Funny and interesting, and it was voted a finalist so I have to post it - but I don't believe this at all.
Main concern: how do you encode memories in a chemical? That is, suppose you want to encode the fact "My name is Tim". It's easy to imagine how this works as neural connections: during infancy, you hear language, your brain forms a predictive model of the language, you get neurons representing the sounds T, I, and M, and the word "name", and you get links between them. Eventually there's some complicated network connecting name, T + I + M, the sight of a nametag, the motor program of shaking hands, and all the other concepts relating to your name being Tim. We know something like this has to work because LLMs can have this type of memory and they're nothing but connections.
How would you implement this in a protein? All I can think of is a binary-like code with literal text - using one amino acid to represent 0, another to represent 1, and coding the ASCII string "M-Y-N-A-M-E-I-S-T-I-M", and then having a decoder somewhere in the brain. But this doesn't match the phenomenology of human memory, which is sensory (eg associated with pictures, smells, etc) and works even without language (eg in preverbal children). CF https://slatestarcodex.com/2017/09/07/how-do-we-get-breasts-out-of-bayes-theorem/ , which asks a similar question about genetic memory.
But also, even supposing you could do this - in order for the organ transplants to make sense, you need for the heart to be producing or storing these chemicals (why?) in such large quantities that they get into the bloodstream, cross the BBB (why would you have a transporter for these?), make it into the brain, and perfectly interface with the host's own memory-decoder system (how?). Then either the heart has to keep making these constantly (why?) or there has to be some mechanism for a single low-dose exposure to get the host brain to start making them itself (why?) And if the cannibalism claim is to make sense, you have to be able to absorb these extremely complex chemicals through the digestive tract, which is usually doing a pretty good job preventing you from absorbing complex chemicals. Why do this, rather than just keep the chemicals in the relevant part of the brain?
Also, if we had these chemicals, a natural application of them would be genetic memory, eg in the womb you get a big dose of your mother's memory chemicals and now you understand the world as well as she does. As far as I know there are far fewer stories of people mysteriously having their mother's memories compared to their organ donors' memories, adjusted for the ratio of people who have mothers vs. organ transplants.
Also, the writer asks whether losing your arm might make you forget something, or eating an animal liver might give you some of the animal's memories. But many people lose limbs and eat animals, and as far as I know nobody reports these phenomena.
Agreed, I think this review goes way too far. Sure there are other mechanisms for memory in the brain besides synapses. No, I don't believe anything about heart transplants transferring memory.
I think the way you would encode "TIM" in protein is in a protein network. You could conceivably do this inside a single cell, AFAIK.
The problem is cells don't have mouths to say "my name is Tim" with. Within-cell memory is absolutely real and important, but it's just not the right scale for organism-level behavior.
You could have a network of lots of different proteins, but I don't understand how the symbols would be grounded. For example, what would it mean to say that arginine = M instead of N? Unless it bottoms out in some kind of behavior like saying an M or hearing an M, it's just arginine.
...maybe this is just the same point you're making.
(Upon closer reread, I think you were making the point I was making, I just hadn't fully understood my own point)
Not a protein expert, but I think many proteins are very weird and have many different possible configurations, and those configurations can be influenced by all sorts of features in the environment. In theory you could have a "name" protein that has a different configuration for each of the 1k most common names in the English language or something. This would be a weird way to do things, though!
Have you read "How Life Works" by Phillip Ball? As someone who had done their best to avoid modern molecular biology in grad school, I found it very useful for making sense of the snippets I had heard. It made molecular biology sound a frightening amount like social science.
I still don't understand how this would be grounded. That is, I 100% agree a protein could have 1000 different states. But what makes those states correspond to 1000 different names, as opposed to 1000 different shades of blue?
Suppose you hear the name Tim, and it equals state #489. You need the brain (which is doing some kind of auditory processing and has decided that "Tim" is important enough to devote a protein state slot to) to encode that by modifying the protein to state #489. Then the protein goes to the heart, and the heart is transplanted into a different person. Then the protein goes to the second person's brain. Unless there is a natural mapping between names and protein states, how does the second person get the same protein state code where it knows that a protein in state #489 equals the name "Tim"?
(for that matter, how does the first person establish a protein-state-to-name code which is consistent across all relevant brain cells?)
I agree! These are all very good reasons why you would not want to encode people's names in a protein. It is just too far away from the physical stimulus, there's no easy way for the encoder and decoder to coordinate on a shared meaning.
But many of these problems are also true for synaptic memory. I don't think cortex or hippocampus transplants would transfer memories either.
The vegetarianism thing might work though? The transplant cells are not used to getting meat bits, and when they get meat bits, they tell the brain, somehow, "hey, knock it off with the meat bits".
I don't know what you mean by cortex or hippocampus transplants. It seems to me that a full-brain transplant would trivially transfer memory.
The reason I don't think these problems are true for synaptic memory is that the brain is already busy building a complex model of the world, and so if you want to encode the concept "my name is Tim", all you need to do is connect pre-existing brain regions holding the concept of self, name, Tim, etc.
I think you guys are thinking too locally. In a sense, evolution stores mechanisms for relevant "memory" of adaptive situations. As we evolve, we get better at linking modules that reduce uncertainty. That uncertainty reduction has to built on a scaffold of systems that more simply reduce uncertainty. But why would you assume the "direction" is up instead of complex reverberations throughout, information possibilities stored in a manner that is somehwhat "black bbox" to us over the scale of the organism. We think of our thoughts and emotions as the highest projections of that, one's we directly experience. But the entire system stays in dynamic interaction with the environment, past and future. More basic markov bound modules are part of the structural foundation of the higher dimensional more sophisticated stuff.
But just like getting healthy changes your feeling and perception, other changes have upstream effects, which lead to downstream effects like epigentic changes. The whole system is constantly in flux regulating an organism in dynamic real time.
And a heart might be a big enough chunk of such a system that it can interact enough with the larger system to make itself heard in a way that is coherent.
The one relevant thing common to everyone is the way nerve impulses work. So perhaps the idea of transferred memories goes from impossible to remotely conceivable if these memories are based on a sequence of sensory nerve signals, i.e. not the letters T, I, M encoded somehow, but the "waveform" (for want of a better word) transmitted by nerves in the ear of the sound of hearing someone say "Hi Tim" or "My name is Tim".
Similarly, I imagine someone drowning would experience a sequence of distressing nerve impulses that might be rapidly encoded somehow in non-neural cells and be characteristic enough that they would be vaguely recognised by another brain as a disquieting recollection of an incident involving water!
As for the encoding mechanism itself, if it involves a cluster of connected cells all you really need is for each cell to be able to record a nerve firing pattern at a given time, and a link to another cell which can record another a short while later, so the whole ensemble is like a linked list or circular tape.
The transplant heart doesn't remember the concept of the word "Tim", but it has some kind of somatic memory of spending a lifetime in a body that responds in a particular way to the physical stimulus of the spoken word "Tim".
This is more plausible to me. I was thinking not so much of codes but just of chemical reactions in the body; somatic information if you like
It made me think of that species of wasp that plants its eggs into spiders, and the spiders start weaving webs that are not useful for their purposes but perfectly useful for holding the chrysalis of the wasp when it’s ready to emerge from the body of the spider which is now completely consumed from the inside . There is a chemical invasion that takes place that alters the behaviour of the spider. One could view a transplanted organ as a chemical invasion.
In the same way certain pattern of electrical charges inside processor represents M or N. But if we are talking about language model on a single cell it probably won’t have single letter representation since our llms don’t use them
Yes, encoding is the real problem. It would need to be something that could diffuse throughout the body. IIUC, extracellular RNA tends to be rapidly digested. (Also, ingested food is reduced to quite simple forms during digestion. We aren't planaria.) Perhaps carbohydrate tagging (of proteins?) would be more plausible. But it would need to be something that could penetrate a cell membranes.
Agreed. It's very hard to imagine a way that, say, a cardiac cell could interface with the outside environment with any specificity. Maybe at the level of "is my organism in a rich environment, or a poor environment? High threat, or low threat?"
That's memory of a kind, but not what people have in mind here.
But the cell alone might not be the vehicle. A heart is connected to the central nervous system and could be participating in all sorts of ways, including meeting the host CNS halfway, translating its own memories, etc. Both sides might have enough intelligence that they can work out the encoding differences, the way your brain adapts after healing from an injury.
re the mother part: even if the heart transplant version worked, it would make sense that it doesn't work at birth, because you would need the model developed by being alive for a while to actually translate those memories into anything useful. it doesn't help to have a signal that resonates with the name "tim" if you don't know how to identify discrete sounds out of the signals from your ears yet.
(on the other hand, hell, maybe this what people thinking they remember their past lives is--oh you thought you drowned and were reincarnated? well maybe your mom had a traumatic experience in a swimming pool once)
If the memory-creating molecules persist in the body for years, shouldn't they cause you to have some of your mother's memories once your brain has developed a sufficient world model?
Children are very similar to their parents in a variety of ways. Perhaps part of that is not only due to genetics and nurture, but also to inherited aversions/affinities (and other very simple sorts of memories) via cells transferred during gestation and breastfeeding. And the reverse: some fetal-originated cells persist and circulate in the mother as well even after the baby is born.
Maybe there's more going on than just molecules in the case of organs. They're big self-contained systems that have some neurons and connect with the host CNS. A mother doesn't transplant a large pre-built system to a fetus.
I feel like we would have noticed the similarity between these past life memories and the mother's memories by now though. Mothers are the most likely people to hear these stories in the first place. As far as I know this isn't a commonly known correlation.
If the encoding of memories depends on the development of the brain, that naturally makes different peoples' brains encode information in incompatible ways. If there were enough of a biologically fixed association between chemical signals and specific concepts that the transplant or cannibalism things worked, I'd expect these concepts to already exist in the brains of babies.
If memory transfer is possible, why do you need to develop your own model by being alive instead of just getting your mother's when (or before) you're born?
Babies are born with tons of instincts. It's easier to notice in other animals, since they are born less prematurely and with more motor skills than baby humans. But even newborn humans know to crawl to the dark circles of the nipples and suck on them, to stay close to Mom's familiar scent and voice, to cry when alone, afraid, uncomfortable, etc.
Perhaps some babies differ in much more subtle instincts based on the cells they received from Mom during gestation and breastfeeding.
Obviously those instinctual aversions and affinities are not all that complex humans can do, plus the kid might encounter a different environment in life and need to adapt to it, so it's still worth developing one's own memories in addition.
I thought the review was fairly clear that a complex memory like "my name is Tim" would be out of scope for the proposed non-synaptic memory, and nobody is saying that the ceullular system would completely replace SPM. The definition from the review is clear in that non-synaptic process would be an addition to, not replacement of, the synaptic system, because the synaptic system clearly exists and works to a great degree:
>The formation, consolidation, and retrieval of learning and memory in biological systems often involves stimulus-dependent, non-synaptic molecular and intracellular processes. These processes do not just serve synaptic-weight-based mechanisms, but provide complementary mechanisms. They are necessary for making and keeping long-term memories, but not sufficient, and interact with synaptic-weight-based mechanisms in nontrivial ways.
A single cell organism doesn't have or need a concept of names, so its memory system has no reason to evolve that kind of capability. As organisms grow more complex, they need more complex memories, and thus more capable memory systems. As long as the old system isn't actively harmful, it sticks around because evolution is a hoarder. It keeps doing whatever it's been doing forever and, biochemistry being as baroque as it is, the old systems interacting in useful and significant ways with the new wouldn't surprise me at all.
As you say, there are some fairly basic objections to forming complex memories through transplants, let alone food, and the review might be overplaying that angle in the introduction. But the basic idea of non-synaptic memory has been well argued, I suppose (speaking as an interested reader).
I didn't think it was clear - the "name is Tim" example is taken directly from the examples of things people anomalously "remembered" after organ transplants. Someone got a organ from a person named Tim and started having dreams about a person named Tim. I thought the author was attributing this to non-synaptic memory.
Maybe it was just my perception then. I considered the intro to be a light joke, a rhetorical device with just enough facts to be enticing rather than silly . I'm pretty sure you've used that device yourself many times in your blog posts, nothing wrong with that. I don't dispute the veracity of those anecdotes either but I feel confident in dismissing them as coincidence if they really happened as described, because what about the millions of recipients you don't hear about that also blurt out random things after waking up from anesthesia? Maybe receiving a transplant does funny things to your biochemistry in general, including switching you between gay and straight in rare cases?
When the review gets from the fancy to the factual parts, however, it argues from experiments with light stimuli and RNA transplants. That seems much more plausible to have an actual mechanism behind it, and that's the basis of how I rated my review of the review, not Power Rangers and dreams.
I think this is a fair reading, but it wasn't clear to me at all. Most of the reasons this idea might be a Very Exciting Big Deal depend on some of the silliness being serious. Otherwise this is just biology as usual. Which TBF can be very exciting in its own right!
These seem to be long-time changes, not just random stuff blurted out when waking up from anesthesia.
We also probably suffer from not enough research here. Only people who (experienced a personality change large enough + were fairly extroverted and open to talk about it + their surgeons were willing to listen to them) are currently recorded. These are three big filters.
"Gay" women end up attracted to and marrying men all the time. It happens so frequently it's not notable in the slightest. Now if it was a previously gay man who suddenly found himself attracted to women for the first time in his life, THAT would be notable. For the woman it's just the norm.
I agree but think the even bigger point is that the reviewer is defining "memory" too loosely to be of interest to me. If I can't recall it then it's not what I call a 'memory'. The only evidence presented towards recallable memory is the transplant evidence, which the reviewer themselves admits is no weirder than many other things we don't believe in.
Still, I love it when people embrace odd theories like this and the gut microbiome one you 'contra'd earlier this year. Aesthetically and emotionally, I want people to go down oddball science interest trajectories and stand in the face of conventional wisdom or even common sense. (Just don't put them in charge of our health system next time, please!)
Neurons themselves are nothing but chemical/biological processes. Is it really weird that some alternative combination of chemical/biological interactions could serve the same purpose ?
At any given point in time, a neural network embodies a particular mathematical algorithm, and any given algorithm can presumably be instantiated in many different ways.
I don't think it's any more intuitively obvious a priori that a memory can be stored in the form of a neural network than in some other form. The obviousness that storing memories in the form of a neural network is possible mostly comes from the confirmation we got from modern AI, and it seems to me that dismissing other mechanisms as much more unlikely a priori is hindsight bias.
Re: "we don't get our mother's memories", uhm, yes we do ? How did you learn to breath ? It was thanks to your mother's and father's DNA, which is not a neural network.
Separately, I think much of the presented evidence in the review is rather weak. The experiments about grinding worms into dust and the like would be big if true, but they seem rather untrustworthy at first glance. But a weak form of the proposed theory is obviously true (re:DNA, individual cells), and a somewhat stronger version does not seem a priori very implausible to me.
"Neurons themselves are nothing but chemical/biological processes. Is it really weird that some alternative combination of chemical/biological interactions could serve the same purpose"
My point is that the brain already contains a model of the world - it's the one we use to think. So all you need to do to encode a memory is to connect things in this pre-existing model. Since you already have a concept of "name", "self", "T", "I" and "M", you can connect those brain cells to form "my name is Tim". Protein molecules don't naturally have any of these concepts, so it's hard.
We have instincts encoded in our DNA, but we don't inherit specific memories from our parents like their names, their biographies, or the book learning that they've absorbed.
Trying to avoid the words 'instinct' and 'memory' for now :
Our parents transmit to us their knowledge of how to breath, and our brain is then able to retrieve and use this information to inform its behavior. This is an existence proof that it is possible to usefully store and retrieve information with a mechanism that has nothing to do with synaptic weights. It so happens that the DNA mechanism is used to carry information that pertains mostly to our subconscious, but that doesn't mean there couldn't be yet another non-synaptic mechanism used by our consciousness to store and retrieve relevant information in daily life.
Having a concept of 'name' or 'TIM' ultimately cashes out in some physical state or process, whether it's in a brain or a LLM or DNA or something else. I agree that the fact we already know these concepts exist in the brain gives a penalty to believing that they were also reproduced in some other medium, but not a huge penalty. I disagree that it is a priori more natural to have these concepts exist as a result of some synaptic weights than in some other form. (This is in answer to you saying 'Protein molecules don't naturally have any of these concepts'. Neither does the brain, or a pre-training LLM)
But neurons don’t have such concepts too? If we oversimplify neurons to bunch of weights or look at LLMs we encounter same problem - blank or random state doesn’t represent anything outside. Electrical impulse isn’t anymore “T” then hypothetical protein signal, fact that it represents T is external encoding
"a neural network embodies a particular mathematical algorithm"
No. Any physical configuration could correspond to an infinite number of mathematical operations--the so-called q-addition which is a slight modification of regular addition. Given any physical machine implementing ordinary addition, you could, by suitably reinterpreting its internal states, say that it is in fact implementing q-addition instead.
Scott: I don't recall you ever posting a top-of-thread opinion on any of the previous reviews (please correct me if I'm wrong). But by saying you don't believe it, you're influencing your readership, and it seems like you're putting your finger on the scales of this contest. It's your blog to do with as you wish, but to be fair to the OP, you should have let the commentariat post their own opinions and only responded as necessary. Just saying this out a sense of fairness to the OP.
I’m not sure if the fairness of the contest is so important that we want to be sacrificing potential good discussions just to preserve it. (Easy for me to say, I’m not a contestant.)
I enjoyed Scott throwing in a large dose of skepticism as I found the writing compelling. But I agree with you wholeheartedly, should this review have won, then would be a great time for Scott's heavy influence.
The examples are so extreme as to raise doubts, but one might still want to explore further, and relevant to the project of this site, I suggest that kidney donors and recipients volunteer for very extensive research into memory and physiological changes.
Joking aside, I think an important question is raised. All cells in the body are joined in the same project, so is there ways that this unity of being and purpose is updated that leave sophisticated intra-cellular traces yet to be identified. And could that affect the quality of our consciousness and our understanding of what it means to be alive and human.
Even if memory can be partly encoded in bodily areas outside the brain, there's yet another problem in transferring it to a donor: It seems highly unlikely that everyone's encoding would work with the same interchangeable protocol, any more than everyone's (brain) memory is likely to be encoded in exactly the same pattern of neurone connections (or however it works).
Drifting slightly off-topic, but on the subject of cannibalism, earlier this year there was a report of bones dated around five thousand years ago being found in the UK somewhere, and these had cut marks and bite marks. But I think the researchers' conclusions that these were evidence of cannibalism may be somewhat hasty.
It was around three thousand BC when farming first took off in the British Isles, and although one tends to think of farmers as being fairly meek and mild compared with warriors, those first farmers must have faced a massive threat from the majority population of hunter gatherers around then. Consequently the farmers were most likely the very opposite of meek. For a parallel, think cocaine growers in a South American jungle. Trespassers are unlikely to be given a pat on the head and released unharmed!
These early farmers probably hired muscle to guard their farms, and kept large fierce dogs to help. Anyone caught trying to raid a farm would have faced a hideous death, to discourage other would be robbers. Quite likely they were dealt with similar to how the Chinese punished pirates: These were tied to a pole on the shore, and had chunks of their flesh cut off to the bone, with torniquets applied to prevent bleeding to death too fast, and the flesh would be thrown to dogs or pigs to be eaten in sight of the victims. I suspect that is a more likely explanation for the cut and gnawed bones.
It seems unlikely to me that a brain has much of a standard encoding scheme, especially one too complex to be interpreted with some work.
Doubly so for the rest of the body, if memory is stored there; memories there would keep very well, and take a long time to recall anyways, so there isn't that much harm in them taking minutes to interpret, or even weeks.
Sarah Constantin asserts here that we know that at least Purkinje cells in humans can encode information by themselves: https://sarahconstantin.substack.com/p/what-does-the-cerebellum-do-anyway (ctrl+f "Even single cells can exhibit learning.") So if she's to be trusted, then at least a mild form of the "not all learning and memory is synapses" hypothesis has to be true. If you disagree with her I'd love to hear it.
LLMs famously have bad memory. And we know that electroshock therapy tends to cause amnesia of the last few weeks, but not anything else. Maybe the last few weeks of memory are present in synaptic connections, but long term memory is somewhere else?
The organ transplant stories sound like they might share similar problems with past life stories- they're prone to suggestion, we don't hear about the null results, and the evidence base could be really bad. Perhaps if any of the stories are actually true, they represent information overheard by the unconscious mind in the operating room ("please bring me Tim's heart").
