These are leaky integrate-and-fire (LIF) neurons. The neuron tuning properties have been selected so there is one ‘on’ and one ‘off’ neuron.
Here we have a simulation of 2 neurons and Nengo. This is much like the simulation of a single neuron. Just to remind you quickly, this is the input. Here we have the subthreshold voltage of the 2 neurons. These are the spike trains of those 2 neurons shown over time. Here we have those neurons shown as if they were on a cortical sheet. And last we have the postsynaptic current that would be induced in a cell that received the signals from these 2 neurons. Although, more appropriately this can now begin to be thought of as a decoding of the input. In other words, it's supposed to be an estimate of about -.45. This estimate currently is not very good but it gets much better as we include more neurons. What is of interest here is to note that the neurons respond very differently to the input. As I increase the input, the top neuron begins to fire more, but the bottom begins to fire less. So you can see that the neurons are encoding information about the input in a complementary, push-pull fashion. This is not an uncommon feature found in cortex. You can also see that the decoded input is following the input. Although it is doing so in a fairly noisy manner. In the neural engineering framework we refer to the preferences of the cells as their encoders. So we would say that the top neuron has a positive encoder because as the input goes up it is more active, and the bottom cell has a negative encoder because as the input goes down it becomes more active.