# Spike Timing Dependent Plasticity interpretation

A paper I'm looking at titled "Synaptic Modifications in Cultured Hippocampal Neurons: Dependence on Spike Timing, Synaptic Strength, and Postsynaptic Cell Type" (can be found here) showed a relationship between spike timing and LTP/LTD. However, I'm having a bit of difficulty understanding what Figure 7 shows:

Since this photo I've seen published in multiple textbooks and reviews, it seems like a very important figure to understand what is going on. According to what they wrote in the introduction:

"Our results showed that postsynaptic spiking that peaked within a time window of 20 msec after synaptic activation resulted in LTP, whereas spiking within a window of 20 msec before synaptic activation led to LTD"

My understanding is that you have two neurons, a presynaptic neuron and a postsynaptic neuron and then based on their relative firing timings, that tells you whether or not there is an increase or decrease in the synaptic weight provided that the timings are close enough together. What confuses me as a non-biologist is how a postsynaptic spike could occur before the presynaptic spike does? I would think the only way a postsynaptic spike occurs is when the presynaptic neuron fires into the postsynaptic one. And then this time interval... is it from when the timing of the last presynaptic spike to the first spike of the postsynaptic one? Or from the first presynaptic spike since you would need multiple of them to induce a spike from the postsynaptic one? I'd mainly like to ask for more clarity over how to interpret what is happening.

The best explanation for STDP that I've ever seen is Nicky Case's Neurotic Neurons. In this Explorable Explanation, I think you'll find that your misunderstanding lies in the fact that both the pre-synaptic (Neuron A) and the post-synaptic neuron (Neuron B) exist in a network of other neurons. They are not only connected to each other. You are correct in saying that it often takes more than one incoming spike to make a neuron fire. These spikes are usually coming from other neurons (Neuron C, D, E... etc). So it's incorrect to think that "the only way a postsynaptic spike occurs is when the presynaptic neuron fires into the postsynaptic one".

So, to summarize, yes the time interval is the time difference between the last pre-synaptic spike to the first spike of the post-synaptic spike. These neurons exist in a network of neurons. It's probably best to think of STDP as a coincidence detector.

• Hi Seanny. Thanks for the great link! So just to check my understanding, if $0$ms is the time that the postsynaptic neuron fires, then any neuron that spiked in the interval $[-20,0]$ (before the postsynaptic spike) will have an increase in the synaptic weight, but anything that fires in the interval $[0,20]$ (i.e. after the postsynaptic spike), then the synaptic weight will lower? So then for one postsynaptic neuron, there is a change in synaptic weights for every presynaptic neuron connected to it that fires when associated with some task? – Brenton Aug 7 '16 at 23:58
• @Brenton that is a correct understanding. However, I would be careful when saying "task", as associating individual neural activities with tasks is a bit difficult. – Seanny123 Aug 8 '16 at 3:04
• One other question. You said " the time interval is the time difference between the last pre-synaptic spike to the first spike of the post-synaptic spike". Let's say the postsynaptic spike is at 0ms again, and I have two presynaptic spikes, one at -15ms and another at -10ms. Then the synaptic weight changes for the synapse for the -10ms neuron, but does it also change for the -15ms? You mentioned that the time interval is the last presynaptic spike, so it sounds like it would only change for the -10ms synapse. I just want to inquire if that's right or if they both would change – Brenton Aug 8 '16 at 18:38
• Both weights would change in different amounts according to STDP. Synapses between different pairs of neurons are not bonded according to STDP. – Seanny123 Aug 9 '16 at 13:10

If you have two unidirectional neurons (ignoring back-propagation), both will elicit an action potential above a certain treshold. As you probably know, action potentials have a certain recovery rate. However, it is the intracellular processes that distinguish LTP from LTD. In LTP, Ca2+ influx is higher than in LTD and activates a protein cascade. LTP activates Ca2+/calmodulin kinase II (CaMKII), while LTD deactivatse CaMKII. Both LTP and LTD are input specificrepeated stimulation at different frequencies will result in different neurotransmitory processes at the synapse, due to different levels of Ca2+ influx, which in turn triggers two different types of protein cascades that have a significant impact on the connectivity as well as the type of memory that the type of plasticity (LTP or LTD) is associated with.

Mostly paraphrased from: Manahan-Vaughan, D. (2010). Hippocampal Long-term Depression as a Declarative Memory Mechanism. In: P. K. Stanton, C. Bramham, & H. E. Scharfman (Eds.). (2010). Synaptic Plasticity and Transsynaptic Signaling, Ch. 18, pp.305-315. NY: Springer.