I want to make a point here that we know less about the brain than is generally acknowledged. Our picture of the functioning of a neuron is taken as more or less settled knowledge; only small refinements are likely. But the refinements that are regularly published are not small. Now we have a paper (citation below) that is extraordinary.
Bywalez and others have shown that the little spines on the dendrite trees of neurons can themselves act as miniature neurons accomplishing computations similar to a full neuron (at least in the olfactory bulb part of the brain and probably other parts too) and that some synapses can be two sided, transmitting signals in both directions. This allows dendrite to dendrite communication. In effect the neck of the spine can isolate the spine from the rest of the neuron, allowing it to reach an action potential level of voltage in its area without interference from the rest of the dendrite tree, and so it is able to send a signal backwards out of the spine.
We are used to thinking of neurons as, in effect, huge add-gates that take a multitude of synapses giving inputs of various strengths and those inputs are combined in the dendrites into a voltage level in the main cell body. If that voltage is above a threshold, an action potential voltage, a signal, is propagated down the neuron’s axon to the dendrites other, usually distant, neurons. There it influences how those other neurons act by contributing a positive or negative voltage to the receiving dendrites’ totals. It is fairly easy to imagine how this works and to mimic it with electronic circuits.
But neuroscience keeps finding exceptions to this theory. There are glial cells assisting and interfering with the process and they can communicate with each other by a different mechanism. There are signals that bypass the whole dendrite calculation and input their signal at the cell body root of the axon, thereby over-riding other inputs. There are axon to axon synapses. Neurons can multitask by calculating and then sending two separate message codes to two separate groups of receiving neurons. Signals can go backwards up the axon. Some neurons can learn timing delays in their signaling. And now this: action potentials can be generated in the little spines of the dendrites and some synapses are not one way transmitters with pre and post halves, but can work both ways. The standard model is getting tattered with exceptions. No doubt there are many more exceptions to come. I venture that we are nowhere near understanding neurons and neuron network behavior.
Bywalez, W., Patirniche, D., Rupprecht, V., Stemmler, M., Herz, A., Pálfi, D., Rózsa, B., & Egger, V. (2015). Local Postsynaptic Voltage-Gated Sodium Channel Activation in Dendritic Spines of Olfactory Bulb Granule Cells Neuron DOI: 10.1016/j.neuron.2014.12.051