...But if I wanted to steelman them: suppose that there are long-term storage cells in certain abdominal organs that encode information like Purkinje cells; they could then transmit the information through the nervous system, like any other signal, rather than via some kind of bloodstream chemical. Maybe the cells don't always survive the death of their original body and the subsequent transplantation, or they fail to re-integrate or are overwritten by the new host's nervous system, but when they aren't, they transmit memories!
Edit: as for why no maternal memory: maternal->fetal microchimerism does happen, including of nervous system cells, but it apparently happens in very small amounts. If that kind of information can't be compressed any further than whole cell configurations, and the body already tends to discourage somatic cell sharing for evolutionary reasons, then maybe it just didn't get a chance to evolve.
The most sensible way I can think of to store such memories is
1) Each neuron picks a random sequence of 20 base pairs (or similar) and generates lots of RNA.
2) When 2 nearby neurons fire, the RNA leaks across, and it gets attached to the RNA produced by nearby neurons.
So each individual strand of RNA just store "Neuron id 7fc35 is next to neuron 4ga1802", but given a whole load of these little strands, you can reconstruct a the entire network structure.
If flatworm brains are all basically identical, but each individual human has their own pattern of neurons and tags, then this would predict that memories could be shared between flatworms, but not humans.
And passing down memories is tricky. There would be a significant risk of forming some distorted nonsense memory. Something useless, but particularly easy to pass down.
You can just make “foreign” RNA concentration representation of connection strength between two cells and you can make cells normally grow in a way to keep concentrations constant. It can be reparation mechanism theoretically capable to restore missing neuron using weights from its neighbours
But to actual transfer knowledge even between worms you will need signatures to be exact same not random
I agree the review goes way too far (the parts about eating memories, or storing them in hearts and arms have to be a joke, I assume they were just being used as a hook to get the review to stand out).
But I don't think the review is disputing that groups of neurons store memories in combination, or that neurons communicate and function primarily through synapses. It's only disputing that long term memories are physically stored in the synapses themselves.
Having said that, although I was initially convinced, I have been unconvinced after further reading. The biggest piece of evidence was that human synapses turn over in weeks to months and so cannot store memories on the order of years. But that is only half true - the most stable dendritic spines last many years. In mice 75% of these types of spines were stable over 1.5 years. There is scarce data on primates but one study showed that the most stable spines in the macaque visual cortex are twice as stable as rodents (with 3% turnover in months.
The second big argument was that neurons can restore synaptic weights that have been erased. Except that these were sea slugs which have a very simple neural structure compared to humans. It's plausible cells might have some ability to do spacial coordination. For instance a cell in the V1 cortex might "know" which direction the V2 cortex is and if you erase it's synapses it might be able to restore them in roughly the right place but not perfectly.
Thoughts on ways to potentially rescue this hypothesis:
1. Neurons don't have any inherent meaning either -- the meaning of neurons is just, like, some sort of network learning thing. (Well, mostly -- obviously you've got sense data to ground it, and also genes can *somehow* control things finely enough to result in some fairly specific instincts. But a lot of it is this sort of network learning thing.) So the same could be true of non-neuronal learning. Of course, that wouldn't allow for the sort of transfer that's posited here, via transplant or ingestion. But that stuff seems (in humans a least) pretty tenuous anyway. Although...
2. Perhaps the cannibalism thing could nonetheless still work *in flatworms* or other such comparatively simple creatures, where the network is shallower and more fixed (consider C. elegans with its famously fixed connectome), so that meaning might be retained across distinct individuals of the same species in such cases, though not in mammals?
It's tangential, but I do have to say, the whole "cellular learning is real, evolution doesn't create things from scratch" line of thinking is pretty suggestive of *something* going on at the individual cellular level, at least. And like, these sorts of lower-level neuron-based workings of the brain (synaptic or otherwise) are certainly helpful for understanding some mental phenomena that seem higher-level -- consider semantic satiation! Sure seems like a lot of what's going on there is "your brain is made of neurons"...
> in order for the organ transplants to make sense, you need for the heart to be producing or storing these chemicals (why?) in such large quantities that they get into the bloodstream, cross the BBB (why would you have a transporter for these?), make it into the brain, and perfectly interface with the host's own memory-decoder system (how?). Then either the heart has to keep making these constantly (why?) or there has to be some mechanism for a single low-dose exposure to get the host brain to start making them itself (why?)
I think you're being too restrictive in how heart-to-brain data transfer could conceivably work. Heart/lungs can certainly send impressions like "excess fluid in bronchial tubes, please cough ASAP" by normal use of sensory nerves, and that channel doesn't just go away outside of emergency conditions. Sensory impressions have a similar enough format to emotional memories for somatization to be a thing: "my heart skipped a beat" as a description of an intense emotional reaction suggests the brain can send such information to the heart quickly, for which stress hormones and so on provide plausible mechanisms.
If I were designing a memory-backup function into the heart, and wasn't allowed to send any chemicals from the heart back through the blood-brain barrier at all, or even add side channels to sensory nerves, next obvious thing to try would be modulating the heartbeat itself to serve as a modem. Relatively high-bandwidth brain-to-heart downloads, then low-bandwidth, meticulously compressed "error correction" responses to keep the archives synchronized, particularly while dreaming. Sleep paralysis means actual blood flow demands are low and predictable, so almost the entire potential range between resting and maximum heart rate can be used to encode information.
Peer-to-peer music piracy is proof of concept for synchronizing large datasets through noisy, intermittent signals. Any good compression algorithm means actual signal would look almost exactly like random noise to someone who doesn't already know the format.
> His name is Tim—I think it’s Tim Leighton, but I’m not sure. I think of him as Tim L.
> She inhaled him. That’s almost a bit too on the nose. She didn’t know it at the time, but her donor’s name was Tim Lamirande. Spooky, right?
Tim Lamirande wakes up after surgery in an unfamiliar ribcage. There are some pipes, and the tools he's expected to use to maintain them, and an eccentric morse-code dictionary (heavy on physiology-related emoji, very light on abstract stuff like math), but not much else he can meaningfully interact with. He thinks he can hear someone else moving around in the next cell over, decides to try communicating by tapping on the pipes. Closest concept in the dictionary to "stuck here together" is "kissing," and closest to "outside this place" is "fresh air source," but he cobbles together a semi-coherent metaphorical explanation of the overall situation, https://bogleech.com/awfulhospital/795 and there's an excellent vocabulary for nuances of self-image and emotive intent, so that part's easy.
Then he tries spelling his name. There are codes for the Latin alphabet, but they seem to have been added as an afterthought, once all the short, easy code sequences were already spoken for. He manages less than one such letter per hour, back-shift during which they can freely communicate (without being distracted by work) only lasts about eight hours, and if he doesn't finish an unfamiliar word in one pass, neighbor gets the internal sequence all garbled, so... "Tim L." Sure, that's close enough.
With much respect, I think you’re missing the point a bit here Scott. I don’t think the author’s main argument is about memories of specific events or facts. They are taking a wider (and more useful, I believe) definition of memory as the state of a system that influences behaviour. That can include aspects of memory encoded in chemistry and biology. You might think of it as “vibe” memory rather than associative memory between concepts.
I agree if wouldn’t be useful to store associative memories like you describe in the heart because the brain is the organ that uses these types of memories in processing. However, processes like biological responses to stimuli e.g. fat storage when getting lots of calories , are processed in a distrusted way to some extent, and it would make sense for there to be state (memories) encoded throughout the body to manage that.
We do indeed see this kind of state (memory) passed from mother to child.
I think that one of the main takeaways of this article is that memory can be much more complex and diverse than things like episodic or associative memory.
The review is very nicely written and raises interesting points. But I want to severely push back on one point. The neuroscientific community is quite misrepresented here.
The author claims that the consensus is the strong SPM hypothesis. But no one believes this. No one. At least not among the neuroscientists who are actually biologists who work with individuals neurons. And I have talked to some real experts in the field, and (as a bit of an outsider) I have done academic research in the field. Not working with actual neurons, but building models of spiking neural systems that were supposed to be pretty close to biology.
Let me try a more accurate description. The human brain is pretty complicated. In order to understand it, we need to look at it at several levels of abstraction.
A) At the lowest levels, we try to understand the chemistry of how a single spike is formed.
B) On slightly higher levels, we may investigate how the input form other neuron excites another, without modelling each single spike on a chemical level.
C) Even higher, how patterns of spiking look like in an ensemble of neurons.
And so on, until we are at regions of the brains like the prefrontal cortex, or even higher up at behavioral level.
By the way, all this is not even covering yet learning, it is just describing a snapshot of the system.
If you are at level B, how a neuron excites another: you can't work with weights on that level. Weights will not explain what's going on. Everything is totally non-linear. When two incoming spikes arrive at the same time at the same dendritic branch, then the effect is much larger than the sum of the two. Especially if the one further from the soma (the blob at the center of the neuron) comes a little before the other. Timing matters a lot. Some people say that neurons should not be described as "integrate-and-fire" (adding up the inputs and fire when they are above a threshold), but rather as "coincidence detectors" (they fire when they get two simultaneous input, screw their weights). Everything is awfully complicated.
But then you want to abstract this whole mess. Because the higher level of abstractions are also important, and you need them.
What is the best abstraction of a neuron and its synapses? Perhaps first you only keep like 10 parameters per neuron and 2-3 per synapse. If you want to make it even simpler, then you keep 3-4 per neuron, and perhaps 1 per synapse. But what if you want to go even more abstract, what do you keep?
And then the consensus probably is: if you really force me to reduce everything to a single number, then probably a "weight" per synapse is our best bet. But it's a simplification.
And now, when it comes to learning: the strong SPM hypothesis claims that only weights change. On higher levels, this is a useful abstraction. But on abstraction level B it is so obviously wrong. Of course the neuron is plastic and changes! All of the time, and all parts of it, not just the synapses. The soma changes, the ion pumps change, the genetic expressions change, the timing of the incoming spikes change. Those are things that we have abstracted away at the higher levels, but they are there. And this does not even touch the formation of new synapses, the growth of the dendrites into new directions, the formation of new neurons. For what we know, even the non-neurons in the brain, like glia cells, may change and play a role. This is why I say that no one believes the strong SPM hypothesis.
Although the author did describe the strong SPM hypothesis as "it's all in the weights", I sorta took their meaning to be more like "it's all in the neurons", where perhaps weights was an (intentional or not) simplification of what that looks like. Would it change your stance if they had de-emphasized the "weights" part?
I could subscribe a lot more to "it's all in the neurons", yes. I would add some caveats that there are other surrounding cells that may play a role, but yes, "it's all in the neurons" is a standard hypothesis. But I don't think the author of the review means "it's all in the neurons" with the strong SPM hypothesis. Right after introducing the strong SPM hypothesis, they are pretty explicit about it:
"The crux of my negative review of the SPM hypothesis is this: cells are extraordinarily complex molecular machines, and there’s a lot going on inside of them that the SPM hypothesis implicitly neglects. We often abstract away most of the biophysical complexity of neurons, which are cells. As cells, they take up physical space, and can have weird, complicated shapes. They talk to other (not necessarily neural) cells. Each individual neuron has a complicated (gene regulatory) network inside it, whose complexity parallels that of many of our models of entire neural circuits. Do we really think that none of this complexity is involved in processes as complicated and multiscale as learning and memory?"
I mean, I agree, they clearly literally meant weights. But if I had to guess they probably didn't know that much about the details and were using a gloss they had heard elsewhere. Probably they now wish they had been more accurate now, because it wasn't that important to the point anyway.
This feels a bit uncharitable. The weight of a spike must be grounded in some molecular changes, and those molecular changes most proximally related to the concept of a weight will have their own supporting dynamics further down the chemical pipeline. The point of saying "its all in the weights" is that the relevant cellular changes for memory are that which modulate the sending/receiving signal strength. This doesn't render the cell irrelevant, but isolates the relevant cellular processes as it relates to memory formation to those functionally upstream of the weight abstraction. But in principle this could relate to a very broad spectrum of cellular processes.
FWIW, as a practicing neuroscientist, I definitely think people believe in the strong SPM hypothesis. Nonlinear integration of EPSPs tells us that simple integrate-and-fire models are wrong, and it's better to think of e.g. single dendritic branches as analogous to units in ANNs. And it's of course true that ion channels, receptor densities, etc. all vary as a result of experience, and this is orthodox neuroscience. But I still think it's coherent to claim that this is all *in the service of* changing synaptic efficacy, and in that sense consistent with the SPM. I think something close to this.
I don't know much about neuroscience but when reading the review I already had a strong feeling the reviewer was misrepresenting how much what they were arguing against was actually a consensus. I came to the comments to ask about this, so I'm glad to see someone has already answered.
> Some people say that neurons should not be described as "integrate-and-fire" (adding up the inputs and fire when they are above a threshold), but rather as "coincidence detectors" (they fire when they get two simultaneous input, screw their weights).
A) I've always found it hard to imagine that mental states might be encoded as anything other than electromagnetic waves.
B) I'd hazard that the "broken rock" definition of memory isn't specific enough. I think the most reasonable, broadest definition of memory would boil down to "encoding *correlation*" (as opposed to just "information persistence") . Information is only useful to the decision-making of an organism insofar as it draws correlations between events. (A while ago, I wrote about how the Correspondence Theory of Truth and the Coherence Theory of Truth are, in fact, complementary theories of knowledge [0]. Correlations would broadly fall under Coherence.) E.g. I can easily imagine ancient microbes inventing molecular ways to encode correlations between being poked and cellular damage, and then using this information to "decide" how to react.
----
Thus, if the most basic type of memory (in the broadest sense) consists of encoding correlation, i wouldn't be surprised to learn (from researchers, with 100% certainty) that neurons (at the synaptic level) were doing some sort of "Fourier-Transform-esque", "music of the body" frequency-matching in order to represent correlation-detection of incoming signals. Feels like a very natural building-block for mental activity, to me.
> How can we remember things for years with synapses that turn over on a time scale of weeks or less?
We don't understand it very well, but this question misses an important point: SOME synapses change very fast, in the order of weeks or less. Especially in regions like the hippocampus where long-term memories are NOT stored.
I don't think we know very much about the synapses that don't change. This is because they are hard to measure. You can manipulate animals so that changing synapses light up. But it's harder to make non-changing synapses light up. Also, it's easy to look into a mouse (uhm, if you are a top-notch neuroscientist) and then look into the same mouse a week later and look for changes. But it's much more expensive, and probably also harder, to make the same experiment with a year time lag instead of a week.
I think the standard model is that indeed some synapses change very fast. Fastest is hippocampus, where many things (not everything) changes within days, in some other regions it is more a matter of weeks or months, and probably some regions like the prefrontal cortex contain subsets of neurons and synapses that don't change in years or decades. This is where our long-term memory sits. Now, this model has not much experimental backup, because it's hard to test experimentally, but it doesn't have a lot of evidence against it either.
UPDATE: In a comment below, Ariel Zeleznikow-Johnston points to some experiments which HAVE followed synpases over months and found stable subsets.
Until you can test it experimentally, it's all a just-so story. Materialists seem very attached to their just-so stories, and they argue about them as if they were valid explanations. The biggest elephant in the room is the assumption that the brain generates the mind.
This is a super important comment, thanks for sharing. A bit more context: when people hear "hippocampus" (HC) they think long-term memory, but the dominant model in the field is that memories are transferred from HC to cortex over a time frame of weeks, matching the empirical dendritic spine turnover results in both areas. That's why HC patients like HM were incapable of forming new memories while retaining their old (pre-operation) ones.
One nitpick: Connectionism is stronger than the claim that "all of the things that make us human—our ability to talk, think, reason, remember, and so on—follow from networks of interacting neurons, and changes in the strengths of connections between those neurons." Classical / symbolic cognitive theories (typically viewed as the competitors to connectionist theories) also claim that all cognitive functions are implemented via neurons in the brain. The difference is that symbolic cognitive theories view those neurons as implementing symbolic computation (like a computer, where, e.g., logic gates are implemented via transistors), while connectionist theories view cognitive functions as directly resulting from sub-symbolic, distributed, associative patterns across those neurons (like an LLM). See https://plato.stanford.edu/entries/connectionism/.
I know this is not central to the post, but seemed worth pointing out. (Source: I'm a PhD in cognitive science.)
If anyone wants a quick rating on the kookiness scale from a licensed neuroscientist*: the basic biological facts reported are not at all kooky, but the speculations and extrapolations are extremely kooky.
It is not at all kooky to say that single cells have "memories" of some kind, and that these are important, but not well-understood. This is universally recognized in neuroscience and not remotely controversial.
But these are cell-memories of cellular things, not people-memories of people things. It is extremely kooky to suggest that that single-cell memories could be responsible for the reported personality changes in transplant patients.
Kooky things are not always wrong. But they are kooky!
*Not a memory or cellular specialist; also not actually licensed.
I found it strange the review didn’t mention that hearts have ~40,000 neurons. What do you think of the claim they could carry some basic memories over? I’d link to a journal article, but it cites precisely the same examples as the review.
What’s your take on the epigenetic “memories”, like rats retaining aversions that their parents learned? There’s lots of skepticism because the mechanism is unclear, I’ve heard, and because many have taken the implications too far. (It gets pretty woo.) But could low-level predispositions be transferred, even if not declarative memories?
Honest questions: I find the subject interesting and don’t know what to think.
I'm not sure that the presence of neurons in the heart makes any difference. Neural activity only acquires "meaning" by dint of relationships with other cells. Brain cells are closely and recurrently connected with sensory and motor cells, so they are well-positioned to acquire "meaning" related to the outside world. Cardiac cells, neurons or otherwise, are mostly connected to other cardiac cells, so I'd guess whatever "meanings" they encode are mostly heart-related.
This doesn’t seem quite right. I might as well share the article I looked at (link below), because it references other apparent not-purely-heart-related functions of neurons in the heart, such as in emotions, and regulating the overall mood/ state of arousal of the body. Most of the “memories” transplant recipients report are not declarative memories, but flashes of light, tactile sensations, and other basic feelings with high emotional salience. Does it seem too crazy that the heart could retain the residue of memories that are salient for arousing the heart, either because that helps the heart react in the future, or because the brain is messy and doesn’t purely separate its functions? (The “Tim” and power ranger examples are more declarative however).
For reference, zebra fish have 100,000 neurons and can form long term memories.
1) I don't think the comparison to zebrafish means very much. 40k neurons is a lot for a zebrafish but not much for a human. The heart may well need most of them just to coordinate cardiac muscle contractions.
2) In biology, everything is connected to everything by one route or another. Arousal and emotion are global states that can probably be "read off" every organ of the body in some way, but this doesn't tell you much about the function of any of those organs.
3) Organ transplants are a big deal physiologically and psychologically, so it makes sense that lots of weird stuff happens. The particular weird stuff may well have something to do with the particular history of the organ.
4) The truism goes, in biology the answer is always yes. I would add: yes, but in a boring way.
“Yes but in a boring way” is biology’s version of “it’s never aliens” in astrology 😂 Ok, these are reasonable points. I will hold my mind open to emotional processing in the heart to manage the body’s macro states (especially since 80% of neurons from the heart signal up to the brain rather than vice versa) - but with cautious skepticism.
But here's a possible "just so" story about low-level dispositions getting transferred. Say Bob is an anxious sort of person with a highly reactive autonomous nervous system. His cardiac cells have their gain ramped way up; it's ready to palpitate at a moment's notice.
Jill somehow gets Bob's heart in her chest. Before she was an even-keeled sort, but now she finds that her heart starts racing much more readily. Because the relationship between emotion and physiology is bi-directional, she gets more anxious. Now her personality is more like Bob's.
This sort of "memory" transmission seems plausible, but purely speculative. The epigenetic memory stuff I haven't followed closely; but I will say, if it were true, you'd think it would be a big thing in the field, and it has distinctly not been a big thing. My guess is that it hasn't replicated, but that's just a guess.
Yes, this is the main other scenario. I feel like the academic job market is brutal enough that people would roll the dice on an iffy area if it gave them a shot at a Nature paper, but I dunno!
Ok, while there was one retraction in this field of research, it seems it’s still active and not crazy, according to this recent paper that explores the mechanisms (below). This article gives a plausible-sounding account of what kind of memory we should expect via heritable epigenetics: “This process is akin to priming or acclimation in plants in response to various stresses like drought, heat, salt, irradiation, and pathogens.” (see section: “Epigenetic transcriptional memory.”)
So here’s a third reason you haven’t heard about it: it’s likely true but in a boring way (for newspapers I mean. For me it’s really interesting).
