Talk:Spike-timing-dependent plasticity

Possible error at 'History' paragraph ?
"Henry Markram used dual patch clamping techniques to repetitively activate pre-synaptic neurons 10 milliseconds before and after the post-synaptic target neurons, and found the strength of the synapse increased".

As later paragraph indicates, 10 milliseconds stimulation before caused strengthen of the synapse whereas 10 milliseconds stimulation after caused weakening of the synapse strength.

Have
Have you seen the article cited on the page? One by Senn, Markram, Tsodyks? Are you sure you know what you are talking about? Out of the three only Markram is wet biologist. The other two are predominantly computer people. --Gene s 09:26, 31 Dec 2004 (UTC) Also look at the title of the journal - Neural Computation. That's NOT what most people would call biology. I believe the bio-stub is completely wrong, and certainly Psychology attribution is wrong. It's cognitive science, computer science, AI. --Gene s 09:30, 31 Dec 2004 (UTC)
 * Gene-s: STDP is a real phenomenon studied by neuroscientists from a variety of labs in a variety of preparations, and is most definitely a biological process. Because of its surprising properties it has also been embraced by computational neuroscientists and incorporated into various models for learning in neural networks.  In contrast to many other fields of biology, in neuroscience the boundaries between "wet" and "non-wet" science are often blurred and closely feed off of each other.  This may have added to your confusion.  Nrets 15:52, 29 July 2005 (UTC)

Maybe add this references:

Debanne D, Shulz DE, Fregnac Y (1995) Temporal constraints in associative synaptic plasticity in hippocampus and neocortex. Can J Physiol Pharmacol, 73(9):1295-311. --Rmarseille 12:11, 29 January 2006 (UTC)

No need, I think just the Markram et al Science paper should put all this to rest. Sakmann is more physiology than most. And really Markram is switching to mostly modeling now. --sluox

Please compare my 1993 invention of the Frustrated Synapse Learning Rule. --David Wallace Croft 12:35, 12 September 2006 (UTC)

Hello, I was wondering if anyone would mind a complete rewrite of this entire article. Unfortunately, the Markram et al 1997 paper did not even coin the term spike timing dependent plasticity. Also, they were looking at an artificially induced, current injected postsynaptic action potential that they postulated was a backpropagating action potential (bAP) that would essentially interact with the excitatory postsynaptic potential (EPSP). There was a timing relationship associated with this interaction of within 100 ms in either direction to get potentiation or depression. If the EPSP preceded the postsynaptic AP by 10 ms (and certainly fewer than 100 ms), then there would be a strengthening of the synapse as measured by the relative amplitude of the EPSP in subsequent trials. Likewise, a depression would occur if the EPSP followed the postsynaptic AP within the same time window. They also noted some features like the 10 Hz stimulation frequency that seemed ideal for this behavior to occur. Thus Markram et al 1997 was one of the first to do these experiments (though see Levy & Steward 1983), but Bi & Poo 1998 and Song et al 2000 are also important experiments. (I think it was Song et al. 2000 that first used the term STDP.) Also see Sjostrom et al (2002), Dan et al 2004, Froemke et al 2005, Izhikevich 2003 and 2006, and for some interesting recent papers, Haas et al 2006 and Zhou 2005.

In fact, Markram et al 1997 certainly raised many more questions than it answered, and Lisman & Spruston have opened up the critical analysis even further, though I don't consider their commentary in NN that is cited in the article as a major reference.

Regardless, as it stands the article is confusing and difficult to read. I would be happy to contribute, but I don't want to step on anyone's "digits." Sirshane13

Sakmann's introduction of new microscrope technology to patch clamping led to ALL of these discoveries that follow closely on the heels of Markram et al's Science paper. And, it raised the basic issue that when pre-synaptic activation is closely followed by postsynaptic activation, there is synaptic strengthening. If the order is reversed, synaptic weakening. There certainly has been a lot of further work on the topic since then, but if Henry Markram had not published it first, I doubt the rest would exist. And if Sakmann had not seen a GI physiologist using infrared microscopy, a decade of discovery in patch clamping would likely not exist. The fact that neither Sakmann nor Markram have become big players in the current investigations of STDP does not weaken the impact of the mid 90s work in Sakmann's lab. --Animalresearcher 17:54, 24 January 2007 (UTC)

