Where Is The Memory Stored In The Neuron.

Yes, maybe, perhaps.

Are you a medical student?

I am inclined to report your threads to admin. as inappropriate.

I beg to disagree. This is utter curiosity with thoughts and questions. Its not a bad thing!

It might be answered by medical studies, but I am more inclined to think that we (mankind) might not even have this knowledge or only very basic.

For OP it would be wise to ask his questions in a rather less technical forum.

My opinion only.

These questions cannot be answered by people without medical training.

Admin. will act, or not.

How confusing and unscientific to claim that some questions should simply not be asked in a general engineering forum and scrubbed from the forum if asked.

I do not have any authority here. Everyone is free to do as they please.

You have medical training. I do not.

Nobody knows what the OP's motivations are.

These may well be questions being asked by a technician, I don't know.

Maybe a medical student?

I'm done interfering. Carry on.

My parents always complained that I was carrying on.

Had a bad day ? The question looks OK IMHO, and I'm fairly sure it says on the CR4 ticket that this is the place for discussion on Science and Engineering.

_{<before anybody points it out, I have no idea how I snuck in >}

I don't think the OP was going along the lines of maybe doing home lobotomy, just asking about the nature of the general topic as stated. It's not my thang, but I'll be interested to read any responces on the topic.

See my contribution above yours.

Brain surgery isn't my cup-o-tea.

Seems a little far afield for the forum, but, I've mad a stab at some information.

Just can't remember why.

...and full credit to you for some mad stabbing .

Staying on topic, as I always do, there is a hilarious story about some Brit MP. Upon somebody pointing out that he had spilled some red sauce over his shirt at a dinner function, he exclaimed," ******* hell, somebody has stabbed me in the front !".

Whilst I would never want to hijack a thread, the question might have been better if asked along the lines of ' how good is your memory, and do you have any specific ways to help yo......Look this is not my fault - when replying to a post, I can't see the original question. Hardly helps that I lack conce_ntra........

I suffer from synapse collapse quite often. Spell chack doesn't understand my neurons so I just rely on people to understand what I mean.

Even those of you who talk funny and use too many U's.

I don't think medical training will be helpful in this. It will take years of very basic research using the simplest of animals with neurons, such as nematodes, to find this out. Typical doctors don't perform any research at all, and it is a stretch for them to even keep up with medical research, let alone basic research like this.

In order to be even a GP doctor you need to go through training. Even premed basic anatomy class provides a lot of insight into these matters. Other classes such as physiology, etc. delve deeper into the subject.

Medical training is actually a very diverse field of studies.

I agree, but the level and type of training has nothing to do with doing research on how memory might be stored. They might find out a little on the way about what people think might be happening, but that has nothing to do with performing medicine. More below.

I think medical training would be absolutely essential to even begin to try to understand the mechanisms involved. Then, add to that electrical engineering, physics and maybe another or two.

I believe that doctors do perform a lot of research. Your typical GP may not, but........

Let's say that I asked you to trace the electrical control circuit of an axillary power unit of a B747 from the cockpit to the APU. Could you do it without training?

Memory is stored evanescently in the synapses; it is all god of the gaps.

Nothing like that at all.

The hippocampus is essentially the coordinator for memory creation and retrieval.

A single simple memory is actually a collection of hundreds of thousands or millions of neurons that form a set of many, many fragments of memory associations.

It is more holographic in nature, which is completely different than the way a computer stores data.

The frontal cortex is pretty much the seat of all cognitive memory, but other areas of the brain are used for other types of memory such a motor skills.

Are you replying to post 4, which seems facetious?

My understanding is that the connections between neurons (synapses) is where learning takes place. Engineers have built artificial neural networks (usually built in software) that are capable of learning patterns by varying the connection between network nodes (model neurons).

http://en.wikipedia.org/wiki/Artificial_neural_network

Neural connections that have been made in the past in response to sensory inputs (say seeing a familiar face) are strengthened and therefore more easily evoked at a later time in response to the same or similar sensory input, and this basically is a memory.

great, thanks for your answer

It is my understanding that we do not yet know the specific mechanism at the neuron level of how a brain keeps a memory. I suspect that some memory will be neuron based data feeding back on itself (short term) while others will be chemical based modifications to neuron connection paths (long term).

great, thank you for your answer

MIT discovers the location of memories: Individual neurons ...

