at possibilities for mapping to biology.