Plagirism
the entire introductory paragraph has been copied from answers.com, which I don't have a problem with, except that it is not very easy to read. I included the section indicating its differences from hebbian theory, but I am not too experienced with this topic. Paskari 18:54, 30 November 2006 (UTC)

This is not a case of plagiarism, since Answers.com takes this from Wikipedia, and cites its source. sirshane13
 * my bad then since it isn't plagarised, however, I still think it's difficult to read. I will try to clear it up

Kostasl: Hello, I am writing a report related to STPD and I have to say that Markram is not in there, it is considered that Zhang et al. 1998 launched the area by pointing out the critical window. Debanne 1994 linked LTP to LTD which was very important aswell. Quoting myself: "The connection between LTP (Long term potentiation) and LTD (Long term depot..)was uknown until a later historically important experiment revealed the connection between LTP-LTD and pre-post spike interactions which was conducted by Debanne et al. in 1994 \cite{debanne1994apa}. It showed that the sign of the change in synaptic strength is a function of the relative timing of pre- and postsynaptic action potentials. Detailed studies of this timing dependence using single pre-postsynaptic spikes soon followed that led to spike timing dependent plasticity and they will be introduced after we look at a model that attempted to capture the LTP-LTD curve observations." The model is Bienenstock-Cooper-Munro. Then follows Zhang on critical window. Computetional Neuroscience is very much related to the biological aspect of things and it is not strictly AI. The experimental observations are used to develop biologicaly feasible models to derive how the real bio-system work. The biological and computation side of neuroscience go hand in hand. --62.6.161.161 02:29, 20 May 2007 (UTC)

some additional comments
Obviously this is a very important article, and while a good start it needs improvement. There's a good, but of course much longer and more technical article at Scholarpedia.

In the section "Mechanisms" it's correctly stated that the Ca signal in largest, ON AVERAGE, when the pre-spike suitably precedes the post-spike. However the best results on this (papers by Koester and Sakmann and Nevian and Sakmann) also show that this "supralinear" signal is only about 50-70% bigger (on average) than the signal expected from the linear sum of the pre=alone and post-alone cases. The actual signals, on a trial-to-trial basis, are rather variable, because of various stochastic effects, and so it's not true that there is always a supralinear signal, and never true that it's "large", as the article says in the next sentence. Presumably what's happening is that trial to trial fluctuations are smoothed out by the integrative action of CaMKinase, which acts as a temporary register of the individual calcium pulses. The spirit of the section is correct, but it's a little bit too sunny: this issue is the crux not only of STDP but probably the whole of learning and intelligence, and it's rather misleading to give the impression that it's been worked out.

The "History" section is not bad, in some ways better than the Scholarpedia article, which mentions the early Levy work, but mispells Steward's name, and does not properly cite their work.

In the last section, it's suggested that the original Hebb formulation needs "tweaking". But the timing aspect is already implicit in the original version: the only way the pre-spike can cause the post-spike is if it precedes it. — Preceding unsigned comment added by 24.188.39.29 (talk) 01:10, 10 November 2012 (UTC)

Connection to synapses and synaptic plasticity?
I find it unclear (after only having read the text before the first section) how this article relates to synapses and synaptic plasticity. The article later mentions synapses, but does not mention synaptic plasticity at all. These two terms should be mentioned in the first paragraph, as this article at least seems to (by judging from the name of it) be very closely related to both of them. —Kri (talk) 20:41, 12 February 2014 (UTC)

Mechanisms
I do not see the logic for the mechanism of STDP and NMDA receptors. The backpropagating AP from the postsynaptic neuron is facilitating the opening of NMDA receptors as there is release of Glu from the presynapse. But then first activating the postsynaptic and then the presynaptic neuron should facilitate STDP because activation of the postsynaptic neuron causes the backpropagating AP. But it is vice versa. Do you see my point? LuxMaryn (talk) 12:09, 29 May 2016 (UTC)