Yes, the location but not the mechanism. They are stored inside neurons, it seems.

As for the mechanism, I have been waiting for someone to discover that for quite a long time. My bet is that they are encoded in microtubules, because they are in all cells and present in large quantities, and have a structure that is conducive to binary information storage, one protein unit per bit. And the amazing length of the axon might be implicated in the production of even larger quantities in neurons.

But that is just my own take on the circumstantial evidence. I'm eager to find out if I'm right.

I am thinking that memories are not actually stored in individual cells as implied by your answer (and by answers from some others), but are actually specific patterns created by a much larger network of neurons working together.

The microtubules in axons, dendrites, etc., are all components that facilitate the process, but not the harborers of the actual memory.

I know another post demonstrated a recall of memory when a rat's photoreceptive nerve cell was struck with a laser, but I think that single cell's stimulation initiated a cascade event that triggered a memory.

Part of the problem is the sheer complexity of the subject and like any complex matter we tend to reduce it into terms that are easier to grasp. In this instance we create analogies to computers and RAM because we are at ease with the simple anatomy of a computer. However, that analogy is really false as the brain doesn't store information in the same way as a silicon computer nor does it process information like a Von Neumann machine.

Absolutely correct but it is not just a convenience to refer back to much simpler RAM, latch and magnetic memory design circuits. These are memory devices that we do understand because we designed them. We naturally compare the unknown with the known. I do take exception to the idea that the human brain does not process information like a Von Neumann machine. We do not know how the brain works. We should not prematurely exclude any proven mechanism as a possibility until it is known. At the very least it is a good basis to compare how a well understood mechanism handles tasks versus an unknown mechanism does.

One clue that the overall architecture is the speed of processing. The brain processes very slowly compared to even a simple PC computer. Neurons simply do not fire nor propagate that fast.

However, managing abstract problems is much easier/faster for the brain to process compared to a PC.

The brain can not process numerical calculations anywhere close to a simple PC.

While I agree we really don't understand the brain, there are clues that tell us that the brain must operate much differently than a modern computer. That tells us that the overall architecture must be different.

The brain has to be massively parallel.

I have long thought that when I hear or see something, my whole brain is listening and watching, and the parts that have something relevant to the input respond to it. The strongest response wins, but all are present.

I've played around a bit with "Artificial Neural Networks", and they do learn by varying the connection strength (weights) between neurons in successive layers. The training is done by "Back Propagation" which is unlikely the process employed by the brain. The connection strengths do store learned information, but it may or may not be true that this is how actual neurons in the brain store information.

I have thought about that. I try to imagine how a group of neurons can store a memory by creating dendrites and attaching them to each other. Would each dendrite be one bit? Or is the location of the dendrite or the location of its connection somehow a factor? But memories have visual, audio, temporal, and spatial components. How would a group of neurons store all that in a pattern of connections among them? And would that mean that an amoeba cannot remember anything because they are only one cell?

It all seems very far-fetched (but it could still be correct.)

An amoeba is not cognitive, but will react to stimuli.

It's been a long time since my med school classes, so a lot of my knowledge in this domain is getting obsolete...

However, what I learned was that every sensation we get creates it's own pattern of stimulated neurons. These complex memory patterns can overlap with other patterns and create associations.

For example, a pleasant scent can be associated with the memory of aural and/or visual stimuli. Walking into a house with a pie backing may evoke the sensations of your mother or grandmother doing the same.

Emotional responses can also be associated with those patterns. You may feel fear when you see or hear a sound because that stimuli caused fright or pain in your past.

There really is no such thing as a simple memory because any one thought has a cascade of other associated thoughts, sensations, or experiences that are all made up of these complex set of neuronal patterns that fire.

Even though a remembrance may evoke an avalanche of related thoughts, that is nothing like having an epileptic fit.

Epilepsy is when neurons fire uncontrollably in a section of the brain or sometimes it spreads over the whole brain. The synapses will simply not stop firing in an uncontrolled way. If it goes on too long (status epilepticus) it can be life threatening.