Reading the post I actually thought they were going to go with the theory that the heart neurons might transfer memories to the rest of the body's neural network just through normal predictive coding interactions, as something similar happens with llms where the behavior of one llm on totally separate questions can change just by being fine tuned on totally unrelated outputs of another llm with the specific behavior. See this great video by welch labs to learn more about the topic:
I think this analogy is useful in that it changes are likely to be weird and unpredictable. Something may change, but it's probably not the "transfer" of a coherent or even well-defined memory.
There's actually an interesting recent paper on the topic of cellular memory: Kukushkin et al., "The massed-spaced learning effect in non-neural human cells" (not sure if I can link it here), which basically showed that “spacing effect” in learning (training works better when it's broken into several sessions over time) emerges in non-neural human cells, like kidney cells. If I understand correctly, they used human cell lines and found that four brief, 10-minute-spaced "pulses" of a chemical signal (forskolin in this case) drove stronger and longer-lasting gene transcription than a single massed pulse, i.e., a cellular analogue of better memory with spacing. I've see some interviews with the lead author of this paper, and he argues, on this basis, that memory fundamentally depends on cellular machinery common to all cells.
I love this line of inquiry if only for all the weird hypotheses it generates
* does eating meat from free-range animals cause you to absorb some of their... wildness? is that why some people like it a lot? does our mass-produced meat industry manage to produce meat that is missing some vital nutrient of "experience"? does eating factory-farmed meat lead to something like anxiety because you vaguely remember the horror of it?
* do blood transfusions confer some of someones's vitality to you? is this why some people are weirdly into them?
* is this why spiders eat their mates or their parents? to learn how to spin webs better?
* do babies absorb some of their mothers' memories through breast milk?
* does thinking that your memories live in your brain cause us to "ignore" our bodily memories, leading to some dysfunction? is this why bodywork-type therapies help, because they remind you to pay attention to the rest of your "mind"?
* is there some concrete experiential benefit to swallowing, um, semen?
no opinion on the veracity tho. just fun to speculate
Kooky hypotheses are fun! The problem is there's a massive "motte and bailey" element to this, and non-specialists might not pick up on the distinction.
Cells definitely have a form of "memory", but these hypotheses require a massive leap from "having memory of some sort" to "transmission of arbitrary information via transplantation or ingestion". These are super different things and the former is not evidence of the latter!
Context: I'm college-educated and work on software engineering that uses LLMs, so I'm vaguely familiar with concepts of synaptic weights from cultural absorption.
I got to section 4 of this review and realized ... "I hadn't thought much about it, but I had always thought that neuron strength, neuron pruning, and other qualities of the neuron itself might be important, alongside just the strength and existence of the synapse." ... Is that the distinction it's fighting against, the literal interpretation of SPM as only being about synapses? Colloquially I had understood it to be more then just synapses, even though I suppose it's in the name. Just providing a data point that some colloquial understanding never abandoned the ideas this review seeks to resurrect.
> Do we really think that none of this complexity is involved in processes as complicated and multiscale as learning and memory?
Natural selection would tend to eliminate brains that were too easily changed by so many processes. Consider how genes of large effect are disfavored relative to lots of genes of small effect, which better enable a target to be hit rather than overshot.
> People think that epigenetic changes like DNA methylation are responsible, since these changes can be fairly stable, and can be heritable
> Relatedly, it seems unlikely that no organism could learn or remember anything before synapses existed. There had to be some other, simpler mechanism that worked before large-scale, synapse-based neural networks were common.
Why must that have been so? Why is it impossible for there to have been organisms without memory, until the capacity for memory evolved later?
> Maybe your memory of your fifth birthday party is in your arm, or leg, or heart
If that were the case then people would LOSE memories on losing a limb (or getting a heart transplant).
> Can you forget something about your fifth birthday party if someone chops your arm off?
There are many people who have lost limbs. Is there any evidence of them losing memories as a result?
Would you realize if you had forgotten the memory? Maybe this does happen but people don’t notice… or it just diffuses around everywhere such that you have a redundant copy in the other arm?
Good point, though I do think that the broader diffusion point still stands. If this hypothesis were true I’d expect important long term memories to get scattered all over and then randomly re-loaded periodically into synaptic memory. So an individual arm wouldn’t be specifically tasked with sole source of truth for the birthday party (presumably read speed on such a system would also be terrible…). I suppose it would accidentally end up as sole source of truth for a few things by luck of the draw, but since those things weren’t currently loaded into your random access layer anyways plus the damage wouldn’t be progressively worsening /systematic I suspect it would be harder to detect (especially if no one around you sees any prior reason to check you for memory loss). That said, I doubt this is actually a thing. I just find the mother-child counter argument stronger than limb loss.
If natural selection eliminates changes of large effect rather than many small effects, it should eliminate synapse-dependent memory systems and favor highly distributed multiformat memory.
he's saying that the causal relationship between any single gene and its phenotypic expression is small. Not that the macroscopic instrument is resilient to damage.
Like a few others here (demost_, maybe others), I am a neuroscientist, I teach classes in learning and memory, and do research in learning and memory. I have been in this field for over 40 years (yes I am old) With that background I believe I can speak to how the synaptic plasticity (in general) and synaptic weight (specifically) hypothesis is viewed and taught in the field. It is a strawman to claim that there is a strict doctrine that synaptic weights are both necessary and sufficient to account for learning and memory that is widely believed and taught.
How does the field treat synaptic plasticity? It is indeed typical to teach about LTP and LTD, but only as one mechanism. It took many years for the field to become persuaded that LTP really was a major mechanism underlying memory, and most courses that have time to do more than just mention it do go into the evidence for LTP being a memory mechanism, as well as what we still don't know about it. In addition most courses acknowledge other interesting examples of plasticity, including white matter plasticity, epigenetics, network dynamics, and even molecular mechanisms! Researchers presenting talks in which they have focused on synaptic plasticity mechanisms in my experience typically acknowledge that there are alternative possible mechanisms as well. I haven't heard anyone in the field claiming that synaptic weights are necessary and sufficient for learning.
I also have not had the experience that AI researchers starting really pushing the idea of synaptic weights underlying learning and memory in the 2010s. The idea was around from much longer, including in the 1980s when I was an undergraduate and connectionism was becoming stronger. At the time there was always an acknowledgement that calling these connections synapses was just by analogy and was never intended to be biologically realistic. I even wrote a paper in undergrad about what we knew then about biological synaptic plasticity and implications of that research for connectionism. I still find that AI researchers who I interact with are quick to acknowledge the lack of biological grounding and the limitations of the synaptic weights view.
With that in mind, the review falls apart for me. The second half argues against a view that may be held in some non-scientific circles but is not an accurate characterization of the field, yet claims to be arguing against a dogma held in the field. The first half is way out on a limb -- many fascinating anecdotes that I did enjoy reading, but probably NOT the best cases to make for a possible molecular basis for memory given the complexity of some of the declarative memories described.
Hey CS, may I ask what are your favorite/seminal papers around these alternative examples of learning and memory? I'm not classically trained in neuroscience so I know of LTD/LTP but not around white matter plasticity, network dynamics, or molecular mechanisms.
Here are some papers that really made me think -- I don't promise that they are the best, just the ones that I read and thought about.
For network dynamics:
Arnsten, A. F. T., Paspalas, C. D., Gamo, N. J., Yang, Y., & Wang, M. (2010). Dynamic Network Connectivity: A new form of neuroplasticity. Trends Cogn Sci, 14, 365–375. https://doi.org/10.1016/j.tics.2010.05.003
I can easily believe some transplant stories while others seem totally fake.
An actual memory of the previous host of an organ? No way.
Something that could be cell-protein-based? Much more plausible. If a transplant stomach previously had a negative reaction / trained biome for a certain diet, that could carry over. Similarly, if a heart always raced when the body sensed meat before, it could again. I could absolutely understand some subconscious patterns being carried over.
Think about it another way - if you received lungs from a long distance runner, you are receiving those lungs with the “memory” of strong performance. Or - just more muscles in the way that ‘matter’. We convolute what ‘counts’ as memory, but I’m not surprised it appears sometimes in unintuitive ways - reacting to foods, other sensations. Feels plausible.
For those muscle growths or pathways to transmit some kind of memory to the brain… seems very unlikely. I could believe it for fruit flies that do not rely on cognition much anyway (more of light/smell seeking robot) but for humans, more likely to just be regurgitating facts they were told that then “feel right”. Hell, I have a memory of flying to Disneyland because my parents told me a story about it when I was a young child - they didn’t realize it was my brother in the story but it’s still in my long term memory. I’m sure a similar effect is outlining the more vivid stories.
That’s my thought too. Even things as disparate as sexual orientation I could believe to have some sort of cellular basis in the organ. But things like names, or episodic memories of the moment of death, seem very far-fetched, even when considering a broader range of mechanisms than pure brain neuron synaptic connections.
This seems especially true when it's not that difficult to believe that people might have been discussing the donor of the organs when the transplant patient was not conscious/aware enough to fully remember the discussion, but picked up part of it.
I could see some of those more vivid stories still coming from cellular memory. Someone drowned, and their heart learned to associate being underwater with bad times, so now the person has that reaction to water. Combine it with a dream on the topic, and I could easily see that physiological response getting jumbled in with memories.
""From a modern vantage point, it’s clear that if groups like the Tupinambá practiced cannibalism for the purpose of acquiring aspects of their enemies, they were misguided. Maybe they could get some protein by doing this, but they certainly couldn’t acquire any of the things that make us fundamentally human, courage and bravery included. ""
To state the obvious, they could by eating immediately acquire certain *negative* aspects of the enemy. i.e. any communicable diseases.
Any cannibalistic pre-medicinal culture would very quickly learn "don't eat the flesh of a diseased person"; then "actually, don't eat any sickly or weak person"; which naturally evolves into "prefer to eat the strong and mighty warriors".
It doesn't feel like any explanation involving non-synaptic cellular memory is needed.
The heart transplant stuff reminds me of something I read a while back that claimed that the heart has a surprisingly high number of neurons in it, and that the count is high enough that some scientists think that it may play a role in cognition, that the brain may not be the exclusive seat of thought as we've believed for a long time.
The heart transplant stories would appear to provide some evidence in support of that theory.
This review reminds me of the work of Eugene McCarthy. Much of the text defends a weak but reasonable hypothesis, "memory encoding is not exclusively performed via synapse weights" or "cross-species hybrids have a greater role in evolutionary history than conventionally thought", all to arrive at the conclusion that humans were descended from pig-chimp hybrids.
You're going to have to do way more work if you want to convince anyone that there is anything in need of explaining about the transplant stories. Half-hearted (ha) disclaimer about psi aside, the review takes for granted that there's something weird going on about the heart transplants, but paranormal forums are stuffed with equally "plausible" tales of reincarnation, ghost sightings, telepathy, and everything else, many of which rely on the same formula of unverifiable, irreproducible personal accounts of remarkable coincidences, accompanied by some suggestive hand-waving.
Some of the stories presented in this review aren't even impressive at face value. What is supposed to be the connection between the college professor's nightmares and the drug bust? I guess we're meant to imagine that being shot involves light?
Unlike the cannibalism, which I'm interpreting as a lighthearted (ha) framing device, the author appears to wholeheartedly (ha) believe in the phenomenon of memory-transferring organs. In absence of good reason to agree with them, this undermines the credibility of the rest of the text.
Another example of the author's unwarranted confidence: "How can we remember things for years with synapses that turn over on a time scale of weeks or less? Maybe we can imagine neural circuits playing an elaborate game of hot potato, and constantly moving information around to prevent synaptic turnover from destroying it. But this explanation seems problematic given the orders of magnitude that separate hours and weeks from decades."
Immune cells also turn over on a time scale of weeks, yet immune memory lasts decades. There's no serious effort to substantiate the claim that synapses can't hold memories because they turn over too fast, just a quick dismissal that mimics logic. This review is peppered with these rapid sleight-of-hand tricks in favor of its central thesis. Fun read, but not convincing.
I'm not sure why you call it a sleight-of-hand. Sure, there are other explanations for how a system with unstable components can itself be stable, but proposing one where the components themselves are more stable seems reasonable.
Is the author's alternative hypothesis a "reasonable" suggestion in the abstract? I guess it could be, but that's hardly the point.
The author fails to provide a compelling reason to doubt the claim that synapses hold long-term memories. They make a broad, unsubstantiated appeal, contradicted by the facts of other theories they endorse in the review, then assert that the claim is "problematic." That's what I'm calling a rhetorical sleight-of-hand.
The foregrounding of the organ transplant cases is an odd choice here. I ultimately found the review pretty convincing as a case for "synaptic weights aren't the whole story when it comes to memory storage." But the organ transplant cases are the weakest evidence in the whole thing, and I still think they're probably not real even if the author is right about the SPM hypothesis being wrong/incomplete.
There is a really obvious fact (at least, I think it's obvious, but maybe not!) that puts this discussion in a different light: the synaptic plasticity & memory hypothesis could be true of neurons without the restriction that the neurons be in the brain.
Almost a full percent of the neurons in the body are outside of the brain (mostly in the enteric nervous system (ENS), maybe half a percent) but also in other organs, including the heart. Is it crazy to imagine that those neurons form memories, along with those in the brain? My guess would be that those memories would be affective rather than strongly "cognitive," but that could fit with many of the transplant stories.
I think these transplant recipients do not need to lie - all it takes is that they've overheard something about their donor's character and took it as a cue to snap into a different attractor state. It's not like people are born to be vegan or meat eaters - it's a choice. In my case I've decided one day I want to be a vegetarian and stopped eating meat. Before I ate meat but vaguely felt that this isn't entirely right thing to do, even though it's tasty, but was to lazy to change my life. I needed a nudge. In my case it was that catering in my company made it very simple to switch to different diet, so I just did it. There's always some cost to switch mental model and identity. Maybe people who were a bit on the fence and thanks to the transplant of heart now got a socially acceptable way to change their habits used that as an excuse to try something different. Or maybe their mental model was compatible with the dancer spinning left or right but not both at the same time, and the event just let them switch to the other attractor.
Also, stories tend to get exaggerated and embellished over time. In some cases, we can even see that in action, like when the people conveniently wrote down accounts and the early accounts are a lot vaguer and more plausible than the later accounts, even from the same person.
One would also think that the receiving a hear transplant might be a disturbing salient reminder that you are made of meat. After all, your meat was taken out of you and now you have meat from a dead person inside you. Such a thing could easily turn one off from the idea of eating meat going forward. Honestly just reading about these transplant stories someone turns me off to the idea of eating meat.
But also, I've eaten enough steaks in my life that one would think I'd wake up screaming every night from nightmares about the slaughter-house, if memories transferred based on what you eat.
The content of this review is interesting, but it makes a couple of major rhetorical mistakes.
First, the writer comes off as a cheerleader for alternatives to the strong SPH. The "review" would be more interesting if it tried harder to engage with the questions raised by alternatives.
Second, it goes down the rabbit hole of physicalism and the philosophical meaning of knowledge when this is utterly irrelevant. Yes, there are valid senses in which any (non-maximally-high-entropy) arrangement of matter "remembers" information, but the actual question here is whether and how the information in question is available to your conscious thought processes (as e.g. immune memory presumably isn't).
"I can’t convey how much I hate this kind of argument and find it stupid. Either you believe everything is ultimately elementary particles obeying the laws of physics, as physicalism posits, or you don’t. If you don’t, fair enough, but I’m not sure how you’d falsify your position. If you do believe this, then thoughts, feelings, and so on are just useful ways of talking about a complex physical system."
Any theory of consciousness would have explanations for the *feeling* of experience, i.e. the hard problem of consciousness. Then, we can just differentiate conscious memory, which does require such an explanation, from immunal memory, which does not. This different will be a feature of the theories themselves and are not just "useful ways of talking about" things.
I loved this whole thing, myself. I didn't take the stories at the beginning as a key part of his argument, but rather as an entertaining way to frame the topic.
I'm no neuroscientist and so hopefully someone will help me understand. If we have neurons throughout our body, which we do, why isn't it plausible that memory could be stored in those multiple places as well?
And then it's surprising to me as a total outsider that we seem to know so little about the actual biological process of learning and memory. This synaptic model seems impoverished and the generally accepted idea, I gather, that there are probably some other molecular processes involved but we have absolutely no idea what they are... is that really the state of what is known?
I naively thought that there was more substantial knowledge under the idea that neurons that fire together wire together. But I'm getting the picture that we don't?
> If we have neurons throughout our body, which we do, why isn't it plausible that memory could be stored in those multiple places as well?
It might. But so far we haven't found indications for that. In contrast, we *have* found indication that the brain is involved in recalling memories. For example, there are neurons which get activate when subjects see a picture of Bill Clinton, or see the name "Bill Clinton" written somewhere, or hear the words "Bill Clinton", or even get some indirect clue which reminds them of Bill Clinton.
Theoretically, it *could* be that the concept of Bill Clinton is really stored in your heart, and only activates the very specific neurons in the brain upon recall. But we have not found indication for that (though to be fair, I doubt that someone has looked hard), and it would be more complicated than a system which stores the memory directly in those neurons in the brain.
> I naively thought that there was more substantial knowledge under the idea that neurons that fire together wire together. But I'm getting the picture that we don't?
We know a lot about the chemical details of how this fire-together-wire-together works. We also know a lot about the local "rules" which govern in which situations synapses become stronger and weaker, and also a bit about mechanisms and rules that change other parts of the neuron. We are pretty sure that this neuronal plasticity is involved in learning, because there are dozens of little puzzle pieces which point into that direction.
For the big question of how long-term memory is formed, we also know some things. For example, that (as far as we understand) our memories are first temporarily stored in the hippocampus, and then during sleep transferred into long-term memories in the prefrontal cortex. But we know very little about how the transfer of memory works in detail. Well, we know a few things, like that the brain replays some memories from the day during sleep, and this seems to be related to the transfer. On an abstract level, this already gives a nice coherent picture. But, which neurons exactly are chosen to store the memory in the long run, how the heck does the brain make sure that it's so nicely embedded that it can be recalled at will? We don't really know.
A disclaimer: I was talking about episodic memory: whom you have met today, what was in the news, what did you have for breakfast. There is also implicit memory (how to speak, how to play the piano) that forms by tons of repetition and probing and probably in a completely different way. We also don't really know the details of this, but it seems a lot less mysterious. Learning by tons of repetition and small changes each time is exactly the stuff that also works elsewhere, for example in LLMs.
With the different roles of REM sleep versus SWS sleep (slow wave sleep, "deep sleep") we are going into terrain with less solid foundation, but there is something like a standard hypothesis.
I mentioned that the brain replays some events from the last day during sleep. These replay events occur both in REM and SWS sleep (and even awake at rest). But the replay is much "cleaner" in SWS sleep. Probably this is the part where the memory is transferred from the ever-changing hippocampus into long-term storage in prefrontal cortex. For REM sleep, replays often occur in variation or together with other patterns. The hypothesis is that the brain uses REM sleep to combine the new memory with other memories and to integrate it into the general knowledge landscape. So when you wake up and suddenly you understand something much better than the day before, probably this happened in REM sleep.
For the connection to Alzheimer, this part is not mysterious. Alzheimer causes neurons to degenerate and die. Cells in the hippocampus are particularly affected, together with some other regions. I don't really know why the hippocampus is especially affected, but my first guess would be because it is the most plastic part of the brain, where many new neurons are born and connections change very fast. Since memories are first temporarily stored in hippocampus before they are transferred into long-term memory, it is not surprising that a sick hippocampus leads to problems with forming new memories. And that is exactly one of the early symptoms in Alzheimer: patients have problems forming new memories, but old memories are much less affected, or become only affected at later stages of Alzheimer.
So it is not mysterious why Alzheimer leads to exactly those symptoms that we observe, it fits nicely into our big picture of memory formation. The big discussion around Alzheimer is: WHY do neurons degenerate and die? This is what the finalist review #3 and the recent guest post here at ACX were about.
Enjoying your knowledgeable replies, I had the thought that Scott could create a corner of his Substack where people get to pose questions, either for advice or just about things they'd like to learn about, and people with knowledge to share could offer it.
The readership here has such a wide range of expertise, it seems like it would be fun to create a space to share it around more. Obviously with no pressure for anyone to respond at any depth to anyone else, just as one is interested or has energy to.
(I think I feel a bit sad about so many people going to ChatGPT for answers that are kind of common denominator answers instead of to other humans who have all this rich color and texture.)
I had exactly the same feeling with the guest post on Alzheimer. I find it just awesome that Scott has so often someone in the readership with this deep level of knowledge for a random question.
There is disagreement about the exact biomolecular mechanisms that encode information (specific receptor configurations? biomolecule modifications? just binary synapse strengths?), to be sure. But it's also the case that ~70% agreed or strongly agreed that "Long-term memory is primarily due to synapse strength", and only 10% disagreed.
