MIT discovers the location of memories: Individual neurons

OMG, it's the ultimate mechanical_fish pet peeve collection! TL;DR: I rant.<p>ONE: I can't find the citation of the scientific paper on ExtremeTech. (Let alone a link. Who would dare to dream of a link?) They do refer obliquely to "the paper" once. (What paper?)<p>And they link to MIT's press office, whose brand is really solid, so everything they write is almost like science! And there you can skim the article twice and finally spot the citation:<p><i>Susumu Tonegawa, the Picower Professor of Biology and Neuroscience at MIT and lead author of the study reported online today in the journal Nature.</i><p>Okay, gotta go find an online <i>Nature</i> subscription to find out what's going on. There's an hour of my day spoken for. At least they're trying to ensure I get some exercise.<p>Why is the actual journal article important? Just look at this thread here on HN. We have people doubting all sorts of things, but these may well be things that are addressed in the <i>actual work</i>. The content of the popular articles means very little: They leave out most of the details. The details matter. The whole <i>point</i> of this study is to try and tease out more details.)<p>TWO: But, wait, there's more. I first tried to read ExtremeTech on an iPad (original edition). The article popped up in one of those insufferable iPad-only JS-powered "mobile editions" with Swiping Action. Unfortunately, there was only the first page of the article. It cut off in midsentence. I tried pressing the giant button marked "Next" on the right side of the screen. I got a big white screen. I flailed around with my fingers. A <i>completely different article</i> eventually rendered itself. I flailed around with my fingers some more. Eventually the original article reappeared, still incomplete. Fortunately, one more roundtrip to the next article and back and I finally got the whole thing to render.<p>Then I pressed the back button and everything seemed to hang. I closed the browser window and thanked the gods for my escape.<p>Why on earth do publications use these broken things, when simple web pages render so nicely on the iPad? The site does have something like a dozen tracking cookies on it; does this imply that they have data showing that swipability is so important that it <i>doesn't even have to work</i> in order to attract more ad impressions/clicks/Tweets/whatever? Or does it merely suggest that they are so busy struggling with glass-cockpit syndrome that they can't perceive that their site is broken on the iPad?

The data doesn't warrant the title of this post. Even if a single neuron is responsible for triggering a memory (which is hard to say based on mice in the first place), it doesn't follow that the information is stored in the neuron. As an analogy, if we erase a specific bit in memory, whole parts can become unreadable. For example imagine changing a bit in a pointer. That doesn't mean that all of the information was stored in that bit.

I can't help but think that what they've discovered is some meatspace equivalent to the hash key or an index key to a memory. By "turning it on or off" you can lose or find a whole table row or hash value, but surely a "single neuron" can't "store" a generalised "memory".

They didn't 'discover the location of memories', they simply discovered that they can trigger a reaction that looks similar to that of the original incident by focusing energy on parts of a mouse brain that were actively stimulated during the incident.<p>This is Pavlovs dog with optics, not revolutionary science.

I always was under impression that there are no memories as we think about photographs, but rather state of the brain is one whole memory system, which react to external signals accordingly to previously "memorized" signals.<p>As I understand, that they just found a way to excite particular small part of the brain which triggered reaction without external signaling. I.e. bypassing all intermediate parts (external sensors, nerves, other neurons, etc). I.e. it is more like to directly feed engine with gas and electricity to create a spark to make it revolve completely skipping engine control unit, ignition key, clutch pedal, fuel pump.<p>And when they refer to memory loss due dying neuron, this is more like part of circuit is removed which supposed to give certain reaction to certain signals, not like "rm /home/user/file.txt" with the rest left in place. So I do not believe thing similar to MiB is possible with this knowledge.

'The mice “quickly entered a defensive, immobile crouch,” strongly suggesting the fear memory was being recalled.'<p>Yes, possibly. Or maybe it had something to do with the hole that had been drilled in their skulls and the frickin' laser beams being fired at their brains. Pretty cool stuff though.

The journal article can be found here: <a href="http://www.nature.com/nature/journal/vnfv/ncurrent/full/nature11028.html" rel="nofollow">http://www.nature.com/nature/journal/vnfv/ncurrent/full/natu...</a><p>It doesn't seem to claim that an <i>individual neuron</i> is the location of a memory, but rather that triggering a small number of specific neurons is necessary and sufficient to cause the behavior that would be consistent with the recall of a particular fearful memory in mice, when optogenetically triggered.<p>In other words, they labeled some neurons with an optogenetic receptor during fear conditioning. Then, thanks to the optogenetic labeling they previously did, they were able to activate this receptor (using light) in a totally different context (one that didn't normally elicit the freezing response associated with mouse fear). When they did so, the mouse exhibited the freezing response. When they ablate these neurons, there is no fear response. The conclusion is that these neurons are necessary and sufficient to encode the fear memory.<p><i>Caveat lector</i>: my summary is based on the <i>abstract</i> so I am just parroting what they have concluded; I haven't read the paper's methods and results for myself. Also, this is the "near-final" version published in advance online today. It may change for final publication.

