It is now generally accepted that magnocellular neurons of the supraoptic and paraventricular nuclei release the neuropeptides oxytocin and vasopressin from their dendrites. Peptide release from their axon terminals in the posterior pituitary and dendrites differ in dynamics suggesting that they may be independently regulated. The dendritic release of peptide within the supraoptic nucleus (SON) is an important part of its physiological function since the local peptides can regulate the electrical activity of magnocellular neurons (MCNs) which possess receptors for these peptides. This direct postsynaptic action would affect the output of peptide in the neurohypophysis. Another way that these peptides can regulate MCN activity would be to modulate afferent inputs unto themselves. Although the influence of afferent inputs (inhibitory and excitatory) on SON magnocellular neuron physiology has been extensively described in the last decade, a role for these locally released peptides on synaptic physiology of this nucleus has been difficult to show until recently, partly because of the difficulty of performing stable synaptic recordings from these cells in suitable preparations that permit extensive examination. We recently showed that under appropriate conditions, oxytocin acts as a retrograde transmitter in the SON. Oxytocin, released from the dendrites of MCNs, decreased evoked excitatory synaptic transmission by inhibiting glutamate release from the presynaptic terminals. It modulated voltage-dependent calcium channels, mainly N-type and to a lesser extent P/Q-type channels, located on glutamatergic terminals. Although evidence is less conclusive, it is possible that vasopressin has similar actions to reduce excitatory transmission. This synaptic depressant effect of oxytocin and/or vasopressin, released from dendrites, would ensure that MCNs regulate afferent input unto themselves using their own firing rate as a gauge. Alternatively, it may only be a subset of afferent terminals that are sensitive to these peptides, thereby providing a means for the MCNs to selectively filter their afferent inputs. Indeed its specificity is partly proven by our observation that oxytocin does not affect spontaneous glutamate release, or GABA release from inhibitory terminals (Brussaard et al., 1996). Thus, the dendrites of MCNs of the supraoptic nucleus serve a dual role as both recipients of afferent input and regulators of the magnitude of afferent input, allowing them to directly participate in the shaping of their output. This adds to a rapidly growing body of evidence in support of the concept of a two-way communication between presynaptic terminals and postsynaptic dendrites, and shows the potential of this nucleus as a model to study such form of synaptic transmission.