A few key pieces of evidence for the synaptic encoding hypothesis that this piece neglects:
1. While the proteins that make up a synapse turnover on the order of days, the actual synapses themselves can be extremely stable in a ship-of-theseus style way. Yang (2009) demonstrates this - they teach mice some motor tasks, watch the dendritic spines that form, and note that a population of them are stable and still present months later.
2. In some studies, selective erasure of synapses has been able to delete one particular memory while sparing others. In Hayashi-Takagi (2015), the authors teach mice to cross a high-rope and also to balance on a rotating rod, while observing which synapses form during the learning of each task. By using optogenetics to erase one of the sets of synapses, they can force the mice to forget how to do one of the tasks while sparing their ability to perform the other.
As per the words of some of the leading researchers in the engram field, “There is a clear consensus on where the memory engram is stored—specific assemblies of synapses activated or formed during memory acquisition”
I feel like a lot of the weird examples from the first section might well be explained by survivor bias. The guy who got a heart from someone named Tim and then has dreams about George, or no different dreams at all, naturally never shows up in the literature.
And the kid who intuits that his donor was a kid about half his age (never mind that that locution seems awfully stilted for a five year old) but was “never told” about that — well, maybe he wasn’t and maybe he was, but even if not, little pitchers have big ears.
I’ll accept that these anecdotes may have been meant more to get our imagination flowing than as any sort of proof, and on that score they are entertaining enough. But I sort of have to classify them as analogous to near-death or out-of-body experiences, or (if you’ll forgive me) UFO sightings.
we all know that it is caused by a prion [protein] that survive being disgested, crossed the blood brain barrier and caused significant changes to neuron connections in the brain and therefore affect memory/personality/thought.
it is therefore possible for other proteins to do all the same thing and the reasons we don't notice are
they are rare
when they happened they are harmless/within expectation
so only when it is extremely harmful that we bother to look for a cause and find something.
Prions don't change your memory, they interfere with the proper functioning of your brain. This is like saying that pouring sugar in your car's gas tank changes its performance therefore the "gas provides energy" theory of automotive design is wrong.
first, it is changing memory according to the stricture laid out by the poster, [i.e. a broken rock has the memory of the thing that broke it.]. But yeah you don't buy it, I don't buy it either.
Second, Scott was arguing above about how poteins tend not to be able to survive being disgested, crossed the blood brain barrier and caused significant changes to neuron connections in the brain and therefore can not be use to store nor transfer memory. I just want to counter that with an example.
lastly, if there is a protein that that survive being disgested, crossed the blood brain barrier and caused significant changes to neurons in the brain, does that not increase your prior that there are more of them with maybe less pronounced effects that we may have not been looking for?
"There are things that affect the functioning of the brain" is very far from "there are things that can manipulate memory in arbitrarily complex ways". Pouring a glass of water on a laptop can break the laptop but it can't reprogram it in a different-but-still-functional way. Destroying function is much simpler than creating function.
I am glad to see us agree on almost everything and now we can narrow our disagreement [if there is any]
almost none of my statement below would be sourced but I assume it would be trivial for you to confirm with google. I thank you for your indulgent.
I agree that there is probably epsilon chance that "there are proteins that can manipulate memory in arbitrarily complex ways"
I hope that you agree that there are memories coded in DNA and the resulting RNA and protein, we call those memories instincts. That is anything a newborn baby can do immediately after being born [e.g. how to breathe, how to suck and swallow]. More complex memories are also coded, most prominently sexual preference. I hope we agree that many instincts and preferences are coded/mediated by protein? and if you have preference of something you have memory of it? what does it mean to have preference for something you have no memory of? [and ehmm, can you turn someone gay by injecting some proteins [?hormone] into their CSF or do you need to be exposed to it at certain development stage?]
And I am pretty sure that there are studies confirming that some/many memories are encoded in chemical signals, probably a protein of some sort.
wait actually do we agre that any protein that affect the neural circuit affect memory? or is the bar is being able to encode abitrary memories and inject it into another person so they can experience it? if it's the second thing then yeah I don't think anything like that exist.
but if we count preference and thoughts patterns then many proteins demonstrated this property. example: alpha-synuclein, implicated in parkinson disease, can cause depression [mostly through dopamine disregulation] and can cross blood brain barrier.
For memory retrieval (eg. where did I put my keys) many of us roll their eyes, which seems to help memory.
Similarly, smells help us evoke memories from long ago.
Do pro athletes know in their brain how to be good at sports vs. how much is in their body?
To me it is obvious that the memories are not only in the brain, but the brain is the last part in the signal-chain and the evocation of a memory is dependent on the set of signals that arrives in the brain (which are dependent on our physical sensors and nerves).
This is also why I believe uploading one’s brain would not do much since the brain probably couldn’t identify itself, without the sensors it’s used to/trained on.
Yes, I've always felt it would be agonizingly disorienting for a disembodied brain or brain emulation to exist, even if that were possible. There's so much of our experience of living that resides in our bodies.
Here is a journal article about how the heart has 40,000 neurons in a little “heart brain.” The article draws on the same stories as in this review and suggests memories may be transferred with this mini-brain in transplants. I find it strange the review did not mention the role of neurons that were transplanted.
These types of reviews are insidious. A maverick dares to challenge the scientific consensus in some highly technical field. It makes for a compelling story. They are usually filled with details, impressive technical knowledge, a compelling narrative, and they make the reader feel special for gaining such clever insights that those pesky scientists missed.
I know I'm being a bit too harsh, because I actually thought this review was on to something. That is until I did some further reading where everything began to fall apart. In particular the supposed smoking gun that synapses turn over in weeks to months and therefore cannot store long term memories. Except that many synapses do last years as many other commenters have pointed out. The fact that the author either didn't know this, or didn't report this is alarming.
The search for intellectual novelty has led to "hey guys, maybe cannibals had something going on there because SCIENCE! No wait, it's probably wrong."
"This controversial thing, often from the past, may actually have truth to it due to SCIENCE! One weird trick can explain everything!"
I think my brain is just physically revolting against the idea at this point. Like this particular article just hits the "parody" button in me, even if it doesn't intend to.
Best read of the year for me! As a CBT therapist that recently moved to Sensorimotor, I always wonder how we actually (phisycally) remember things and embody memories. After reading this post, now I know to know even less than before. Fascinating! :)
Took me several sittings to slog through. Dry like burnt toast, humour didn't land, absolute thicket of links and referenes. Which I'd want in an essay attempting to prove a particular point, but feels out of place in a not-a-book-review...the focus needs to stay on the thing being reviewed, with a minimum of other distractions and digressions. Just enough to establish the floor, as it were. Bold but frankly stomach-churning choice to lead off with cannibalism stories, and then get a redux halfway through, then a zinger at the end. This might be "insanely cool" for some people, for me I regret reading while eating.
I notice that only one of the Four Axioms got (arguably) refuted. I notice that a fair amount of the supporting evidence is not itself well-corroborated. Like yeah, you can interpret "half the mouse brain studies succeeded!" as optimistic, but I'm thinking "wow, a 50% failure rate, that sounds like random noise". Which is weird coming off a tangent about McCannibal's results being hard to replicate! Same thing with the transplant just-so stories, they're too cute by half to not trigger my epistemic red flag detector. It's eminently clear, from stuff in this review and just generally anyone familiar with basic psych (so like...most LW/ACX readers), that human memories are highly fallible, and we love overfitting events into plausible narratives. Moreover, it just cries out for base rate comparison: how many transplantees *don't* experience weirdly specific changes that correlate well with donor foibles? (Self-reported, no less!) The literature on anesthaesia is also pretty clear that it does all kinds of systemically weird shit to people...given all the semi-plausible hints here that memory can be stored and transmitted throughout the body, why wouldn't we expect some eerie concordances to arise by random chance?
On and on and on. It's a shot across the bow, to be sure, but with enough holes and weak links that I'm just not comfortable writing off the discrepancies as "well of course if it got the same amount of funding and hype as Big Brain Dogma then we'd fill in the blanks". That's veering into Fully General Counterargument territory.
I've been circling around the CPM for the better part of 10 years now. In addition to the many other excellent comments below pointing out flaws in the argument, I'd like to mention a few other important issues, some of which go beyond the article itself but might be of interest:
1. Readout. I can ask you to define a word, recall your most recent birthday party, or perform some known motor skill, and you can do so instantly. DNA transcription/RNA translation occurs on the timescale of at least minutes, and sometimes days. If RNA (or protein post-translational modifications, or histone acetylation, or whatever) is the substrate of human memory, how does this readout process take place?
2. Developmental biology. Neurons contain a dazzling array of cell surface proteins that help determine which two neurons make a synapse in the first place (these have names like "cadherins," "neuroligins," etc.). When they make a synapse, lots more intracellular machinery is recruited to help stabilize it. If you do things that naively "disrupt" the synapse (perhaps this is what is happening during metamorphosis, although I have no idea), these proteins still stick around and could in principle allow the same circuit to reform. Is this "molecular memory"? I wouldn't say so, considering that formation of the synapse is both necessary and sufficient for the memory to be expressed.
3. Theory. If you go into the history of these arguments, particularly Randy Gallistel's entertaining "Memory and the Computational Brain: Why Cognitive Science will Transform Neuroscience" (2010), a key hangup is the claim that neural networks are "write-only" memory systems. Sure, experience can alter some weights, but how can you possibly go in after the fact and read out that change? How can you store and retrieve the value of some variable? In my opinion, the success of artificial neural networks has destroyed this objection, so the argument has had to be resuscitated in a different guise. Mechanistic interpretability has taught us that it can be very difficult to trace the causal path from input to output in a massive, densely connected system, but the memory is nonetheless there.
That said, I have lots of respect for Sam Gershman and many of the other folks mentioned in this post. One of my favorite talks ever was an internal neuroscience seminar at Harvard where Gershman shared the thoughts that would later be included in his 2023 review article -- only for the audience to strenuously object to essentially every claim. For those thinking that academia has descended into a morass of conformity and groupthink, I think this provides a welcome corrective.
Your point 1 was my big question after reading this post. The timescales just don’t seem to add up, but I’m ignorant of the subject and can’t form an educated opinion of whether it’s an insurmountable obstacle or not.
"Among other reasons, he claimed that their God didn’t like the idea of him being eaten."
The following quote: "Staden immediately insisted that this in fact was the case: 'I told him yes, my God was angry, because he wanted to eat me.' "
'Their god' vs 'my god' is not hugely relevant to the review, but it put me in mind of the relative position of religious outlook in that episode. To Staden, their god might as well be *his* god, too, in essence. Auspicious slip-up?
In grad school we had a guest seminar from a researcher (sorry I forget who) who had developed techniques that were successful in helping paralyzed patients regain mobility. The technique involved having the patient think about moving their limb and simultaneously applying shocks to the gray matter in the spine, essentially training these regions to take on roles from the damaged motor cortex. This was the first time I heard that neural processing doesn't just occur in the brain. I don't have a biology background, do we have a good understanding of what the roles and capabilities of the extra-brain neurons are?
- extremely strawmaning positions of mainstream science (for example, no one thinking about actual neurons as bunch of weights it is an abstraction useful for large collectives of neurons)
- ignoring clear but maybe hard to formulate distinction between memories and state of system affected by previous events. If my arm get cut of fact that it’s still missing is not a memory mechanism. If my cells have epigenetic markers from malnutrition and now store more energy it’s not a memory it is a state of their metabolism. Of course it is a spectrum but I think we can’t ignore clear differences between information stored on hard drive and in the current processor state
- ignoring elephant in the room: mechanisms of transition from information stored in single cell to actual behaviour
- not providing reasonable questions about worms and sea slugs experiments - their learned behaviour is extremely simple and can be encoded by several bits. It is quite plausible that single protein amount affecting light sensitivity is enough to “encode” this behaviour
- not doubting memory from transplant observations. I see unnoticed information about donor leaking and reaction to it under stress much more plausible than actual memory transfer. I do believe that transplant can affect recipient behaviour in an unexpected ways because brain isn’t isolated system and some reactions can be casually connected to signals from other organs but I don’t believe in any information transfer
This post strongly reminded me of bioelectricity, i.e. the idea that non-brain cells also have the kind of electrical circuits as brain cells, just way fewer of them, and that they form a sort of low-level intelligence responsible for assembling cells into larger creatures. You can learn about it in Prof. Michael Levins Youtube Channel, for example here: https://www.youtube.com/watch?v=s-SL8qy-6kA
Perhaps such networks somehow interact with the proper intelligence in the brain?
The author can afford to be more precise with their definitions of memory. There is a loose definition of memory, where some arbitrary type of information is stored. Biochemistry is trivially implicated in this kind of information storage, as information is stored in rna/dna/etc. However, they conflate this loose definition of memory with "memory", the rememberance of complex visual and linguistic information (such as names and images). The problem is, they begin the review with anecdotes on the latter kind of memory, but all of their scientific evidence points towards the former, with no effort to bridge the two. I can quite easily believe that biochemistry stores some behavioral tendencies via epigenetics, but biochemistry routinely storing the names of people and places is much harder to believe (and evidence of this kind of memory being stored is not really included in this review).
> The most popular quantitative models of the brain, artificial neural networks (ANNs), assume that the strengths of the connections between neurons, or “weights”, completely determine how networks behave. These are the things that are assumed to change during learning, maybe through an algorithm like backpropagation, or maybe through something that looks more like Hebbian learning. If these models learn something, or store a memory, it has to be through changes in weights.
> In AI, too, weights are king. The term “weights” is even used as a synonym for “model parameters”. Weights are the numbers that completely characterize what a state-of-the-art model has learned through expensive training, and these days they can be ferociously protected trade secrets.
This is a popular oversimplification but technically incorrect - two main sets of parameters are learned during training. One is the connection weights, which determine how neurons influence each other, but there is a second, oft neglected set - the biases, which determine the activation threshold for a given neuron.
Thank you. I have no view on the correctness of any of this nor am I competent to form one, but I greatly appreciated the ride and I did learn a lot, including some wholly new (albeit low probability!) hypothesis!
With a lot of respect, I think you’re missing the point a bit here Scott. I don’t think the author’s main argument is about memories of specific events or facts. They are taking a wider (and more useful, I believe) definition of memory as the state of a system that influences behaviour. That can include aspects of memory encoded in chemistry and biology. You might think of it as “vibe” memory rather than associative memory between concepts.
I agree if wouldn’t be useful to store associative memories like you describe in the heart because the brain is the organ that uses these types of memories in processing. However, processes like biological responses to stimuli e.g. fat storage when getting lots of calories , are processed in a distrusted way to some extent, and it would make sense for there to be state (memories) encoded throughout the body to manage that.
We do indeed see this kind of state (memory) passed from mother to child.
I think that one of the main takeaways of this article is that memory can be much more complex and diverse than things like episodic or associative memory.
I'm surprised the author didn't mention Roger Penrose's ( the physicist) proposal that the microtubules in neurons store/delete information & are the basis of consciousness.
> But if any of these examples is at least kind of true …
These two phrases are dubious. I could list many historical instances in which there have been lots and lots of purported examples of a phenomenon, but actually the phenomenon doesn’t exist, and every last one of those purported examples turns out to be complete BS, no exceptions. This kind of thing happens all the time, unfortunately. Examples include UFO sightings, or Scott's old posts on parapsychology and 5-HTTLPR, among many others.
(Tbc, maybe in this case the phenomenon is real. Or maybe it isn’t. I’m just making a narrower claim that “hey look at how many striking examples I just listed!” is a weak argument in itself.)
When psychologists refer to one model of neuronal learning as a "mere-casual mechanism", and contrast this with a hypothesized mechanism that isn't mere causal, they don't mean that they think memory operates "acausually" or without the interaction of molecular elements. They mean that causal functioning is direct and mechanical, and involves no symbolic information. We can define symbolic information as an arrangement of patterns elements in some system that represents conditions outside of that system. Information becomes symbolic when it can be rearranged from within the system to represent some external condition that doesn't currently exist. If the elements within the system can be rearranged to represent what conditions might look like at some point in the future, we can call that a prediction. If they can be rearranged such that they represent conditions as they might have existed at some point in the past, we can call that a memory.
A mere-causal system, by contrast, does not contain symbolic information, but only a direct mechanical response to conditions as they currently exist. A thermostat, for example contains only elements that react mechanically (the thermometer, the circuit to the heater, the set point mechanism) - they don't "represent" anything other than what they are, and can't be used to forecast or recreate anything. A thermostat doesn't predict anything or remember anything; it just adapts to changing conditions according to it's set point (as determined by a human brain, not the thermostat itself). "Mere causal" is a bad term for this - I would have called it "mechanical", or "bio-mechanical" in the case of bacteria (assuming bacteria cannot process symbolic information).
The immune system appears to be another case of a bio-mechanical response system. While immune cells possess the capacity to adapt to new intrusive pathogens, this seems to be a chemical structural adaptation. The immune system isn't predicting anything, and not remembering anything in the form of symbolic information either, anymore than the thermostat is (because elements within the immune cells are not being rearranged "independent of current conditions" - they get rearranged when they encounter a new invasive cell). Immune cells do not "think".
A brain is different. We know that the human brain is a prediction machine, constantly checking current conditions against a template of what was expected to occur a few micro-seconds to years ago, and therefore must contain a symbolic representation of these predictions. We also know they contain memories (albeit reconstructed to some degree) which symbolically represent conditions as they are thought to have been, again from a few micro-seconds to years ago. Therefore, somewhere in the brain there must be a capacity for "information processing", the ability to rearrange systemic elements independent of current conditions.
The question is where this information processing capability is located. One possibility is within each individual neuron. In that case, neurons would be storing information and manipulating it somehow to help produce predictions and memories. Another possibility is in patterns of connections between neurons. This would consist of the growth and decay of physical connections, or changes in the weights between them.
There is very strong evidence that patterns of neural connections process symbolic information. The brain could hardly operate otherwise. The question is then whether or not this is sufficient to explain what the brain can do, or is an additional mechanism needed. Since this would be a more complex model, it requires additional evidence in support of it. Does it exist?
I think the evidence is consistent with a low level within neuron effect, which could then go on and affect macro-behavior, esp. in more primitive organisms. But it seems clear that the majority of the information stored and processed in the human brain is going on between cells, and that if intracell processing is going on, it works closely with intercellular processes. Strictly speaking, then, synaptic weights are not sufficient to explain all information processing, but do account for most of it.
Wait, if this was true, wouldn't it be very easy to test with blood transplants? Set up an experiment where people get transplanted with e.g. the blood of the nobel-prize-winning physicist, and some placebo blood, and check if they get better at solving physics problems?
(Or use a Navy SEAL's blood and test endurance against tests, or a drowning survivor's blood and test for hydrophobia, etc.)
In the essay and comment section, it seems part of the problem was that behavioral changes were characterized as memories. Food and sexual preferences may be better characterized as a novel reaction to a stimulus. Say, I lift a 30-pound dumbbell over my head daily, but I develop pain and on subsequent days I can no longer do so. I try, but can’t. The inability isn’t memory, but to an observer, it may appear to be.
Food and music preferences change without a heart transplant, sometimes suddenly. Behavior and mood are variable, and do not consistently correlate with any particular memory.
"If synaptic weights turn over on a time scale of weeks, maybe there’s something slower that turns over on a time scale of months. And maybe there’s something else that’s slower than that."
It would be interesting to analyze what various drugs that impact memory are doing and how they might impact this. Psychedelics in particular. If they increase neuroplasticity, why is it often so hard to remember trips? Whereas if they are disrupting the flow of memories from short term to long term storage, it makes sense
I lack the relevant background to be able to tell if this is plausible or if this is more like the biology version of Velikovskyism, so I plead epistemic learned helplessness ( https://slatestarcodex.com/2019/06/03/repost-epistemic-learned-helplessness/ ) and just shrug my shoulders and move on. Probably a lot of the other readers should too.
Interesting review on Synaptic Plasticity! It's fascinating how our brains constantly adapt. Reminds me of the evolving strategies you need in many io games – always learning and adjusting to stay competitive. Some of those io games even mirror the complex networks our neurons create! Keep exploring these connections, both in science and those addictive browser games. https://iogamesonl.com/
It blows my mind that the stuff about memories from the things you eat made it in here. Did we just skip past digestion breaking down pretty much everything that would give you a memory? It makes the rest of the article seem incredibly credulous.
And the “if even 25% of these stories are true” argument is so bad.
I njoyed reading this review, even as someones who studies neuroscience in uni and knew most of the facts.