Impressive as this may sound this is not the first study of its kind. Studies since 2009 (mentioned in the abstract) in the amygdala have been able to direct and inactivate fear memories in a reversible manner via optogenetics again (see <a href="http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2844777/" rel="nofollow">http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2844777/</a>) . In these studies killing as much as 15% of the cells did not erase the memory, indicating that it's ensembles, not single neurons that encode fear memories.<p>To complicate the picture even more, each neuron is part of more than one memory engrams, and memories are stored in synaptic connections which are formed in the vast dendritic trees of neurons. So you have a very sparse and reduntant encoding of this associative memory.<p>Finally, note that both the amygdala and the hippocampus are very old structures so we don't know if the same processes take place in the neocortex (although it's likely so).<p>I don't mean to belittle the article, but it's mostly proof of concept if you have followed the relative literature. Tonegawa's lab had some even more fascinating papers published recently that probe the process of memory encoding to the level of single dendrites.

Here is a simple explanation of the experiment:<p>1. The set of neurons that was active only during learning was determined.<p>2. The genes activated in those neurons were determined.<p>3. Genetic engineering was done to make the activation of those genes always happen in conjunction with activation of the gene being responsible for the neuron becoming sensitive to light.<p>4. The mouse was put through a learning experience, during which the small group of neurons affected became sensitive to light.<p>5. Via stimulating those neurons with light, the experience was reproduced in the mouse in a completely different environment (so the comparisions to Pavlov are not justified).<p>In this way the abstract concept of a memory and of the process of remembering something was related very closely to a specific physical phenomena. Even the methods used to make the experiment are interesting by themselves (at least for a lay person) and I think you cannot easily dismiss the importance of this discovery as some people do here. Please read the article here to get a better picture:<p><a href="http://www.mit.edu/newsoffice/2012/conjuring-memories-artificially-0322.html" rel="nofollow">http://www.mit.edu/newsoffice/2012/conjuring-memories-artifi...</a>

I wish I could block all extremetech stories from My HN feed. The are universally overhyped, never the original source, often draw unjustified conclusions, and are a pain to read (at least on an iPad - they use a non standard interface and pop overs).

MIT scientist Sebastian Seung talked about this possibility in his 2010 TED talk, "I am my connectome" (<a href="http://www.ted.com/talks/sebastian_seung.html" rel="nofollow">http://www.ted.com/talks/sebastian_seung.html</a>).

I think this has been overhyped just a bit. I realize that a discovery of this nature is a big deal, but it's not what journalists are making it out to be.

&#62;MIT researchers have shown, for the first time ever, that memories are stored in specific brain cells. By triggering a single neuron, the researchers were able to force the subject to recall a specific memory. By removing this neuron, the subject would lose that memory.<p>That's incredibly poor reasoning. Using the same logic, I can "show, for the first time ever" that C structures are stored in individual memory pointers.

I read the MIT news piece and didn't find any reference to the removal of a neuron that would eliminate the memory. Was that in a longer research paper?

My brother, who is studying in medical school right now pointed this out to me: "the whole time i was reading this though, i was bothered by what usually bothers me about studies regarding the mechanics of the brain, which is that we dont know our measures of "activation" are sufficient for determining a causal relationship between activation of certain neurons and the recall of certain memories."

Off topic: I live across the river from MIT. I work in asset management and worked for an angel investment fund. I'm working on a startup at night. If you live in Boston and want to meet for coffee email me.<p>I'll meet anyone from HN, even if you just want some career advice for working in asset management or just want to talk about random stuff.<p>My email address is my hn username at gmail.

Maybe this means we can finally build a machine like the one in the matrix that lets us bootstrap our brains with kungfu.<p>But more importantly, I think the most significant result of this research, should even a bit of it prove to be true, is the possibility that it could help us understand, and maybe even prevent dementia. That's exciting!

&#62; The question now, though, is how memories are actually encoded — can we programmatically create new memories and thus learn entire subjects by inserting a laser into our brain?<p>That's a pretty scary area of technology. I hope that as a civilization we will become responsible enough to use that before we discover how.

I wonder if by deleting not so nice memories (i'm guessing that would be one of the first applications) the person will continue to make the same decisions over and over, because they might not remember them and keep re-experiencing (and learning) repeatedly?

If you are interested in this sort of thing (memory storage in the brain, recall etc.) you might be interested in Wilder Penfield's work. (e.g. <a href="http://primal-page.com/penfield.htm" rel="nofollow">http://primal-page.com/penfield.htm</a>)

"The mice 'quickly entered a defensive, immobile crouch,' strongly suggesting the fear memory was being recalled."<p>Perhaps the defensive, immobile crouch had something to do with having a hole drilled through their skull and a laser shined through said hole?

