Neurons of the ventromedial nucleus (VMN) are essential for energy homeostasis. We have made extensive in vivo recordings from these neurons to study how they respond to appetite related signals, including from leptin, ghrelin and cholecystokinin (CCK)¹. These neurons display a variety of intriguing activity patterns. Analysis of interspike interval data reveals varied features amongst different subpopulations, such as doublets, long refractory periods, and, of particular interest, a distinct 3Hz oscillation in excitability². We have been using modelling in an attempt to determine whether these are more likely based on intrinsic electrophysiological properties, or are a result of network interactions. Our first attempt, using a network of random synaptically connected model neurons, with long duration hyperpolarising after-potential (HAP), could reproduce the oscillation, but not completely match the recorded neurons, which do not show the long refractory period necessary to generate the rhythm. The model was also not robust, with the quality of the match dependent on the random connections. A new model solves this by using a network of two cell types, one with a short HAP and the other with a long HAP. The long HAP cells synchronise to produce a 3Hz rhythm. This forms a network input signal to the short HAP cells, producing firing that closely, and consistently, matches recorded oscillatory cells. The firing patterns in the non-oscillatory cells producing the rhythm also match other types identified in the recorded cells. Disrupting the rhythm by reducing the slow HAP’s duration produces a very similar change in firing pattern to experimental intervention with CCK. Thus we can replicate many of the observed features of the VMN using a relatively simple network model. We are now using this model to relate spiking properties to physiological function, by testing their responses to simulations of the in vivo signalling.