My main issue is that:
a) there is a distinction to be made between cognitive and none cognitive memory. I don't think I need to go into much detial to explain that immune system memory is not the same as an episodic memory of me as a child tasting strawberries for the first time, or even semantic memory of knowing a fact. Though there might be grey areas, I think its a distinction that is super easy to build a classifier for.
b) the review doesn't make a super compelling case as to why wouldn't the objective truth be something like: cognitive memory is govenred mostly by synaptic weights updating, while non cognitive memory is goverend by all sorts of mechanisms the we evolutionary had before having a brain (and cognition, without getting into the weeds of what exactly is cognition).
That's what most biologists think today, at least in my experience. And sure, at the begining the author talks about organ transplants and memory, but more of as a hook to the point of the review than actually zooming into the evidence and what this could entail.
I was a student almost a decade ago, and back then we were taught:
1. cognitive memory is probably mostly connections between neurons, but neurons have all sorts of complex modulating behaviors that we don't yet understand
2. there are, of course, other forms of "memory" in an organism
Both of these weren't controversial, and this review didn't really change my priors on that too much.
> there are two common arguments against ‘counting’ immune memory as a cognitive-like form of memory ... the Error Argument... the Mere Causal Argument.
If I blow off your leg with explosives, it will still be gone tomorrow. Is that "memory"? So I propose a third argument, the "Not Cognitive" argument: any form of memory that isn't accessed in the mind in a vaguely memory-like way isn't "memory". So while maybe your mind registers some leg pain, the continued absence of the limb doesn't count as "memory". Immune memory can be called "memory" informally, but in a discussion about how the mind works it seems like a red herring. (meanwhile the Error Argument seems to imply that computer memory isn't memory because the computer doesn't misremember.)
> Activity-dependent synaptic plasticity is ... both necessary and sufficient for the information storage underlying the type of memory mediated by the brain area in which that plasticity is observed.
But soon afterward it starts treating SPM as a strawman that ALL organisms on Earth use and need synapses for memory, so that we can "falsify" it by creating doubt that insects rely exclusively on synaptic memory. This is silly! What scientist has ever claimed that all organisms rely exclusively on synaptic memory? I think it would be rare to find a scientist saying this about all organisms with brains, but the author looks all the way down to single-celled organisms as if this isn't banal and unrelated to the subject of interest.
No, almost everyone talking about SPM is talking about humans. I expect other mammals to be studied mainly due to a certain social pressure not to do human vivisection, not because we care that much about how rats' brains work.
In section VI it is asserted that synaptic weight changes are insufficient for learning and memory, but the text prior to the sentence "And there are other reasons to think synapses are insufficient..." shows merely that other forms of memory may/do exist in some insects. I would point out that this doesn't even disconfirm SPM in the insects themselves. Synaptic storage can potentially be sufficient, or even necessary, even if we know for a fact that memories are also stored elsewhere, because redundancies exist in nature. Granted, if synaptic storage *is* sufficient and necessary then it would be surprising for evolution to preserve other mechanisms in these insects.
But it's possible for a mechanism to be sufficient and necessary and *gimpy*: a mechanism can *work* on its own but be suboptimal, just we can function with one eye but it's suboptimal. We can also function with only two hands, which as a father of two toddlers, I think is also very much less than ideal, but it's all evolution could manage. If a primary mechanism is sufficient, necessary and gimpy, there is selection pressure to preserve a secondary mechanism.
After that sentence, it is asked "How can we remember things for years with synapses that turn over on a time scale of weeks or less?"
This reminds me that computer DRAM can potentially remember things for years using memory cells that degrade within one second or so. Which, in turn, reminds me of a potential biological analogue, because I've been having an odd experience lately. I'm 45 years old, and just this year I noticed something: occasionally, just as I wake up, when my eyes are closed, I see a sequence of images roughly 0.22 seconds apart, with a ~0.05 second fade between the images. There are no thoughts or feelings associated with them; I see only the images themselves. Sometimes they are color, but they seem more often closer to monochrome. Every image tends to be visually similar to the next one in the sequence, with some overlap between the two. Some images depict something clearly, like a face, but more typically they depict parts of objects. And I have seen two kinds of image sequences: sometimes, the images are primarily foveatic (clear images in the center of my view) while other times they are primarily peripheral (blurry images in peripheral vision). It hasn't happened for a week or two now, but IIRC, in the most recent instance I tried thinking about a specific object to see if the object would appear in my view. It did not. These experiences, of course, led me to hypothesize that during sleep my brain could be systematically refreshing visual memories sort of like how DRAM does, and that a glitch was letting me see the process consciously.
The author lists a series of papers challenging elements of SPM, but oddly doesn't explain what any of the papers propose as an alternative. Maybe they agree with his CPM (or Martin et al 2000's hypothesis which evidently is not called CPM: "I’m just the only one calling it this") ― but he doesn't say so.
> What are these molecular and intracellular processes I allude to? Is learning and memory stored in RNA, as McConnell thought and Glanzman et al.’s experiments suggest? Is it stored in post-translational modifications to proteins, like Crick thought? Is it stored via epigenetic mechanisms, or in the stable states of gene regulatory networks, or in the action of transcription factors? The truth is that no one really knows. But I think there’s evidence for all of these...
Well, as I said, it would be surprising for evolution to preserve other mechanisms if one is necessary and sufficient. Is it possible? Sure. Is it worth investigating? Maybe. Do I trust this particular author to give me a clear-headed, unbiased view of the matter? Not at all. The vibes, the "reasoning", and the rhetorical approach I see in this article feel similar to the articles I have read by global warming dismissives.
"What causes global warming? Is it the sun? cosmic rays? Natural internal variability? Volcanoes? Increasing humidity? The truth is that no one really knows, but I think there’s evidence for all of these!" Well, I can tell you why each of these does not cause global warming and how we know it's mostly greenhouse gases and mainly CO₂. I don't know much about SPM, but the vibes he gives off? The vibes are very reminiscent. He's just not thinking clearly.
Funny and interesting, and it was voted a finalist so I have to post it - but I don't believe this at all.
Main concern: how do you encode memories in a chemical? That is, suppose you want to encode the fact "My name is Tim". It's easy to imagine how this works as neural connections: during infancy, you hear language, your brain forms a predictive model of the language, you get neurons representing the sounds T, I, and M, and the word "name", and you get links between them. Eventually there's some complicated network connecting name, T + I + M, the sight of a nametag, the motor program of shaking hands, and all the other concepts relating to your name being Tim. We know something like this has to work because LLMs can have this type of memory and they're nothing but connections.
How would you implement this in a protein? All I can think of is a binary-like code with literal text - using one amino acid to represent 0, another to represent 1, and coding the ASCII string "M-Y-N-A-M-E-I-S-T-I-M", and then having a decoder somewhere in the brain. But this doesn't match the phenomenology of human memory, which is sensory (eg associated with pictures, smells, etc) and works even without language (eg in preverbal children). CF https://slatestarcodex.com/2017/09/07/how-do-we-get-breasts-out-of-bayes-theorem/ , which asks a similar question about genetic memory.
But also, even supposing you could do this - in order for the organ transplants to make sense, you need for the heart to be producing or storing these chemicals (why?) in such large quantities that they get into the bloodstream, cross the BBB (why would you have a transporter for these?), make it into the brain, and perfectly interface with the host's own memory-decoder system (how?). Then either the heart has to keep making these constantly (why?) or there has to be some mechanism for a single low-dose exposure to get the host brain to start making them itself (why?) And if the cannibalism claim is to make sense, you have to be able to absorb these extremely complex chemicals through the digestive tract, which is usually doing a pretty good job preventing you from absorbing complex chemicals. Why do this, rather than just keep the chemicals in the relevant part of the brain?
Also, if we had these chemicals, a natural application of them would be genetic memory, eg in the womb you get a big dose of your mother's memory chemicals and now you understand the world as well as she does. As far as I know there are far fewer stories of people mysteriously having their mother's memories compared to their organ donors' memories, adjusted for the ratio of people who have mothers vs. organ transplants.
Also, the writer asks whether losing your arm might make you forget something, or eating an animal liver might give you some of the animal's memories. But many people lose limbs and eat animals, and as far as I know nobody reports these phenomena.
Agreed, I think this review goes way too far. Sure there are other mechanisms for memory in the brain besides synapses. No, I don't believe anything about heart transplants transferring memory.
I think the way you would encode "TIM" in protein is in a protein network. You could conceivably do this inside a single cell, AFAIK.
The problem is cells don't have mouths to say "my name is Tim" with. Within-cell memory is absolutely real and important, but it's just not the right scale for organism-level behavior.
You could have a network of lots of different proteins, but I don't understand how the symbols would be grounded. For example, what would it mean to say that arginine = M instead of N? Unless it bottoms out in some kind of behavior like saying an M or hearing an M, it's just arginine.
...maybe this is just the same point you're making.
(Upon closer reread, I think you were making the point I was making, I just hadn't fully understood my own point)
Not a protein expert, but I think many proteins are very weird and have many different possible configurations, and those configurations can be influenced by all sorts of features in the environment. In theory you could have a "name" protein that has a different configuration for each of the 1k most common names in the English language or something. This would be a weird way to do things, though!
Have you read "How Life Works" by Phillip Ball? As someone who had done their best to avoid modern molecular biology in grad school, I found it very useful for making sense of the snippets I had heard. It made molecular biology sound a frightening amount like social science.
I still don't understand how this would be grounded. That is, I 100% agree a protein could have 1000 different states. But what makes those states correspond to 1000 different names, as opposed to 1000 different shades of blue?
Suppose you hear the name Tim, and it equals state #489. You need the brain (which is doing some kind of auditory processing and has decided that "Tim" is important enough to devote a protein state slot to) to encode that by modifying the protein to state #489. Then the protein goes to the heart, and the heart is transplanted into a different person. Then the protein goes to the second person's brain. Unless there is a natural mapping between names and protein states, how does the second person get the same protein state code where it knows that a protein in state #489 equals the name "Tim"?
(for that matter, how does the first person establish a protein-state-to-name code which is consistent across all relevant brain cells?)
I agree! These are all very good reasons why you would not want to encode people's names in a protein. It is just too far away from the physical stimulus, there's no easy way for the encoder and decoder to coordinate on a shared meaning.
But many of these problems are also true for synaptic memory. I don't think cortex or hippocampus transplants would transfer memories either.
The vegetarianism thing might work though? The transplant cells are not used to getting meat bits, and when they get meat bits, they tell the brain, somehow, "hey, knock it off with the meat bits".
I don't know what you mean by cortex or hippocampus transplants. It seems to me that a full-brain transplant would trivially transfer memory.
The reason I don't think these problems are true for synaptic memory is that the brain is already busy building a complex model of the world, and so if you want to encode the concept "my name is Tim", all you need to do is connect pre-existing brain regions holding the concept of self, name, Tim, etc.
I think you guys are thinking too locally. In a sense, evolution stores mechanisms for relevant "memory" of adaptive situations. As we evolve, we get better at linking modules that reduce uncertainty. That uncertainty reduction has to built on a scaffold of systems that more simply reduce uncertainty. But why would you assume the "direction" is up instead of complex reverberations throughout, information possibilities stored in a manner that is somehwhat "black bbox" to us over the scale of the organism. We think of our thoughts and emotions as the highest projections of that, one's we directly experience. But the entire system stays in dynamic interaction with the environment, past and future. More basic markov bound modules are part of the structural foundation of the higher dimensional more sophisticated stuff.
But just like getting healthy changes your feeling and perception, other changes have upstream effects, which lead to downstream effects like epigentic changes. The whole system is constantly in flux regulating an organism in dynamic real time.
And a heart might be a big enough chunk of such a system that it can interact enough with the larger system to make itself heard in a way that is coherent.
The one relevant thing common to everyone is the way nerve impulses work. So perhaps the idea of transferred memories goes from impossible to remotely conceivable if these memories are based on a sequence of sensory nerve signals, i.e. not the letters T, I, M encoded somehow, but the "waveform" (for want of a better word) transmitted by nerves in the ear of the sound of hearing someone say "Hi Tim" or "My name is Tim".
Similarly, I imagine someone drowning would experience a sequence of distressing nerve impulses that might be rapidly encoded somehow in non-neural cells and be characteristic enough that they would be vaguely recognised by another brain as a disquieting recollection of an incident involving water!
As for the encoding mechanism itself, if it involves a cluster of connected cells all you really need is for each cell to be able to record a nerve firing pattern at a given time, and a link to another cell which can record another a short while later, so the whole ensemble is like a linked list or circular tape.
This I think is the answer.
The transplant heart doesn't remember the concept of the word "Tim", but it has some kind of somatic memory of spending a lifetime in a body that responds in a particular way to the physical stimulus of the spoken word "Tim".
That could also be why she got most of the last name wrong. Not enough fidelity in the signal to resolve the whole thing correctly.
This is more plausible to me. I was thinking not so much of codes but just of chemical reactions in the body; somatic information if you like
It made me think of that species of wasp that plants its eggs into spiders, and the spiders start weaving webs that are not useful for their purposes but perfectly useful for holding the chrysalis of the wasp when it’s ready to emerge from the body of the spider which is now completely consumed from the inside . There is a chemical invasion that takes place that alters the behaviour of the spider. One could view a transplanted organ as a chemical invasion.
There are more things between heaven and earth than are dreamt of in your philosophy Horatio
In the same way certain pattern of electrical charges inside processor represents M or N. But if we are talking about language model on a single cell it probably won’t have single letter representation since our llms don’t use them
Yes, encoding is the real problem. It would need to be something that could diffuse throughout the body. IIUC, extracellular RNA tends to be rapidly digested. (Also, ingested food is reduced to quite simple forms during digestion. We aren't planaria.) Perhaps carbohydrate tagging (of proteins?) would be more plausible. But it would need to be something that could penetrate a cell membranes.
Agreed. It's very hard to imagine a way that, say, a cardiac cell could interface with the outside environment with any specificity. Maybe at the level of "is my organism in a rich environment, or a poor environment? High threat, or low threat?"
That's memory of a kind, but not what people have in mind here.
But the cell alone might not be the vehicle. A heart is connected to the central nervous system and could be participating in all sorts of ways, including meeting the host CNS halfway, translating its own memories, etc. Both sides might have enough intelligence that they can work out the encoding differences, the way your brain adapts after healing from an injury.
re the mother part: even if the heart transplant version worked, it would make sense that it doesn't work at birth, because you would need the model developed by being alive for a while to actually translate those memories into anything useful. it doesn't help to have a signal that resonates with the name "tim" if you don't know how to identify discrete sounds out of the signals from your ears yet.
(on the other hand, hell, maybe this what people thinking they remember their past lives is--oh you thought you drowned and were reincarnated? well maybe your mom had a traumatic experience in a swimming pool once)
If the memory-creating molecules persist in the body for years, shouldn't they cause you to have some of your mother's memories once your brain has developed a sufficient world model?
I mean, I have no idea. maybe not cause by the time they would you've replaced them with your own somehow.
Children are very similar to their parents in a variety of ways. Perhaps part of that is not only due to genetics and nurture, but also to inherited aversions/affinities (and other very simple sorts of memories) via cells transferred during gestation and breastfeeding. And the reverse: some fetal-originated cells persist and circulate in the mother as well even after the baby is born.
Maybe there's more going on than just molecules in the case of organs. They're big self-contained systems that have some neurons and connect with the host CNS. A mother doesn't transplant a large pre-built system to a fetus.
I feel like we would have noticed the similarity between these past life memories and the mother's memories by now though. Mothers are the most likely people to hear these stories in the first place. As far as I know this isn't a commonly known correlation.
If the encoding of memories depends on the development of the brain, that naturally makes different peoples' brains encode information in incompatible ways. If there were enough of a biologically fixed association between chemical signals and specific concepts that the transplant or cannibalism things worked, I'd expect these concepts to already exist in the brains of babies.
If memory transfer is possible, why do you need to develop your own model by being alive instead of just getting your mother's when (or before) you're born?
Babies are born with tons of instincts. It's easier to notice in other animals, since they are born less prematurely and with more motor skills than baby humans. But even newborn humans know to crawl to the dark circles of the nipples and suck on them, to stay close to Mom's familiar scent and voice, to cry when alone, afraid, uncomfortable, etc.
Perhaps some babies differ in much more subtle instincts based on the cells they received from Mom during gestation and breastfeeding.
Obviously those instinctual aversions and affinities are not all that complex humans can do, plus the kid might encounter a different environment in life and need to adapt to it, so it's still worth developing one's own memories in addition.
I thought the review was fairly clear that a complex memory like "my name is Tim" would be out of scope for the proposed non-synaptic memory, and nobody is saying that the ceullular system would completely replace SPM. The definition from the review is clear in that non-synaptic process would be an addition to, not replacement of, the synaptic system, because the synaptic system clearly exists and works to a great degree:
>The formation, consolidation, and retrieval of learning and memory in biological systems often involves stimulus-dependent, non-synaptic molecular and intracellular processes. These processes do not just serve synaptic-weight-based mechanisms, but provide complementary mechanisms. They are necessary for making and keeping long-term memories, but not sufficient, and interact with synaptic-weight-based mechanisms in nontrivial ways.
A single cell organism doesn't have or need a concept of names, so its memory system has no reason to evolve that kind of capability. As organisms grow more complex, they need more complex memories, and thus more capable memory systems. As long as the old system isn't actively harmful, it sticks around because evolution is a hoarder. It keeps doing whatever it's been doing forever and, biochemistry being as baroque as it is, the old systems interacting in useful and significant ways with the new wouldn't surprise me at all.
As you say, there are some fairly basic objections to forming complex memories through transplants, let alone food, and the review might be overplaying that angle in the introduction. But the basic idea of non-synaptic memory has been well argued, I suppose (speaking as an interested reader).
I didn't think it was clear - the "name is Tim" example is taken directly from the examples of things people anomalously "remembered" after organ transplants. Someone got a organ from a person named Tim and started having dreams about a person named Tim. I thought the author was attributing this to non-synaptic memory.
Maybe it was just my perception then. I considered the intro to be a light joke, a rhetorical device with just enough facts to be enticing rather than silly . I'm pretty sure you've used that device yourself many times in your blog posts, nothing wrong with that. I don't dispute the veracity of those anecdotes either but I feel confident in dismissing them as coincidence if they really happened as described, because what about the millions of recipients you don't hear about that also blurt out random things after waking up from anesthesia? Maybe receiving a transplant does funny things to your biochemistry in general, including switching you between gay and straight in rare cases?
When the review gets from the fancy to the factual parts, however, it argues from experiments with light stimuli and RNA transplants. That seems much more plausible to have an actual mechanism behind it, and that's the basis of how I rated my review of the review, not Power Rangers and dreams.
I think this is a fair reading, but it wasn't clear to me at all. Most of the reasons this idea might be a Very Exciting Big Deal depend on some of the silliness being serious. Otherwise this is just biology as usual. Which TBF can be very exciting in its own right!
These seem to be long-time changes, not just random stuff blurted out when waking up from anesthesia.
We also probably suffer from not enough research here. Only people who (experienced a personality change large enough + were fairly extroverted and open to talk about it + their surgeons were willing to listen to them) are currently recorded. These are three big filters.
"Gay" women end up attracted to and marrying men all the time. It happens so frequently it's not notable in the slightest. Now if it was a previously gay man who suddenly found himself attracted to women for the first time in his life, THAT would be notable. For the woman it's just the norm.
I agree but think the even bigger point is that the reviewer is defining "memory" too loosely to be of interest to me. If I can't recall it then it's not what I call a 'memory'. The only evidence presented towards recallable memory is the transplant evidence, which the reviewer themselves admits is no weirder than many other things we don't believe in.
Still, I love it when people embrace odd theories like this and the gut microbiome one you 'contra'd earlier this year. Aesthetically and emotionally, I want people to go down oddball science interest trajectories and stand in the face of conventional wisdom or even common sense. (Just don't put them in charge of our health system next time, please!)
Neurons themselves are nothing but chemical/biological processes. Is it really weird that some alternative combination of chemical/biological interactions could serve the same purpose ?
At any given point in time, a neural network embodies a particular mathematical algorithm, and any given algorithm can presumably be instantiated in many different ways.
I don't think it's any more intuitively obvious a priori that a memory can be stored in the form of a neural network than in some other form. The obviousness that storing memories in the form of a neural network is possible mostly comes from the confirmation we got from modern AI, and it seems to me that dismissing other mechanisms as much more unlikely a priori is hindsight bias.
Re: "we don't get our mother's memories", uhm, yes we do ? How did you learn to breath ? It was thanks to your mother's and father's DNA, which is not a neural network.