We had a dog here that I could sometimes break his seizure by stimulating him by touch if could intervene quickly enough. The reverse was true where a loud noise would induce a seizure.

I am rambling, but most of my undergraduate work with my mentor doctor was in epilepsy.

Turns out amoebae can remember. Google "can an amoeba remember a stimulus?" It does not follow that they use the same mechanism as humans, but it is possible they do. But microtubules can provide a detailed and compact way of storing and recalling a memory with a lot of data, such as a picture or sound, that needs to stay in a particular structure, such as a visual field or a temporal sequence. In fact, if I wanted to I could invent a reasonable (but entirely hypothetical) mechanism of turning a neural stimulus into a stored memory, and retrieving it again, using microtubules.

Yes, complex memory patterns can be created and overlap with other memory patterns, and they can stimulate other memory patterns. Interestingly, some memories can be stored with the emotional part and factual parts separated into different hemispheres, or at least that is one interpretation of the data. PTSD can result when a vivid and dangerous event is saved (apparently in the right hemisphere) without without the proper massaging and reduction of the emotional data. When the memory is recalled, it can "play" again like a movie, with all the traumatic feelings being replayed back just as vividly as the original experience. But when the traumatic memory is processed properly, the memory becomes just a "bad memory" from the past instead of a relived experience.

I think a single neuron can "store" a specific excitation response, but I would not call that a detailed memory.

For example, I don't think you could find your home phone number encoded into a single neuron. For that it would take thousands and thousands of neurons working together to form a set of patterns that represent that number.

At least that is how I understand it.

Yeah, that's the part I can't really accept. Using thousands of neurons to store a phone number sounds incredibly inefficient. If you had just a few neurons to store each single digit, such as the concept of "3," and then another neuron stores 9 links to 9 of those neurons, that would be much more efficient. I used to do object oriented programming as a professional.

Not really. A single neuron can support many different memory patterns.

Think of a chess board. If every square could only store one memory then the board's limit is only 64.

However, if a memory was composed of a combination of squares, then the number of potential memories is much, much higher.

True. You could get a lot of data out of the permutations of a low number of neurons, but that also means we are talking about a very large number of dendrites. I think you are saying that somehow a few neurons (few like in thousands?) get connected in a way such that only that one combination somehow "means" his phone number.

If someone asks him his for his phone number, that triggers that combination of those neurons to be lit up, and that in turn stimulates a signal to be sent that says "XXX-YYY-ZZZZ."

I think they have found that memories are stored inside single neurons, because when they experimented with stimulating single neurons the subject remembered a single memory. Was it the "Angelina Jolie" neuron, or something like that?

I don't see a conflict with stimulating one neuron as it cascades to complete a pattern.

I would bet that in the end we will find that what you propose is not true. Anyway, if one neuron contained a phone number, how is that extracted? CAN-bus?

I do see a conflict with stimulating one neuron that cascades to complete a pattern. Which pattern does it complete? If only a combination of many neurons can represent a memory, and each neuron belongs to many patterns, then which of the many patterns that it belongs to does it complete?

I only have hypotheses about how a memory would be extracted. If one neuron contained a phone number encoded in microtubules, then it is extracted by having a "read" mechanism run along the microtubule, reading each bit in a given time interval. Whenever it hits an "on" bit, it stimulates a firing of the neuron. The pattern of firings encodes a complex signal that is picked up by other neurons. The complex signal may indicate a packet of information like a phone number, or the identify of a particular neuron that contains one element of that phone number. Or it could be sent to many other neurons. Those neurons would listen to the pattern of firings in the signal and read the stored bits in its own microtubules simultaneously as the firings come in. If there is a mismatch, it stops and ignores the signal. If there is a match of the entire signal, the second neuron would say, "I know that signal! It is familiar to me somehow," and send out its own signal saying it remembers that phone number and what it is associated with in itself.

Purely hypothetical.

"...then which of the many patterns that it belongs to does it complete?"

I am fuzzy on this, but neurons can contain multiple dendrites as well as multiple terminal buttons. In other words, a single neuron has connections to more than one other neuron on both input and output sides of its dendrites and axiom.

Again, I am not sure if my own memory on this is correct, but the axiom is the output and the hillock is the gate which determines at what action potential the nerve fires and which terminal buttons are excited.

