Some remarks on a spike train model of interacting neurons