Separately, I think much of the presented evidence in the review is rather weak. The experiments about grinding worms into dust and the like would be big if true, but they seem rather untrustworthy at first glance. But a weak form of the proposed theory is obviously true (re:DNA, individual cells), and a somewhat stronger version does not seem a priori very implausible to me.
"Neurons themselves are nothing but chemical/biological processes. Is it really weird that some alternative combination of chemical/biological interactions could serve the same purpose"
My point is that the brain already contains a model of the world - it's the one we use to think. So all you need to do to encode a memory is to connect things in this pre-existing model. Since you already have a concept of "name", "self", "T", "I" and "M", you can connect those brain cells to form "my name is Tim". Protein molecules don't naturally have any of these concepts, so it's hard.
We have instincts encoded in our DNA, but we don't inherit specific memories from our parents like their names, their biographies, or the book learning that they've absorbed.
Trying to avoid the words 'instinct' and 'memory' for now :
Our parents transmit to us their knowledge of how to breath, and our brain is then able to retrieve and use this information to inform its behavior. This is an existence proof that it is possible to usefully store and retrieve information with a mechanism that has nothing to do with synaptic weights. It so happens that the DNA mechanism is used to carry information that pertains mostly to our subconscious, but that doesn't mean there couldn't be yet another non-synaptic mechanism used by our consciousness to store and retrieve relevant information in daily life.
Having a concept of 'name' or 'TIM' ultimately cashes out in some physical state or process, whether it's in a brain or a LLM or DNA or something else. I agree that the fact we already know these concepts exist in the brain gives a penalty to believing that they were also reproduced in some other medium, but not a huge penalty. I disagree that it is a priori more natural to have these concepts exist as a result of some synaptic weights than in some other form. (This is in answer to you saying 'Protein molecules don't naturally have any of these concepts'. Neither does the brain, or a pre-training LLM)
But neurons don’t have such concepts too? If we oversimplify neurons to bunch of weights or look at LLMs we encounter same problem - blank or random state doesn’t represent anything outside. Electrical impulse isn’t anymore “T” then hypothetical protein signal, fact that it represents T is external encoding
"a neural network embodies a particular mathematical algorithm"
No. Any physical configuration could correspond to an infinite number of mathematical operations--the so-called q-addition which is a slight modification of regular addition. Given any physical machine implementing ordinary addition, you could, by suitably reinterpreting its internal states, say that it is in fact implementing q-addition instead.
Am I correct that this applies equally well to neural networks or any other physical medium which may or may not be used for memories ?
Also, can you point to some useful reading related to your assertion ? It seems interesting and it's surprisingly difficult to find anything relevant.
This argument is given by Edward Feser on his blog
http://edwardfeser.blogspot.com/2017/01/revisiting-ross-on-immateriality-of.html
Scott: I don't recall you ever posting a top-of-thread opinion on any of the previous reviews (please correct me if I'm wrong). But by saying you don't believe it, you're influencing your readership, and it seems like you're putting your finger on the scales of this contest. It's your blog to do with as you wish, but to be fair to the OP, you should have let the commentariat post their own opinions and only responded as necessary. Just saying this out a sense of fairness to the OP.
That's a good point and I hadn't thought of it. I'll try to refrain from this for the rest of the contest.
I’m not sure if the fairness of the contest is so important that we want to be sacrificing potential good discussions just to preserve it. (Easy for me to say, I’m not a contestant.)
I enjoyed Scott throwing in a large dose of skepticism as I found the writing compelling. But I agree with you wholeheartedly, should this review have won, then would be a great time for Scott's heavy influence.
I agree, that his comment could be a strong influence on the competition and thus unfair.
But as a host, it is also his general responsible to prevent unreliable facts (or "bad epistemology"? or "potential untruths"?) to spread too far.
I think his top-level comment was a good compromise between these two goals ("fairness" and "not spreading potential untruths").
The examples are so extreme as to raise doubts, but one might still want to explore further, and relevant to the project of this site, I suggest that kidney donors and recipients volunteer for very extensive research into memory and physiological changes.
Joking aside, I think an important question is raised. All cells in the body are joined in the same project, so is there ways that this unity of being and purpose is updated that leave sophisticated intra-cellular traces yet to be identified. And could that affect the quality of our consciousness and our understanding of what it means to be alive and human.
Even if memory can be partly encoded in bodily areas outside the brain, there's yet another problem in transferring it to a donor: It seems highly unlikely that everyone's encoding would work with the same interchangeable protocol, any more than everyone's (brain) memory is likely to be encoded in exactly the same pattern of neurone connections (or however it works).
Drifting slightly off-topic, but on the subject of cannibalism, earlier this year there was a report of bones dated around five thousand years ago being found in the UK somewhere, and these had cut marks and bite marks. But I think the researchers' conclusions that these were evidence of cannibalism may be somewhat hasty.
It was around three thousand BC when farming first took off in the British Isles, and although one tends to think of farmers as being fairly meek and mild compared with warriors, those first farmers must have faced a massive threat from the majority population of hunter gatherers around then. Consequently the farmers were most likely the very opposite of meek. For a parallel, think cocaine growers in a South American jungle. Trespassers are unlikely to be given a pat on the head and released unharmed!
These early farmers probably hired muscle to guard their farms, and kept large fierce dogs to help. Anyone caught trying to raid a farm would have faced a hideous death, to discourage other would be robbers. Quite likely they were dealt with similar to how the Chinese punished pirates: These were tied to a pole on the shore, and had chunks of their flesh cut off to the bone, with torniquets applied to prevent bleeding to death too fast, and the flesh would be thrown to dogs or pigs to be eaten in sight of the victims. I suspect that is a more likely explanation for the cut and gnawed bones.
Craniopagus twins that share thoughts exist, so maybe the encodings are interpretable? https://en.wikipedia.org/wiki/Craniopagus_twins#Tatiana_and_Krista (see "thalamic bridge")
Yes, I sketched a rough idea of how it might work elsewhere in the thread
It seems unlikely to me that a brain has much of a standard encoding scheme, especially one too complex to be interpreted with some work.
Doubly so for the rest of the body, if memory is stored there; memories there would keep very well, and take a long time to recall anyways, so there isn't that much harm in them taking minutes to interpret, or even weeks.
Sarah Constantin asserts here that we know that at least Purkinje cells in humans can encode information by themselves: https://sarahconstantin.substack.com/p/what-does-the-cerebellum-do-anyway (ctrl+f "Even single cells can exhibit learning.") So if she's to be trusted, then at least a mild form of the "not all learning and memory is synapses" hypothesis has to be true. If you disagree with her I'd love to hear it.
LLMs famously have bad memory. And we know that electroshock therapy tends to cause amnesia of the last few weeks, but not anything else. Maybe the last few weeks of memory are present in synaptic connections, but long term memory is somewhere else?
The organ transplant stories sound like they might share similar problems with past life stories- they're prone to suggestion, we don't hear about the null results, and the evidence base could be really bad. Perhaps if any of the stories are actually true, they represent information overheard by the unconscious mind in the operating room ("please bring me Tim's heart").
...But if I wanted to steelman them: suppose that there are long-term storage cells in certain abdominal organs that encode information like Purkinje cells; they could then transmit the information through the nervous system, like any other signal, rather than via some kind of bloodstream chemical. Maybe the cells don't always survive the death of their original body and the subsequent transplantation, or they fail to re-integrate or are overwritten by the new host's nervous system, but when they aren't, they transmit memories!
Edit: as for why no maternal memory: maternal->fetal microchimerism does happen, including of nervous system cells, but it apparently happens in very small amounts. If that kind of information can't be compressed any further than whole cell configurations, and the body already tends to discourage somatic cell sharing for evolutionary reasons, then maybe it just didn't get a chance to evolve.
The most sensible way I can think of to store such memories is
1) Each neuron picks a random sequence of 20 base pairs (or similar) and generates lots of RNA.
2) When 2 nearby neurons fire, the RNA leaks across, and it gets attached to the RNA produced by nearby neurons.
So each individual strand of RNA just store "Neuron id 7fc35 is next to neuron 4ga1802", but given a whole load of these little strands, you can reconstruct a the entire network structure.
If flatworm brains are all basically identical, but each individual human has their own pattern of neurons and tags, then this would predict that memories could be shared between flatworms, but not humans.
And passing down memories is tricky. There would be a significant risk of forming some distorted nonsense memory. Something useless, but particularly easy to pass down.
You can just make “foreign” RNA concentration representation of connection strength between two cells and you can make cells normally grow in a way to keep concentrations constant. It can be reparation mechanism theoretically capable to restore missing neuron using weights from its neighbours
But to actual transfer knowledge even between worms you will need signatures to be exact same not random
I agree the review goes way too far (the parts about eating memories, or storing them in hearts and arms have to be a joke, I assume they were just being used as a hook to get the review to stand out).
But I don't think the review is disputing that groups of neurons store memories in combination, or that neurons communicate and function primarily through synapses. It's only disputing that long term memories are physically stored in the synapses themselves.
Having said that, although I was initially convinced, I have been unconvinced after further reading. The biggest piece of evidence was that human synapses turn over in weeks to months and so cannot store memories on the order of years. But that is only half true - the most stable dendritic spines last many years. In mice 75% of these types of spines were stable over 1.5 years. There is scarce data on primates but one study showed that the most stable spines in the macaque visual cortex are twice as stable as rodents (with 3% turnover in months.
The second big argument was that neurons can restore synaptic weights that have been erased. Except that these were sea slugs which have a very simple neural structure compared to humans. It's plausible cells might have some ability to do spacial coordination. For instance a cell in the V1 cortex might "know" which direction the V2 cortex is and if you erase it's synapses it might be able to restore them in roughly the right place but not perfectly.
Thoughts on ways to potentially rescue this hypothesis:
1. Neurons don't have any inherent meaning either -- the meaning of neurons is just, like, some sort of network learning thing. (Well, mostly -- obviously you've got sense data to ground it, and also genes can *somehow* control things finely enough to result in some fairly specific instincts. But a lot of it is this sort of network learning thing.) So the same could be true of non-neuronal learning. Of course, that wouldn't allow for the sort of transfer that's posited here, via transplant or ingestion. But that stuff seems (in humans a least) pretty tenuous anyway. Although...
2. Perhaps the cannibalism thing could nonetheless still work *in flatworms* or other such comparatively simple creatures, where the network is shallower and more fixed (consider C. elegans with its famously fixed connectome), so that meaning might be retained across distinct individuals of the same species in such cases, though not in mammals?
It's tangential, but I do have to say, the whole "cellular learning is real, evolution doesn't create things from scratch" line of thinking is pretty suggestive of *something* going on at the individual cellular level, at least. And like, these sorts of lower-level neuron-based workings of the brain (synaptic or otherwise) are certainly helpful for understanding some mental phenomena that seem higher-level -- consider semantic satiation! Sure seems like a lot of what's going on there is "your brain is made of neurons"...
> in order for the organ transplants to make sense, you need for the heart to be producing or storing these chemicals (why?) in such large quantities that they get into the bloodstream, cross the BBB (why would you have a transporter for these?), make it into the brain, and perfectly interface with the host's own memory-decoder system (how?). Then either the heart has to keep making these constantly (why?) or there has to be some mechanism for a single low-dose exposure to get the host brain to start making them itself (why?)
I think you're being too restrictive in how heart-to-brain data transfer could conceivably work. Heart/lungs can certainly send impressions like "excess fluid in bronchial tubes, please cough ASAP" by normal use of sensory nerves, and that channel doesn't just go away outside of emergency conditions. Sensory impressions have a similar enough format to emotional memories for somatization to be a thing: "my heart skipped a beat" as a description of an intense emotional reaction suggests the brain can send such information to the heart quickly, for which stress hormones and so on provide plausible mechanisms.
If I were designing a memory-backup function into the heart, and wasn't allowed to send any chemicals from the heart back through the blood-brain barrier at all, or even add side channels to sensory nerves, next obvious thing to try would be modulating the heartbeat itself to serve as a modem. Relatively high-bandwidth brain-to-heart downloads, then low-bandwidth, meticulously compressed "error correction" responses to keep the archives synchronized, particularly while dreaming. Sleep paralysis means actual blood flow demands are low and predictable, so almost the entire potential range between resting and maximum heart rate can be used to encode information.
Peer-to-peer music piracy is proof of concept for synchronizing large datasets through noisy, intermittent signals. Any good compression algorithm means actual signal would look almost exactly like random noise to someone who doesn't already know the format.
> His name is Tim—I think it’s Tim Leighton, but I’m not sure. I think of him as Tim L.
> She inhaled him. That’s almost a bit too on the nose. She didn’t know it at the time, but her donor’s name was Tim Lamirande. Spooky, right?
Tim Lamirande wakes up after surgery in an unfamiliar ribcage. There are some pipes, and the tools he's expected to use to maintain them, and an eccentric morse-code dictionary (heavy on physiology-related emoji, very light on abstract stuff like math), but not much else he can meaningfully interact with. He thinks he can hear someone else moving around in the next cell over, decides to try communicating by tapping on the pipes. Closest concept in the dictionary to "stuck here together" is "kissing," and closest to "outside this place" is "fresh air source," but he cobbles together a semi-coherent metaphorical explanation of the overall situation, https://bogleech.com/awfulhospital/795 and there's an excellent vocabulary for nuances of self-image and emotive intent, so that part's easy.
Then he tries spelling his name. There are codes for the Latin alphabet, but they seem to have been added as an afterthought, once all the short, easy code sequences were already spoken for. He manages less than one such letter per hour, back-shift during which they can freely communicate (without being distracted by work) only lasts about eight hours, and if he doesn't finish an unfamiliar word in one pass, neighbor gets the internal sequence all garbled, so... "Tim L." Sure, that's close enough.
With much respect, I think you’re missing the point a bit here Scott. I don’t think the author’s main argument is about memories of specific events or facts. They are taking a wider (and more useful, I believe) definition of memory as the state of a system that influences behaviour. That can include aspects of memory encoded in chemistry and biology. You might think of it as “vibe” memory rather than associative memory between concepts.
I agree if wouldn’t be useful to store associative memories like you describe in the heart because the brain is the organ that uses these types of memories in processing. However, processes like biological responses to stimuli e.g. fat storage when getting lots of calories , are processed in a distrusted way to some extent, and it would make sense for there to be state (memories) encoded throughout the body to manage that.
We do indeed see this kind of state (memory) passed from mother to child.
I think that one of the main takeaways of this article is that memory can be much more complex and diverse than things like episodic or associative memory.
For argument sake, you can ask the same about DNA, eg. how does DNA encode blue eyes.
I can imagine a similar complex system of encoding and decoding neural states to molecules.
The review is very nicely written and raises interesting points. But I want to severely push back on one point. The neuroscientific community is quite misrepresented here.
The author claims that the consensus is the strong SPM hypothesis. But no one believes this. No one. At least not among the neuroscientists who are actually biologists who work with individuals neurons. And I have talked to some real experts in the field, and (as a bit of an outsider) I have done academic research in the field. Not working with actual neurons, but building models of spiking neural systems that were supposed to be pretty close to biology.
Let me try a more accurate description. The human brain is pretty complicated. In order to understand it, we need to look at it at several levels of abstraction.
A) At the lowest levels, we try to understand the chemistry of how a single spike is formed.
B) On slightly higher levels, we may investigate how the input form other neuron excites another, without modelling each single spike on a chemical level.
C) Even higher, how patterns of spiking look like in an ensemble of neurons.
And so on, until we are at regions of the brains like the prefrontal cortex, or even higher up at behavioral level.
By the way, all this is not even covering yet learning, it is just describing a snapshot of the system.
If you are at level B, how a neuron excites another: you can't work with weights on that level. Weights will not explain what's going on. Everything is totally non-linear. When two incoming spikes arrive at the same time at the same dendritic branch, then the effect is much larger than the sum of the two. Especially if the one further from the soma (the blob at the center of the neuron) comes a little before the other. Timing matters a lot. Some people say that neurons should not be described as "integrate-and-fire" (adding up the inputs and fire when they are above a threshold), but rather as "coincidence detectors" (they fire when they get two simultaneous input, screw their weights). Everything is awfully complicated.
But then you want to abstract this whole mess. Because the higher level of abstractions are also important, and you need them.
What is the best abstraction of a neuron and its synapses? Perhaps first you only keep like 10 parameters per neuron and 2-3 per synapse. If you want to make it even simpler, then you keep 3-4 per neuron, and perhaps 1 per synapse. But what if you want to go even more abstract, what do you keep?
And then the consensus probably is: if you really force me to reduce everything to a single number, then probably a "weight" per synapse is our best bet. But it's a simplification.
And now, when it comes to learning: the strong SPM hypothesis claims that only weights change. On higher levels, this is a useful abstraction. But on abstraction level B it is so obviously wrong. Of course the neuron is plastic and changes! All of the time, and all parts of it, not just the synapses. The soma changes, the ion pumps change, the genetic expressions change, the timing of the incoming spikes change. Those are things that we have abstracted away at the higher levels, but they are there. And this does not even touch the formation of new synapses, the growth of the dendrites into new directions, the formation of new neurons. For what we know, even the non-neurons in the brain, like glia cells, may change and play a role. This is why I say that no one believes the strong SPM hypothesis.
Although the author did describe the strong SPM hypothesis as "it's all in the weights", I sorta took their meaning to be more like "it's all in the neurons", where perhaps weights was an (intentional or not) simplification of what that looks like. Would it change your stance if they had de-emphasized the "weights" part?
I could subscribe a lot more to "it's all in the neurons", yes. I would add some caveats that there are other surrounding cells that may play a role, but yes, "it's all in the neurons" is a standard hypothesis. But I don't think the author of the review means "it's all in the neurons" with the strong SPM hypothesis. Right after introducing the strong SPM hypothesis, they are pretty explicit about it:
"The crux of my negative review of the SPM hypothesis is this: cells are extraordinarily complex molecular machines, and there’s a lot going on inside of them that the SPM hypothesis implicitly neglects. We often abstract away most of the biophysical complexity of neurons, which are cells. As cells, they take up physical space, and can have weird, complicated shapes. They talk to other (not necessarily neural) cells. Each individual neuron has a complicated (gene regulatory) network inside it, whose complexity parallels that of many of our models of entire neural circuits. Do we really think that none of this complexity is involved in processes as complicated and multiscale as learning and memory?"
I mean, I agree, they clearly literally meant weights. But if I had to guess they probably didn't know that much about the details and were using a gloss they had heard elsewhere. Probably they now wish they had been more accurate now, because it wasn't that important to the point anyway.
This feels a bit uncharitable. The weight of a spike must be grounded in some molecular changes, and those molecular changes most proximally related to the concept of a weight will have their own supporting dynamics further down the chemical pipeline. The point of saying "its all in the weights" is that the relevant cellular changes for memory are that which modulate the sending/receiving signal strength. This doesn't render the cell irrelevant, but isolates the relevant cellular processes as it relates to memory formation to those functionally upstream of the weight abstraction. But in principle this could relate to a very broad spectrum of cellular processes.
FWIW, as a practicing neuroscientist, I definitely think people believe in the strong SPM hypothesis. Nonlinear integration of EPSPs tells us that simple integrate-and-fire models are wrong, and it's better to think of e.g. single dendritic branches as analogous to units in ANNs. And it's of course true that ion channels, receptor densities, etc. all vary as a result of experience, and this is orthodox neuroscience. But I still think it's coherent to claim that this is all *in the service of* changing synaptic efficacy, and in that sense consistent with the SPM. I think something close to this.
I don't know much about neuroscience but when reading the review I already had a strong feeling the reviewer was misrepresenting how much what they were arguing against was actually a consensus. I came to the comments to ask about this, so I'm glad to see someone has already answered.
> Some people say that neurons should not be described as "integrate-and-fire" (adding up the inputs and fire when they are above a threshold), but rather as "coincidence detectors" (they fire when they get two simultaneous input, screw their weights).
A) I've always found it hard to imagine that mental states might be encoded as anything other than electromagnetic waves.
B) I'd hazard that the "broken rock" definition of memory isn't specific enough. I think the most reasonable, broadest definition of memory would boil down to "encoding *correlation*" (as opposed to just "information persistence") . Information is only useful to the decision-making of an organism insofar as it draws correlations between events. (A while ago, I wrote about how the Correspondence Theory of Truth and the Coherence Theory of Truth are, in fact, complementary theories of knowledge [0]. Correlations would broadly fall under Coherence.) E.g. I can easily imagine ancient microbes inventing molecular ways to encode correlations between being poked and cellular damage, and then using this information to "decide" how to react.