So it is possible for the neuron to act as complex routing switch to selectively fire its outputs based on which inputs are stimulated.

When I get time I will check on that to make sure I am not full of BS.

Just some cursory research that confirms my thoughts on memory storage.

• The brain has about 100 billion neurons
• The brain has about 1 trillion support cells (glia, astrocytes, etc.)
• Each neuron has from 1 to 10,000 synaptical connections
• The brain has about 1,000 trillion synapses
• Each neuron action potential fires at a rate of about 50 Hz
  
• Estimated memory capacity of the brain ranges from a low of 1 to 1,000 terabytes

Reasons why a memory is not in a single cell:

1. If it was true that a single neuron held a memory, the rate that information could be accessed is bound by the action potential frequency of 50 Hz. 50 bits of information is too slow for memory retrieval.

2. Even if all 100 billion neurons contained memory information that means that each neuron would need to contain 9 bits of data if the capacity of the brain was 1 terabyte and as much as 9,000 bits if the capacity is 1,000 terabytes.

Obviously, not every neuron in the brain is used for memory, so that number would have to be much higher.

Again, the rate that information flows out of a neuron limits the volume of date that it can contain and still function with enough speed so you could cross the street and not get killed.

3. The neural network is not organized in a row/column format, so addressing single memories would imply a CPU or MMU processing unit to address, store, and retrieve information. We do know the brain doesn't quite function in that way (Von Neumann architecture)

This strongly suggests that the brain stores information in neural nets rather than in discrete cells. Although the exact mechanism is not clear, there is a preponderance of evidence that suggests that memory is a complex network of associated neurons working as a system rather than in independent storage units in each cell.

That's what I know so far.

Some thoughts:

1. I think 50 Hz cannot be too slow for memory retrieval, because it uses that rate regardless of how the memory is stored. Storing it in a pattern of cells sounds even slower than storing it in one cell. What mechanism are you thinking of that would make it faster?

2. I don't know how they made that estimate of the brain's capacity. Clearly they did not start filling it up until it overflowed, so it must be based on some assumptions. I would be surprised if each neuron stored only 9 bits. Amoebae must be storing far more that that in order to remember unpleasant stimuli. A single microtubule must be able to store millions of bits.

I don't know how you are judging how fast information can flow out of a neuron. The speed that a neuron can fire is already pretty slow, and it is a complicated chemical system with pumping sodium ions, calcium ions, and potassium ions in and out of the cell, and that is a pretty slow process, relying on diffusion. In contrast, a mechanism inside the cell need not rely on diffusion, but on proteins attached to a linear memory storage substance like a microtubule or a protein, and that sounds like it is probably waiting around a lot for the cell to fire. It seems to me that without knowing how a neuron stores a memory you can't know whether that is the limiting factor in retrieving a memory.

3. I don't see how you can make that assumption about needing a CPU without knowing how the brain operates. I think memory must be massively parallel, so a central CPU may be unnecessary.

I will agree that it is possible that memories are stored in a network of associated neurons, but I disagree that the evidence points to that.

1. Parallel processing is faster than serial.

2. It was a pretty wide margin, so they obviously do not know with any precision. However, even the low end estimate substantiates the point.

I am wondering on the role of the microtubules, too. Maybe they control which synapses fire at the end of an axiom? I need to do more research, but it is an interesting approach.

3. the CPU or memory management unit is the postman for a Von Neumann architecture. You need some mechanism to address the location of data and a row/column addressing system is how RAM is addressed. The brain doesn't work that way.

You have made some very good posts here, but a chess board storing 64 chunks of data ? I'm, sure you are right, but the analogy escapes me. Binary, I might just about understand, but you seem to suggest a unary type mechanism.

Each square either does, or does not, store info. In other words it must be binary. What am I failing to grasp here ?

The chessboard was a metaphor that I used to describe that more information is possible to be stored as a pattern of cells rather than each cell containing a unit of information.

Think of it this way. If each square contained a unit of information, then only 64 units of information can be stored.

If each unit of information is represented as a pattern, you not only get the same 64 cases of unique information, but you could have all possible combinations of two squares representing a unit of information, all possible combinations of 3 squares, and all possible combinations of 4 squares, etc., until all possible combinations are exhausted.

