Fredo <fredo from hotmail.com> wrote: I've been reading a bit about neurophysiology and neurobiology. A number of
texts refer to backpropagation in real neurons, where the signal
backpropagates, though significantly attenuated, up the dendrites.
I can't seem to find out much detail on the matter, beyond the fact that it
exists. In particular, what is the significance? The backpropagation might
affect gap junction-connected neurons, but it can't back-propagate through
synapses, can it? If it can, what's the mechanism? Surely not
neurotransmitter release in the dendrites...
Back-propagation is the natural result of injecting any ionic current into a neuron. These ions will spread out in both directions. In neurons 3 ionic currents do this: Sodium ions, Potassium ions, and Calcium ions. Sodium ions are the "normal" neural charge, Potassium ions reverse the Sodium effects and thus are considered an inhibitory response, Calcium ions are involved in neural modulation and adaptability.
As others have mentioned the Sodium back-propagation can be actively amplified (the NMDA receptors) and modulated further via dendritic micro-circuits. Yet you asked about the significance of all this. The answer is that the interactions of these back-propagating currents determine the response characteristics of the neuron which in turn is governed by the purpose of the neuron its local circuit which is in turn governed by the behavioral needs of the animal. By response characteristics I mean control over latency, time horizon, burstiness, frequency, connective type (more Sum-like or OR-like) etc. for a given set of input types,
If you are really interested in how these back-propagating currents interact and want to play with their various control parameters I have a demo brain circuit simulation program for windows computers at my site (softstatemagic.com) where you can create your own neuron or use a neuron in an example brain circuit that you can download.
And in an answer to Stephen Wolstenholme I hope you can see that the backpropagation technique used in artificial neuron networks is nothing like the neural backpropagation in question here.
Brain Circuit Simulation Resources: http://www.softstatemagic.com