----
Thus, if the most basic type of memory (in the broadest sense) consists of encoding correlation, i wouldn't be surprised to learn (from researchers, with 100% certainty) that neurons (at the synaptic level) were doing some sort of "Fourier-Transform-esque", "music of the body" frequency-matching in order to represent correlation-detection of incoming signals. Feels like a very natural building-block for mental activity, to me.
[0] https://fromthechair.substack.com/p/magic-runes-and-sand-dunes-the-binary
> How can we remember things for years with synapses that turn over on a time scale of weeks or less?
We don't understand it very well, but this question misses an important point: SOME synapses change very fast, in the order of weeks or less. Especially in regions like the hippocampus where long-term memories are NOT stored.
I don't think we know very much about the synapses that don't change. This is because they are hard to measure. You can manipulate animals so that changing synapses light up. But it's harder to make non-changing synapses light up. Also, it's easy to look into a mouse (uhm, if you are a top-notch neuroscientist) and then look into the same mouse a week later and look for changes. But it's much more expensive, and probably also harder, to make the same experiment with a year time lag instead of a week.
I think the standard model is that indeed some synapses change very fast. Fastest is hippocampus, where many things (not everything) changes within days, in some other regions it is more a matter of weeks or months, and probably some regions like the prefrontal cortex contain subsets of neurons and synapses that don't change in years or decades. This is where our long-term memory sits. Now, this model has not much experimental backup, because it's hard to test experimentally, but it doesn't have a lot of evidence against it either.
UPDATE: In a comment below, Ariel Zeleznikow-Johnston points to some experiments which HAVE followed synpases over months and found stable subsets.
https://open.substack.com/pub/astralcodexten/p/your-review-the-synaptic-plasticity?commentId=155384468
Until you can test it experimentally, it's all a just-so story. Materialists seem very attached to their just-so stories, and they argue about them as if they were valid explanations. The biggest elephant in the room is the assumption that the brain generates the mind.
This is a super important comment, thanks for sharing. A bit more context: when people hear "hippocampus" (HC) they think long-term memory, but the dominant model in the field is that memories are transferred from HC to cortex over a time frame of weeks, matching the empirical dendritic spine turnover results in both areas. That's why HC patients like HM were incapable of forming new memories while retaining their old (pre-operation) ones.
One nitpick: Connectionism is stronger than the claim that "all of the things that make us human—our ability to talk, think, reason, remember, and so on—follow from networks of interacting neurons, and changes in the strengths of connections between those neurons." Classical / symbolic cognitive theories (typically viewed as the competitors to connectionist theories) also claim that all cognitive functions are implemented via neurons in the brain. The difference is that symbolic cognitive theories view those neurons as implementing symbolic computation (like a computer, where, e.g., logic gates are implemented via transistors), while connectionist theories view cognitive functions as directly resulting from sub-symbolic, distributed, associative patterns across those neurons (like an LLM). See https://plato.stanford.edu/entries/connectionism/.
I know this is not central to the post, but seemed worth pointing out. (Source: I'm a PhD in cognitive science.)
If anyone wants a quick rating on the kookiness scale from a licensed neuroscientist*: the basic biological facts reported are not at all kooky, but the speculations and extrapolations are extremely kooky.
It is not at all kooky to say that single cells have "memories" of some kind, and that these are important, but not well-understood. This is universally recognized in neuroscience and not remotely controversial.
But these are cell-memories of cellular things, not people-memories of people things. It is extremely kooky to suggest that that single-cell memories could be responsible for the reported personality changes in transplant patients.
Kooky things are not always wrong. But they are kooky!
*Not a memory or cellular specialist; also not actually licensed.
I found it strange the review didn’t mention that hearts have ~40,000 neurons. What do you think of the claim they could carry some basic memories over? I’d link to a journal article, but it cites precisely the same examples as the review.
What’s your take on the epigenetic “memories”, like rats retaining aversions that their parents learned? There’s lots of skepticism because the mechanism is unclear, I’ve heard, and because many have taken the implications too far. (It gets pretty woo.) But could low-level predispositions be transferred, even if not declarative memories?
Honest questions: I find the subject interesting and don’t know what to think.
I'm not sure that the presence of neurons in the heart makes any difference. Neural activity only acquires "meaning" by dint of relationships with other cells. Brain cells are closely and recurrently connected with sensory and motor cells, so they are well-positioned to acquire "meaning" related to the outside world. Cardiac cells, neurons or otherwise, are mostly connected to other cardiac cells, so I'd guess whatever "meanings" they encode are mostly heart-related.
This doesn’t seem quite right. I might as well share the article I looked at (link below), because it references other apparent not-purely-heart-related functions of neurons in the heart, such as in emotions, and regulating the overall mood/ state of arousal of the body. Most of the “memories” transplant recipients report are not declarative memories, but flashes of light, tactile sensations, and other basic feelings with high emotional salience. Does it seem too crazy that the heart could retain the residue of memories that are salient for arousing the heart, either because that helps the heart react in the future, or because the brain is messy and doesn’t purely separate its functions? (The “Tim” and power ranger examples are more declarative however).
For reference, zebra fish have 100,000 neurons and can form long term memories.
https://pmc.ncbi.nlm.nih.gov/articles/PMC11061817/
Without getting too deep into the weeds, I'd say:
1) I don't think the comparison to zebrafish means very much. 40k neurons is a lot for a zebrafish but not much for a human. The heart may well need most of them just to coordinate cardiac muscle contractions.
2) In biology, everything is connected to everything by one route or another. Arousal and emotion are global states that can probably be "read off" every organ of the body in some way, but this doesn't tell you much about the function of any of those organs.
3) Organ transplants are a big deal physiologically and psychologically, so it makes sense that lots of weird stuff happens. The particular weird stuff may well have something to do with the particular history of the organ.
4) The truism goes, in biology the answer is always yes. I would add: yes, but in a boring way.
“Yes but in a boring way” is biology’s version of “it’s never aliens” in astrology 😂 Ok, these are reasonable points. I will hold my mind open to emotional processing in the heart to manage the body’s macro states (especially since 80% of neurons from the heart signal up to the brain rather than vice versa) - but with cautious skepticism.
But here's a possible "just so" story about low-level dispositions getting transferred. Say Bob is an anxious sort of person with a highly reactive autonomous nervous system. His cardiac cells have their gain ramped way up; it's ready to palpitate at a moment's notice.
Jill somehow gets Bob's heart in her chest. Before she was an even-keeled sort, but now she finds that her heart starts racing much more readily. Because the relationship between emotion and physiology is bi-directional, she gets more anxious. Now her personality is more like Bob's.
This sort of "memory" transmission seems plausible, but purely speculative. The epigenetic memory stuff I haven't followed closely; but I will say, if it were true, you'd think it would be a big thing in the field, and it has distinctly not been a big thing. My guess is that it hasn't replicated, but that's just a guess.
I have the same priors on the epigenic rat studies. Though maybe there’s a small chance people avoid studying it because it has a stink of woo on it.
Edit: oh look a retraction. Google’s AI says it was retracted for methodological weakness, not fraud, but anyways, not promising.
https://pmc.ncbi.nlm.nih.gov/articles/PMC12372920/
Yes, this is the main other scenario. I feel like the academic job market is brutal enough that people would roll the dice on an iffy area if it gave them a shot at a Nature paper, but I dunno!
Ok, while there was one retraction in this field of research, it seems it’s still active and not crazy, according to this recent paper that explores the mechanisms (below). This article gives a plausible-sounding account of what kind of memory we should expect via heritable epigenetics: “This process is akin to priming or acclimation in plants in response to various stresses like drought, heat, salt, irradiation, and pathogens.” (see section: “Epigenetic transcriptional memory.”)
So here’s a third reason you haven’t heard about it: it’s likely true but in a boring way (for newspapers I mean. For me it’s really interesting).
https://pmc.ncbi.nlm.nih.gov/articles/PMC10947848/
Reading the post I actually thought they were going to go with the theory that the heart neurons might transfer memories to the rest of the body's neural network just through normal predictive coding interactions, as something similar happens with llms where the behavior of one llm on totally separate questions can change just by being fine tuned on totally unrelated outputs of another llm with the specific behavior. See this great video by welch labs to learn more about the topic:
https://youtu.be/NUAb6zHXqdI
I think this analogy is useful in that it changes are likely to be weird and unpredictable. Something may change, but it's probably not the "transfer" of a coherent or even well-defined memory.
There's actually an interesting recent paper on the topic of cellular memory: Kukushkin et al., "The massed-spaced learning effect in non-neural human cells" (not sure if I can link it here), which basically showed that “spacing effect” in learning (training works better when it's broken into several sessions over time) emerges in non-neural human cells, like kidney cells. If I understand correctly, they used human cell lines and found that four brief, 10-minute-spaced "pulses" of a chemical signal (forskolin in this case) drove stronger and longer-lasting gene transcription than a single massed pulse, i.e., a cellular analogue of better memory with spacing. I've see some interviews with the lead author of this paper, and he argues, on this basis, that memory fundamentally depends on cellular machinery common to all cells.
The How the Hippies Saved Physics link goes to a "critical error on this website" page. I suggest replacing it with an Amazon (https://a.co/d/cSo6tgX) or Goodreads link (https://www.goodreads.com/book/show/11048785-how-the-hippies-saved-physics).
(Yes, I had to click the link for a book with such a name)
I love this line of inquiry if only for all the weird hypotheses it generates
* does eating meat from free-range animals cause you to absorb some of their... wildness? is that why some people like it a lot? does our mass-produced meat industry manage to produce meat that is missing some vital nutrient of "experience"? does eating factory-farmed meat lead to something like anxiety because you vaguely remember the horror of it?
* do blood transfusions confer some of someones's vitality to you? is this why some people are weirdly into them?
* is this why spiders eat their mates or their parents? to learn how to spin webs better?
* do babies absorb some of their mothers' memories through breast milk?
* does thinking that your memories live in your brain cause us to "ignore" our bodily memories, leading to some dysfunction? is this why bodywork-type therapies help, because they remind you to pay attention to the rest of your "mind"?
* is there some concrete experiential benefit to swallowing, um, semen?
no opinion on the veracity tho. just fun to speculate
Kooky hypotheses are fun! The problem is there's a massive "motte and bailey" element to this, and non-specialists might not pick up on the distinction.
Cells definitely have a form of "memory", but these hypotheses require a massive leap from "having memory of some sort" to "transmission of arbitrary information via transplantation or ingestion". These are super different things and the former is not evidence of the latter!
> does eating meat from free-range animals cause you to absorb some of their... wildness?
does eating vegetables make you more of a vegetable?
Yes, this is well-known in the carnivore community.
Context: I'm college-educated and work on software engineering that uses LLMs, so I'm vaguely familiar with concepts of synaptic weights from cultural absorption.
I got to section 4 of this review and realized ... "I hadn't thought much about it, but I had always thought that neuron strength, neuron pruning, and other qualities of the neuron itself might be important, alongside just the strength and existence of the synapse." ... Is that the distinction it's fighting against, the literal interpretation of SPM as only being about synapses? Colloquially I had understood it to be more then just synapses, even though I suppose it's in the name. Just providing a data point that some colloquial understanding never abandoned the ideas this review seeks to resurrect.
> Do we really think that none of this complexity is involved in processes as complicated and multiscale as learning and memory?
Natural selection would tend to eliminate brains that were too easily changed by so many processes. Consider how genes of large effect are disfavored relative to lots of genes of small effect, which better enable a target to be hit rather than overshot.
> People think that epigenetic changes like DNA methylation are responsible, since these changes can be fairly stable, and can be heritable
Those tend to be erased in mammals https://www.razibkhan.com/i/82002047/if-only-it-werent-for-that-blank-slate Again, there are reasons for natural selection to favor that.
> Relatedly, it seems unlikely that no organism could learn or remember anything before synapses existed. There had to be some other, simpler mechanism that worked before large-scale, synapse-based neural networks were common.
Why must that have been so? Why is it impossible for there to have been organisms without memory, until the capacity for memory evolved later?
> Maybe your memory of your fifth birthday party is in your arm, or leg, or heart
If that were the case then people would LOSE memories on losing a limb (or getting a heart transplant).
> Can you forget something about your fifth birthday party if someone chops your arm off?
There are many people who have lost limbs. Is there any evidence of them losing memories as a result?
Would you realize if you had forgotten the memory? Maybe this does happen but people don’t notice… or it just diffuses around everywhere such that you have a redundant copy in the other arm?
People with dementia forget things, and that gets noticed.
Good point, though I do think that the broader diffusion point still stands. If this hypothesis were true I’d expect important long term memories to get scattered all over and then randomly re-loaded periodically into synaptic memory. So an individual arm wouldn’t be specifically tasked with sole source of truth for the birthday party (presumably read speed on such a system would also be terrible…). I suppose it would accidentally end up as sole source of truth for a few things by luck of the draw, but since those things weren’t currently loaded into your random access layer anyways plus the damage wouldn’t be progressively worsening /systematic I suspect it would be harder to detect (especially if no one around you sees any prior reason to check you for memory loss). That said, I doubt this is actually a thing. I just find the mother-child counter argument stronger than limb loss.
If natural selection eliminates changes of large effect rather than many small effects, it should eliminate synapse-dependent memory systems and favor highly distributed multiformat memory.
he's saying that the causal relationship between any single gene and its phenotypic expression is small. Not that the macroscopic instrument is resilient to damage.
(1) No he isn't. (2) That would still cut *against* the claim that we should expect memory to be concentrated in one method, not for it.
Like a few others here (demost_, maybe others), I am a neuroscientist, I teach classes in learning and memory, and do research in learning and memory. I have been in this field for over 40 years (yes I am old) With that background I believe I can speak to how the synaptic plasticity (in general) and synaptic weight (specifically) hypothesis is viewed and taught in the field. It is a strawman to claim that there is a strict doctrine that synaptic weights are both necessary and sufficient to account for learning and memory that is widely believed and taught.
How does the field treat synaptic plasticity? It is indeed typical to teach about LTP and LTD, but only as one mechanism. It took many years for the field to become persuaded that LTP really was a major mechanism underlying memory, and most courses that have time to do more than just mention it do go into the evidence for LTP being a memory mechanism, as well as what we still don't know about it. In addition most courses acknowledge other interesting examples of plasticity, including white matter plasticity, epigenetics, network dynamics, and even molecular mechanisms! Researchers presenting talks in which they have focused on synaptic plasticity mechanisms in my experience typically acknowledge that there are alternative possible mechanisms as well. I haven't heard anyone in the field claiming that synaptic weights are necessary and sufficient for learning.
I also have not had the experience that AI researchers starting really pushing the idea of synaptic weights underlying learning and memory in the 2010s. The idea was around from much longer, including in the 1980s when I was an undergraduate and connectionism was becoming stronger. At the time there was always an acknowledgement that calling these connections synapses was just by analogy and was never intended to be biologically realistic. I even wrote a paper in undergrad about what we knew then about biological synaptic plasticity and implications of that research for connectionism. I still find that AI researchers who I interact with are quick to acknowledge the lack of biological grounding and the limitations of the synaptic weights view.
With that in mind, the review falls apart for me. The second half argues against a view that may be held in some non-scientific circles but is not an accurate characterization of the field, yet claims to be arguing against a dogma held in the field. The first half is way out on a limb -- many fascinating anecdotes that I did enjoy reading, but probably NOT the best cases to make for a possible molecular basis for memory given the complexity of some of the declarative memories described.
Hey CS, may I ask what are your favorite/seminal papers around these alternative examples of learning and memory? I'm not classically trained in neuroscience so I know of LTD/LTP but not around white matter plasticity, network dynamics, or molecular mechanisms.
Thanks!
Here are some papers that really made me think -- I don't promise that they are the best, just the ones that I read and thought about.
For network dynamics:
Arnsten, A. F. T., Paspalas, C. D., Gamo, N. J., Yang, Y., & Wang, M. (2010). Dynamic Network Connectivity: A new form of neuroplasticity. Trends Cogn Sci, 14, 365–375. https://doi.org/10.1016/j.tics.2010.05.003
For molecular memory:
Gallistel, C. R. (2020). The physical basis of memory. Cognition, 104533. https://doi.org/10.1016/j.cognition.2020.104533
For white matter:
Sampaio-Baptista, C., & Johansen-Berg, H. (2017). White Matter Plasticity in the Adult Brain. Neuron, 96(6), 1239–1251. https://doi.org/10.1016/j.neuron.2017.11.026
Thanks!
I can easily believe some transplant stories while others seem totally fake.
An actual memory of the previous host of an organ? No way.
Something that could be cell-protein-based? Much more plausible. If a transplant stomach previously had a negative reaction / trained biome for a certain diet, that could carry over. Similarly, if a heart always raced when the body sensed meat before, it could again. I could absolutely understand some subconscious patterns being carried over.
Think about it another way - if you received lungs from a long distance runner, you are receiving those lungs with the “memory” of strong performance. Or - just more muscles in the way that ‘matter’. We convolute what ‘counts’ as memory, but I’m not surprised it appears sometimes in unintuitive ways - reacting to foods, other sensations. Feels plausible.
For those muscle growths or pathways to transmit some kind of memory to the brain… seems very unlikely. I could believe it for fruit flies that do not rely on cognition much anyway (more of light/smell seeking robot) but for humans, more likely to just be regurgitating facts they were told that then “feel right”. Hell, I have a memory of flying to Disneyland because my parents told me a story about it when I was a young child - they didn’t realize it was my brother in the story but it’s still in my long term memory. I’m sure a similar effect is outlining the more vivid stories.
That’s my thought too. Even things as disparate as sexual orientation I could believe to have some sort of cellular basis in the organ. But things like names, or episodic memories of the moment of death, seem very far-fetched, even when considering a broader range of mechanisms than pure brain neuron synaptic connections.
This seems especially true when it's not that difficult to believe that people might have been discussing the donor of the organs when the transplant patient was not conscious/aware enough to fully remember the discussion, but picked up part of it.
I could see some of those more vivid stories still coming from cellular memory. Someone drowned, and their heart learned to associate being underwater with bad times, so now the person has that reaction to water. Combine it with a dream on the topic, and I could easily see that physiological response getting jumbled in with memories.
""From a modern vantage point, it’s clear that if groups like the Tupinambá practiced cannibalism for the purpose of acquiring aspects of their enemies, they were misguided. Maybe they could get some protein by doing this, but they certainly couldn’t acquire any of the things that make us fundamentally human, courage and bravery included. ""
To state the obvious, they could by eating immediately acquire certain *negative* aspects of the enemy. i.e. any communicable diseases.
Any cannibalistic pre-medicinal culture would very quickly learn "don't eat the flesh of a diseased person"; then "actually, don't eat any sickly or weak person"; which naturally evolves into "prefer to eat the strong and mighty warriors".
It doesn't feel like any explanation involving non-synaptic cellular memory is needed.
Oh, this process of elimination makes sense, and I'd never thought of it before.
The heart transplant stuff reminds me of something I read a while back that claimed that the heart has a surprisingly high number of neurons in it, and that the count is high enough that some scientists think that it may play a role in cognition, that the brain may not be the exclusive seat of thought as we've believed for a long time.
The heart transplant stories would appear to provide some evidence in support of that theory.
Just found an article on that: https://pmc.ncbi.nlm.nih.gov/articles/PMC11061817/
This review reminds me of the work of Eugene McCarthy. Much of the text defends a weak but reasonable hypothesis, "memory encoding is not exclusively performed via synapse weights" or "cross-species hybrids have a greater role in evolutionary history than conventionally thought", all to arrive at the conclusion that humans were descended from pig-chimp hybrids.
You're going to have to do way more work if you want to convince anyone that there is anything in need of explaining about the transplant stories. Half-hearted (ha) disclaimer about psi aside, the review takes for granted that there's something weird going on about the heart transplants, but paranormal forums are stuffed with equally "plausible" tales of reincarnation, ghost sightings, telepathy, and everything else, many of which rely on the same formula of unverifiable, irreproducible personal accounts of remarkable coincidences, accompanied by some suggestive hand-waving.
Some of the stories presented in this review aren't even impressive at face value. What is supposed to be the connection between the college professor's nightmares and the drug bust? I guess we're meant to imagine that being shot involves light?
Unlike the cannibalism, which I'm interpreting as a lighthearted (ha) framing device, the author appears to wholeheartedly (ha) believe in the phenomenon of memory-transferring organs. In absence of good reason to agree with them, this undermines the credibility of the rest of the text.
Another example of the author's unwarranted confidence: "How can we remember things for years with synapses that turn over on a time scale of weeks or less? Maybe we can imagine neural circuits playing an elaborate game of hot potato, and constantly moving information around to prevent synaptic turnover from destroying it. But this explanation seems problematic given the orders of magnitude that separate hours and weeks from decades."
