Cognitive Sciences Stack Exchange is a question and answer site for practitioners, researchers, and students in cognitive science, psychology, neuroscience, and psychiatry. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I have rather technical question regarding STDP dynamics. I am working on a neural network implementing an STDP learning algorithm, and have noticed that it is extremely anti-reciprocal. When two neurons have bidirectional connectivity between them, the conditions that will strengthen one direction will inevitably weaken the other, as far as I can understand. If B consistently fires immediately following a spike in A, the A->B synapse will strengthen, however the B->A synapse would weaken, as the pre-synapse / post-synapse roles are reverse. This wouldn't be much of a problem if it weren't for empirical evidence of high reciprocity in cortical connectivity.

Am I missing something? Is this a known shortcoming of the model?

share|improve this question
Couldn't this problem be solved by having a certain transduction delay (signal transduction, synaptic transmission) in both connections? – H.Muster Jul 19 '12 at 19:13
Yes I just got that answer from a friend in the field as well. It seems to be the solution - aligning the STDP function not on time 0 but rather on a certain -Xms to the left. Do you have any idea what transmission delay is reasonable? – Amir Jul 20 '12 at 10:27
This surely depends on what you intend to model (i.e., cells, which part of the brain). Unfortunately, I do not have a reverence discussing this topic at hand. – H.Muster Jul 20 '12 at 11:26
A synapse does not exist in isolation in the brain. That does limit how correlated the firing rates of two neurons can become. Furthermore, reciprocal connectivity is may be to the same neurons but not the same place on the neuron. For example neuron A could synapse far from B's soma but B could synapse close to A's soma. – mac389 Aug 25 '12 at 0:52
Not all synapses use STDP as a learning rule. Some synapses have normal Hebbian learning. – honi Nov 19 '15 at 2:44

Consider that elimination of such a phenomena is not ideal. It's been proposed that actual neuronal networks exist under the tension of synchronous decoupling.

To answer your question though, you probably should consider that reciprocal connections might not be between two individual neurons. Rather, axons in the cortex are clumped into minicolumns, each one having 20+ pyramidals in supragranular layers. It's conjecture, but it's possible that the reciprocal connections are between different pyramidals within minicolumns (or even different pyramidals within columns or whatever other macro-structure you want).

share|improve this answer
You are exactly right. You should edit your answer to demonstrate more certainty: reciprocity is not a property of single neurons. It's a property of nuclei. But a caveat... this is only for chemical synapses. Electrically coupled (gap junction) neurons can be reciprocal. An another caveat.... retrograde signalers like THC act backwards on chemical synapses... so there is some form of reciprocity at the single neuron level, but it's not part of standard models really. And what they do is modulate the connection... not send electrical signals. – Keegan Keplinger Aug 5 '12 at 6:16
@Xurtio I see. I was equivocating a bit because I couldn't produce a direct source. It's just inference from other things I know of the nervous system. I'll try and find sources and edit the answer to be more certain. – Preece Aug 6 '12 at 4:37

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.