&#60;you drill a hole through the subject’s skull and point the laser at a small cluster of neurons, the mice quickly entered a defensive state&#62;<p><pre><code> This is a expected response and perhaps has nothing to do with remembering.</code></pre>

This article sucks, the quality of online journalism has taken a new low...there are no citations...just rambling that eventually one finds to be a total misunderstanding of the results and of science in general...ahhhh...

There was a Ted talk some time ago about the work on optogenetics (in fruit flies I think) <a href="http://www.ted.com/talks/gero_miesenboeck.html" rel="nofollow">http://www.ted.com/talks/gero_miesenboeck.html</a>

That title is incorrect. These researchers did not stimulate one neuron to activate one memory. They stimulated a group of neurons in one region using a fiber optic.<p>I'll go through the experiment for those who don't really get the write up. Quick background:<p>0. Neurons are specialized cells in your brain, their firing is the basis of cognition. Neurons that fire strongly together tend to get linked. Not all your neurons are firing all at once. Not all neurons participate in a memory, but a large collection (0.1-4% depending on brain region) are used for each memory. (at least in naive mice) Memories are highly distributed across brain regions and within brain regions.<p>1. There is a gene 'cFos' that turns on in neurons that undergo activity. It is very short lasting, about an hour, and then it is back down. It is very cell specific, only cells used have been activated will show this gene.<p>2. There is a genetically inducable protein you can put in cells so that when you shine them with light they will start firing.<p>3. Anything genetically inducible can have a tag added that will make it impossible to induce when an animal is on a specific drug. (We'll call it Dox.)<p>The researchers made it so that when cFos is activated, it will transcribe the inducible light activated channel. But, it will only do this when the animal isn't on Dox. So they took the animal off Dox, exposed it to fear conditioning (they shocked it in a unique box), then put it back on Dox. So only the cells that were active during the fear conditioning will be turn on when they shine light.<p>They then put the animal in a new box and shined light, and the animal froze. (A sign it was afraid) They concluded that activating cells that had been active during the storing of the memory can 'reactivate' the memory, even out of it's context.<p>Of course there are a few major problems with this. The region of the brain they worked in is a relatively minor area of the brain. They admit in the course of the paper that each cell is used to store many memories. (They claim the assembly of cells activate the memory? But they never bother testing this.) Their firing pattern is far from physiological. The mice who underwent this 'memory reactivation' did not freeze as much as mice in normal fear conditioning and did not seem to learn the 'reactivated memory' at all. (Though again they didn't really bother testing this.) Normal fear-memory is learned quite well. This is a very preliminary but very important study in the field. It will be interesting to see how they follow this up.<p>Also, the Npas4 stuff is completely irrelevant to this article. There are countless markers that when 'knocked out' make it impossible to store memory.<p>Also here is a (paywalled) link to the paper:<a href="http://www.nature.com/nature/journal/vnfv/ncurrent/full/nature11028.html" rel="nofollow">http://www.nature.com/nature/journal/vnfv/ncurrent/full/natu...</a><p>Here is a link to another (paywalled) paper which came out today which did almost the same thing and say almost opposite results (the details explain this oddity): <a href="http://www.sciencemag.org/content/335/6075/1513.full?rss=1" rel="nofollow">http://www.sciencemag.org/content/335/6075/1513.full?rss=1</a><p>Here is a (paywalled) more layperson write up to the second article: <a href="http://www.sciencemag.org/content/335/6075/1455.full" rel="nofollow">http://www.sciencemag.org/content/335/6075/1455.full</a>

<a href="http://blogs.nature.com/actionpotential/2012/03/fear-of-the-light.html" rel="nofollow">http://blogs.nature.com/actionpotential/2012/03/fear-of-the-...</a>

Besides the fact that the scientists used trauma to retrieve their data, what a beautiful notion: collecting a memory at the tip of every neuron.

I don't believe a word of this. It isn't <i>proof</i> of any such a process. This to me appears as inappropriate reductionism.<p>The brain is highly volatile both chemically and physically and is changing minute by minute. The notion that a memory belongs to a single group or cluster of neurons is absurd. They are constantly being reordered and reprioritised and changed throughout the lifetime of the animal. And being lost due to chemical damage and the effects of ageing and free radicals.<p>The nature of memory and conciousness isn't just the physical existence of neurons. Who <i>you</i> are is more than the sum of the parts.<p>The mind is also as much a process of the body -- your conciousness and mind grows within it from birth and a large part of who you are is as much a product of that, its hormones and its metabolic chemistry as anything that resides solely between the ears alone.

The Nature abstract is about a "population of neurons" and triggering a memory related behavior while the extremetech article talks about "specific brain cells" and memory storage.<p>Nothing to see here, move along.

I always take these with a grain of salt. One day a study comes out saying that memories are not stored in a specific location, that scientists can cut out large portions of mice brain and the mice will retain memories, and the next day a study comes out that specific memories can be created or destroyed with specific neurons.

There's an obligatory Eternal Sunshine of the Spotless Mind reference to be made here...