That's a crude example, but I wanted to demonstrate the power that a pattern holds over a "bit" or unit type storage.

I am also hesitant to describe this as binary because the brain doesn't work like a typical digital computer. Unlike a computer where a unique RAM location holds one unit of information, the brain can use that same location for part of more than one memory.

When I get more time I will need to refresh my own brain on the details and also do some research on what the latest understanding on this is. I am under multiple deadlines that are cramping my free time this weekend.

I think I get your point.

Somewhere on the dusty shelves I have a book that goes into great detail on the topic. I never managed to read the whole thing, but I recall some mention of the brain storing info as a series of cascading images. I can`t vouch for that at all, but when reading it, the general message seemed a bit like `a picture can describe a thousand words`. That still does nothing to help answer the original question, but perhaps supports your mention that a human brain does not fuction like a computer. The architecture is just plain different.

Even sitting here doing very little, my brain is processing huge quantities of data. It`s only recently that people have created some of those incredible walking/running robotic critters, and that is just the simple act of walking or running. Somehow the human brain can process data far better than any logic we apply to computer programs. Whoever can figure out why will end up making Bill Gates look like a pauper.

"Cells that fire together, wire together"

See Hebbian learning:

http://en.wikipedia.org/wiki/Hebbian_theory

What do you think of the link in #14?

MIT discovers the location of memories: Individual neurons ...

I am not convinced. Just because one neuron is involved in a memory (assuming that is what is happening here) does not mean that the memory is contained in only that neuron. Each neuron is connected to numerous other neurons.

(If you remove the right rear tire from a car, the car cannot move, but more is involved in motion of the car than that wheel.)

It certainly is a puzzle.

Between the two "n"s.

Hebbian learning: "Cells that fire together, wire together"

http://www.nbb.cornell.edu/neurobio/linster/lecture4.pdf

Here read this maybe it will help.

http://science.howstuffworks.com/life/inside-the-mind/human-brain/human-memory1.htm

If synapse connections with dendrites, and neurotransmitters are how memories are stored/encoded, because there are more synapes/dendrites than neurons, then that makes more sense in terms of storing memories.

There are 100 billion neurons in the human brain, and trillions of synapses. Because if just groups of neurons held memories, that would mean lots of groups of neurons would need to hold memories, because there are more synapses than neurons, then there is more space to hold more memories.

I haven't seen a reasonable mechanism in which a group of cells can store a memory, no matter how many synapses and dendrites are involved. A synapse is a structure that allows one neuron to pass a neural impulse to another neuron, and I don't see anything in particular that suggests that it can help store a memory. Same with neurotransmitters. They allow a neural impulse to cross a synapse, but why would that be related to storing a memory?

We tend to think of memory in terms of computer memory, which is nothing more than a collection of electronic gates. That is little different than a synapse.

However it seems pretty clear to me that the brain doesn't store ones and zeros, so the similarity to computers and RAM that we tend to assign to the brain ends there.

It is not correct to say that an electronic gate is little different than a synapse. As far as we know a synapse has no other capabiliity than to pass an impulse from one neuron to another, not to remember what it passed. Yes, some electronic gates are able to store binary data and are used as memory, but many just modify the transmission of data being passed back and forth by other electronic gates.

I don't think of human memory in terms of computer memory, because I think evolution is far more creative than we are, and capable of leaps of function exactly because it does not use logic in its design. It uses random mutation and recombination, and whatever works sticks around, and whatever does not work is lost.

Regarding ones, and zeroes, what a neuron receives as input is a neural depolarization, which is just a "one" as compared with the normal resting state, which is a "zero". I haven't heard that a neuron can transmit a neural impulse in anything but all or nothing mode.

What the neuron (or group of neurons) stores is not yet known, but it must be able to create a new neural impulse as its output, which is again a "one" or "zero". I don't see any reason to think that what it stores is not just ones and zeroes, considering that its input and output are both ones and zeroes. Why does it seem pretty clear to you that it doesn't?

No, I said that memory storage for the brain is or appears to be very different than the way a computer stores memory, despite any similarity between an electronic gate and a synapse.