Magnocellular oxytocin (OT) neurons in the supraoptic nucleus of the hypothalamus project to the posterior pituitary, where they secrete the hormone into the blood stream. OT is essential for breastfeeding and important for parturition. In these cases spike activity is organised into intense bursts[1]. OT has also anorexigenic effects and is involved in osmotic regulation, showing then a continuous increase in spike activity rather than a bursting pattern[1]. In OT cells, spikes are generated in response to afferent inputs that produce fluctuations in membrane potential. That spontaneous spiking activity can be closely matched by models which assume that this afferent input arrives randomly[2]. When spikes propagate to the axon terminals in the posterior pituitary, they trigger exocytosis of hormone containing vesicles. However, OT secretion is a non-linear function of firing rate: the same number of spikes provokes much more secretion if they are close together[3]. It is this feature that makes bursts such an important feature of the milk-ejection reflex. Because of this non-linearity, variance in firing rate that results from the randomly fluctuating synaptic inputs would be expected to be amplified to produce a more variable secretion. Thus there seems to be a conflict: the non-linearity of stimulus-secretion coupling that makes bursts so effective at releasing large pulses of OT during reflex milk ejection will also make secretion very variable in response to a “steady” input. Here we present a mathematical model that replicates the behaviour of spike production and axonal secretion in single OT neurones. The secretion model, based on an existing vasopressin secretion model[4], was matched to in vitro[3][5] experimental data, and accurately reproduces the non-linearity of stimulus-secretion coupling in the OT terminals. The spiking model accurately reproduces the spiking behaviour of OT cells in a wide range of experimental circumstances. Combining the spiking and secretion models allows us to study the responses of the OT system to a fixed transient challenge, mimicking the excitatory response to systemic injection of the gut peptide cholecystokinin (CCK), and in particular, it allows us to study how the response magnitude is affected by factors that affect the basal firing rate of OT neurones. We show that a key feature of the electrophysiological phenotype of OT neurones – their expression of a slow afterhyperpolarisation (AHP) – is a critically important determinant of the variability of the plasma OT concentration that results from secretion. The AHP moderates the variability of spike activity in OT neurones, with a resulting substantial impact on the variability of secretion. The AHP “smooths” the mean firing rate of OT cells over a time scale of a few seconds, avoiding extreme excursions that would result in large fluctuations in secretion.