Immune cells also turn over on a time scale of weeks, yet immune memory lasts decades. There's no serious effort to substantiate the claim that synapses can't hold memories because they turn over too fast, just a quick dismissal that mimics logic. This review is peppered with these rapid sleight-of-hand tricks in favor of its central thesis. Fun read, but not convincing.
I'm not sure why you call it a sleight-of-hand. Sure, there are other explanations for how a system with unstable components can itself be stable, but proposing one where the components themselves are more stable seems reasonable.
Is the author's alternative hypothesis a "reasonable" suggestion in the abstract? I guess it could be, but that's hardly the point.
The author fails to provide a compelling reason to doubt the claim that synapses hold long-term memories. They make a broad, unsubstantiated appeal, contradicted by the facts of other theories they endorse in the review, then assert that the claim is "problematic." That's what I'm calling a rhetorical sleight-of-hand.
The foregrounding of the organ transplant cases is an odd choice here. I ultimately found the review pretty convincing as a case for "synaptic weights aren't the whole story when it comes to memory storage." But the organ transplant cases are the weakest evidence in the whole thing, and I still think they're probably not real even if the author is right about the SPM hypothesis being wrong/incomplete.
I disagree. They're intriguing case studies which draw us into the possibility of CPM and make us wonder whether they could be correct.
There is a really obvious fact (at least, I think it's obvious, but maybe not!) that puts this discussion in a different light: the synaptic plasticity & memory hypothesis could be true of neurons without the restriction that the neurons be in the brain.
Almost a full percent of the neurons in the body are outside of the brain (mostly in the enteric nervous system (ENS), maybe half a percent) but also in other organs, including the heart. Is it crazy to imagine that those neurons form memories, along with those in the brain? My guess would be that those memories would be affective rather than strongly "cognitive," but that could fit with many of the transplant stories.
I think these transplant recipients do not need to lie - all it takes is that they've overheard something about their donor's character and took it as a cue to snap into a different attractor state. It's not like people are born to be vegan or meat eaters - it's a choice. In my case I've decided one day I want to be a vegetarian and stopped eating meat. Before I ate meat but vaguely felt that this isn't entirely right thing to do, even though it's tasty, but was to lazy to change my life. I needed a nudge. In my case it was that catering in my company made it very simple to switch to different diet, so I just did it. There's always some cost to switch mental model and identity. Maybe people who were a bit on the fence and thanks to the transplant of heart now got a socially acceptable way to change their habits used that as an excuse to try something different. Or maybe their mental model was compatible with the dancer spinning left or right but not both at the same time, and the event just let them switch to the other attractor.
Also, stories tend to get exaggerated and embellished over time. In some cases, we can even see that in action, like when the people conveniently wrote down accounts and the early accounts are a lot vaguer and more plausible than the later accounts, even from the same person.
One would also think that the receiving a hear transplant might be a disturbing salient reminder that you are made of meat. After all, your meat was taken out of you and now you have meat from a dead person inside you. Such a thing could easily turn one off from the idea of eating meat going forward. Honestly just reading about these transplant stories someone turns me off to the idea of eating meat.
But also, I've eaten enough steaks in my life that one would think I'd wake up screaming every night from nightmares about the slaughter-house, if memories transferred based on what you eat.
The content of this review is interesting, but it makes a couple of major rhetorical mistakes.
First, the writer comes off as a cheerleader for alternatives to the strong SPH. The "review" would be more interesting if it tried harder to engage with the questions raised by alternatives.
Second, it goes down the rabbit hole of physicalism and the philosophical meaning of knowledge when this is utterly irrelevant. Yes, there are valid senses in which any (non-maximally-high-entropy) arrangement of matter "remembers" information, but the actual question here is whether and how the information in question is available to your conscious thought processes (as e.g. immune memory presumably isn't).
"I can’t convey how much I hate this kind of argument and find it stupid. Either you believe everything is ultimately elementary particles obeying the laws of physics, as physicalism posits, or you don’t. If you don’t, fair enough, but I’m not sure how you’d falsify your position. If you do believe this, then thoughts, feelings, and so on are just useful ways of talking about a complex physical system."
Any theory of consciousness would have explanations for the *feeling* of experience, i.e. the hard problem of consciousness. Then, we can just differentiate conscious memory, which does require such an explanation, from immunal memory, which does not. This different will be a feature of the theories themselves and are not just "useful ways of talking about" things.
*do* we know that there's no conscious experience of immune memory?
I loved this whole thing, myself. I didn't take the stories at the beginning as a key part of his argument, but rather as an entertaining way to frame the topic.
I'm no neuroscientist and so hopefully someone will help me understand. If we have neurons throughout our body, which we do, why isn't it plausible that memory could be stored in those multiple places as well?
And then it's surprising to me as a total outsider that we seem to know so little about the actual biological process of learning and memory. This synaptic model seems impoverished and the generally accepted idea, I gather, that there are probably some other molecular processes involved but we have absolutely no idea what they are... is that really the state of what is known?
I naively thought that there was more substantial knowledge under the idea that neurons that fire together wire together. But I'm getting the picture that we don't?
> If we have neurons throughout our body, which we do, why isn't it plausible that memory could be stored in those multiple places as well?
It might. But so far we haven't found indications for that. In contrast, we *have* found indication that the brain is involved in recalling memories. For example, there are neurons which get activate when subjects see a picture of Bill Clinton, or see the name "Bill Clinton" written somewhere, or hear the words "Bill Clinton", or even get some indirect clue which reminds them of Bill Clinton.
Theoretically, it *could* be that the concept of Bill Clinton is really stored in your heart, and only activates the very specific neurons in the brain upon recall. But we have not found indication for that (though to be fair, I doubt that someone has looked hard), and it would be more complicated than a system which stores the memory directly in those neurons in the brain.
> I naively thought that there was more substantial knowledge under the idea that neurons that fire together wire together. But I'm getting the picture that we don't?
We know a lot about the chemical details of how this fire-together-wire-together works. We also know a lot about the local "rules" which govern in which situations synapses become stronger and weaker, and also a bit about mechanisms and rules that change other parts of the neuron. We are pretty sure that this neuronal plasticity is involved in learning, because there are dozens of little puzzle pieces which point into that direction.
For the big question of how long-term memory is formed, we also know some things. For example, that (as far as we understand) our memories are first temporarily stored in the hippocampus, and then during sleep transferred into long-term memories in the prefrontal cortex. But we know very little about how the transfer of memory works in detail. Well, we know a few things, like that the brain replays some memories from the day during sleep, and this seems to be related to the transfer. On an abstract level, this already gives a nice coherent picture. But, which neurons exactly are chosen to store the memory in the long run, how the heck does the brain make sure that it's so nicely embedded that it can be recalled at will? We don't really know.
A disclaimer: I was talking about episodic memory: whom you have met today, what was in the news, what did you have for breakfast. There is also implicit memory (how to speak, how to play the piano) that forms by tons of repetition and probing and probably in a completely different way. We also don't really know the details of this, but it seems a lot less mysterious. Learning by tons of repetition and small changes each time is exactly the stuff that also works elsewhere, for example in LLMs.
Thank you so much for taking the time to explain this!
The role of sleep in all this seems very interesting. Is there research on the role of REM sleep to that process? I think there is as I ask that.
I imagine all this connects over to Alzheimer's research.
With the different roles of REM sleep versus SWS sleep (slow wave sleep, "deep sleep") we are going into terrain with less solid foundation, but there is something like a standard hypothesis.
I mentioned that the brain replays some events from the last day during sleep. These replay events occur both in REM and SWS sleep (and even awake at rest). But the replay is much "cleaner" in SWS sleep. Probably this is the part where the memory is transferred from the ever-changing hippocampus into long-term storage in prefrontal cortex. For REM sleep, replays often occur in variation or together with other patterns. The hypothesis is that the brain uses REM sleep to combine the new memory with other memories and to integrate it into the general knowledge landscape. So when you wake up and suddenly you understand something much better than the day before, probably this happened in REM sleep.
For the connection to Alzheimer, this part is not mysterious. Alzheimer causes neurons to degenerate and die. Cells in the hippocampus are particularly affected, together with some other regions. I don't really know why the hippocampus is especially affected, but my first guess would be because it is the most plastic part of the brain, where many new neurons are born and connections change very fast. Since memories are first temporarily stored in hippocampus before they are transferred into long-term memory, it is not surprising that a sick hippocampus leads to problems with forming new memories. And that is exactly one of the early symptoms in Alzheimer: patients have problems forming new memories, but old memories are much less affected, or become only affected at later stages of Alzheimer.
So it is not mysterious why Alzheimer leads to exactly those symptoms that we observe, it fits nicely into our big picture of memory formation. The big discussion around Alzheimer is: WHY do neurons degenerate and die? This is what the finalist review #3 and the recent guest post here at ACX were about.
Thank you again, totally interesting!
Enjoying your knowledgeable replies, I had the thought that Scott could create a corner of his Substack where people get to pose questions, either for advice or just about things they'd like to learn about, and people with knowledge to share could offer it.
The readership here has such a wide range of expertise, it seems like it would be fun to create a space to share it around more. Obviously with no pressure for anyone to respond at any depth to anyone else, just as one is interested or has energy to.
(I think I feel a bit sad about so many people going to ChatGPT for answers that are kind of common denominator answers instead of to other humans who have all this rich color and texture.)
That's an interesting idea!
I had exactly the same feeling with the guest post on Alzheimer. I find it just awesome that Scott has so often someone in the readership with this deep level of knowledge for a random question.
This year I published the first formal survey of neuroscientists on their views on the physical basis of long-term memory.
https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0326920
There is disagreement about the exact biomolecular mechanisms that encode information (specific receptor configurations? biomolecule modifications? just binary synapse strengths?), to be sure. But it's also the case that ~70% agreed or strongly agreed that "Long-term memory is primarily due to synapse strength", and only 10% disagreed.
A few key pieces of evidence for the synaptic encoding hypothesis that this piece neglects:
1. While the proteins that make up a synapse turnover on the order of days, the actual synapses themselves can be extremely stable in a ship-of-theseus style way. Yang (2009) demonstrates this - they teach mice some motor tasks, watch the dendritic spines that form, and note that a population of them are stable and still present months later.
http://www.nature.com/doifinder/10.1038/nature08577
2. In some studies, selective erasure of synapses has been able to delete one particular memory while sparing others. In Hayashi-Takagi (2015), the authors teach mice to cross a high-rope and also to balance on a rotating rod, while observing which synapses form during the learning of each task. By using optogenetics to erase one of the sets of synapses, they can force the mice to forget how to do one of the tasks while sparing their ability to perform the other.
https://www.nature.com/articles/nature15257
As per the words of some of the leading researchers in the engram field, “There is a clear consensus on where the memory engram is stored—specific assemblies of synapses activated or formed during memory acquisition”
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4874022/
I feel like a lot of the weird examples from the first section might well be explained by survivor bias. The guy who got a heart from someone named Tim and then has dreams about George, or no different dreams at all, naturally never shows up in the literature.
And the kid who intuits that his donor was a kid about half his age (never mind that that locution seems awfully stilted for a five year old) but was “never told” about that — well, maybe he wasn’t and maybe he was, but even if not, little pitchers have big ears.
I’ll accept that these anecdotes may have been meant more to get our imagination flowing than as any sort of proof, and on that score they are entertaining enough. But I sort of have to classify them as analogous to near-death or out-of-body experiences, or (if you’ll forgive me) UFO sightings.
one confirmed phenomenon where the thing you eat directly affect your memory:
Mad cow disease [Bovine spongiform encephalopathy (BSE)]
we all know that it is caused by a prion [protein] that survive being disgested, crossed the blood brain barrier and caused significant changes to neuron connections in the brain and therefore affect memory/personality/thought.
it is therefore possible for other proteins to do all the same thing and the reasons we don't notice are
they are rare
when they happened they are harmless/within expectation
so only when it is extremely harmful that we bother to look for a cause and find something.
^this
Prions don't change your memory, they interfere with the proper functioning of your brain. This is like saying that pouring sugar in your car's gas tank changes its performance therefore the "gas provides energy" theory of automotive design is wrong.
ahh,
first, it is changing memory according to the stricture laid out by the poster, [i.e. a broken rock has the memory of the thing that broke it.]. But yeah you don't buy it, I don't buy it either.
Second, Scott was arguing above about how poteins tend not to be able to survive being disgested, crossed the blood brain barrier and caused significant changes to neuron connections in the brain and therefore can not be use to store nor transfer memory. I just want to counter that with an example.
lastly, if there is a protein that that survive being disgested, crossed the blood brain barrier and caused significant changes to neurons in the brain, does that not increase your prior that there are more of them with maybe less pronounced effects that we may have not been looking for?
"There are things that affect the functioning of the brain" is very far from "there are things that can manipulate memory in arbitrarily complex ways". Pouring a glass of water on a laptop can break the laptop but it can't reprogram it in a different-but-still-functional way. Destroying function is much simpler than creating function.
I am glad to see us agree on almost everything and now we can narrow our disagreement [if there is any]
almost none of my statement below would be sourced but I assume it would be trivial for you to confirm with google. I thank you for your indulgent.
I agree that there is probably epsilon chance that "there are proteins that can manipulate memory in arbitrarily complex ways"
I hope that you agree that there are memories coded in DNA and the resulting RNA and protein, we call those memories instincts. That is anything a newborn baby can do immediately after being born [e.g. how to breathe, how to suck and swallow]. More complex memories are also coded, most prominently sexual preference. I hope we agree that many instincts and preferences are coded/mediated by protein? and if you have preference of something you have memory of it? what does it mean to have preference for something you have no memory of? [and ehmm, can you turn someone gay by injecting some proteins [?hormone] into their CSF or do you need to be exposed to it at certain development stage?]
And I am pretty sure that there are studies confirming that some/many memories are encoded in chemical signals, probably a protein of some sort.
wait actually do we agre that any protein that affect the neural circuit affect memory? or is the bar is being able to encode abitrary memories and inject it into another person so they can experience it? if it's the second thing then yeah I don't think anything like that exist.
but if we count preference and thoughts patterns then many proteins demonstrated this property. example: alpha-synuclein, implicated in parkinson disease, can cause depression [mostly through dopamine disregulation] and can cross blood brain barrier.
no time to read article. may I eat your heart instead?
The evidence for this thesis seems roughly on par with the evidence for past lives.
I’ve wondered about this for many years:
For memory retrieval (eg. where did I put my keys) many of us roll their eyes, which seems to help memory.
Similarly, smells help us evoke memories from long ago.
Do pro athletes know in their brain how to be good at sports vs. how much is in their body?
To me it is obvious that the memories are not only in the brain, but the brain is the last part in the signal-chain and the evocation of a memory is dependent on the set of signals that arrives in the brain (which are dependent on our physical sensors and nerves).
This is also why I believe uploading one’s brain would not do much since the brain probably couldn’t identify itself, without the sensors it’s used to/trained on.
Thanks Scott for everything 🙏 Hi from Austria.
Yes, I've always felt it would be agonizingly disorienting for a disembodied brain or brain emulation to exist, even if that were possible. There's so much of our experience of living that resides in our bodies.
Here is a journal article about how the heart has 40,000 neurons in a little “heart brain.” The article draws on the same stories as in this review and suggests memories may be transferred with this mini-brain in transplants. I find it strange the review did not mention the role of neurons that were transplanted.
https://pubmed.ncbi.nlm.nih.gov/38694651/
These types of reviews are insidious. A maverick dares to challenge the scientific consensus in some highly technical field. It makes for a compelling story. They are usually filled with details, impressive technical knowledge, a compelling narrative, and they make the reader feel special for gaining such clever insights that those pesky scientists missed.
I know I'm being a bit too harsh, because I actually thought this review was on to something. That is until I did some further reading where everything began to fall apart. In particular the supposed smoking gun that synapses turn over in weeks to months and therefore cannot store long term memories. Except that many synapses do last years as many other commenters have pointed out. The fact that the author either didn't know this, or didn't report this is alarming.
Seconded, for all that I enjoyed reading it.
The search for intellectual novelty has led to "hey guys, maybe cannibals had something going on there because SCIENCE! No wait, it's probably wrong."
"This controversial thing, often from the past, may actually have truth to it due to SCIENCE! One weird trick can explain everything!"
I think my brain is just physically revolting against the idea at this point. Like this particular article just hits the "parody" button in me, even if it doesn't intend to.
Best read of the year for me! As a CBT therapist that recently moved to Sensorimotor, I always wonder how we actually (phisycally) remember things and embody memories. After reading this post, now I know to know even less than before. Fascinating! :)
That was a bad idea to start reading this during breakfast. I couldn’t stop though
No mention of prions?
Took me several sittings to slog through. Dry like burnt toast, humour didn't land, absolute thicket of links and referenes. Which I'd want in an essay attempting to prove a particular point, but feels out of place in a not-a-book-review...the focus needs to stay on the thing being reviewed, with a minimum of other distractions and digressions. Just enough to establish the floor, as it were. Bold but frankly stomach-churning choice to lead off with cannibalism stories, and then get a redux halfway through, then a zinger at the end. This might be "insanely cool" for some people, for me I regret reading while eating.
I notice that only one of the Four Axioms got (arguably) refuted. I notice that a fair amount of the supporting evidence is not itself well-corroborated. Like yeah, you can interpret "half the mouse brain studies succeeded!" as optimistic, but I'm thinking "wow, a 50% failure rate, that sounds like random noise". Which is weird coming off a tangent about McCannibal's results being hard to replicate! Same thing with the transplant just-so stories, they're too cute by half to not trigger my epistemic red flag detector. It's eminently clear, from stuff in this review and just generally anyone familiar with basic psych (so like...most LW/ACX readers), that human memories are highly fallible, and we love overfitting events into plausible narratives. Moreover, it just cries out for base rate comparison: how many transplantees *don't* experience weirdly specific changes that correlate well with donor foibles? (Self-reported, no less!) The literature on anesthaesia is also pretty clear that it does all kinds of systemically weird shit to people...given all the semi-plausible hints here that memory can be stored and transmitted throughout the body, why wouldn't we expect some eerie concordances to arise by random chance?
On and on and on. It's a shot across the bow, to be sure, but with enough holes and weak links that I'm just not comfortable writing off the discrepancies as "well of course if it got the same amount of funding and hype as Big Brain Dogma then we'd fill in the blanks". That's veering into Fully General Counterargument territory.
I've been circling around the CPM for the better part of 10 years now. In addition to the many other excellent comments below pointing out flaws in the argument, I'd like to mention a few other important issues, some of which go beyond the article itself but might be of interest:
1. Readout. I can ask you to define a word, recall your most recent birthday party, or perform some known motor skill, and you can do so instantly. DNA transcription/RNA translation occurs on the timescale of at least minutes, and sometimes days. If RNA (or protein post-translational modifications, or histone acetylation, or whatever) is the substrate of human memory, how does this readout process take place?
2. Developmental biology. Neurons contain a dazzling array of cell surface proteins that help determine which two neurons make a synapse in the first place (these have names like "cadherins," "neuroligins," etc.). When they make a synapse, lots more intracellular machinery is recruited to help stabilize it. If you do things that naively "disrupt" the synapse (perhaps this is what is happening during metamorphosis, although I have no idea), these proteins still stick around and could in principle allow the same circuit to reform. Is this "molecular memory"? I wouldn't say so, considering that formation of the synapse is both necessary and sufficient for the memory to be expressed.
3. Theory. If you go into the history of these arguments, particularly Randy Gallistel's entertaining "Memory and the Computational Brain: Why Cognitive Science will Transform Neuroscience" (2010), a key hangup is the claim that neural networks are "write-only" memory systems. Sure, experience can alter some weights, but how can you possibly go in after the fact and read out that change? How can you store and retrieve the value of some variable? In my opinion, the success of artificial neural networks has destroyed this objection, so the argument has had to be resuscitated in a different guise. Mechanistic interpretability has taught us that it can be very difficult to trace the causal path from input to output in a massive, densely connected system, but the memory is nonetheless there.
That said, I have lots of respect for Sam Gershman and many of the other folks mentioned in this post. One of my favorite talks ever was an internal neuroscience seminar at Harvard where Gershman shared the thoughts that would later be included in his 2023 review article -- only for the audience to strenuously object to essentially every claim. For those thinking that academia has descended into a morass of conformity and groupthink, I think this provides a welcome corrective.
Your point 1 was my big question after reading this post. The timescales just don’t seem to add up, but I’m ignorant of the subject and can’t form an educated opinion of whether it’s an insurmountable obstacle or not.
A small nitpick about the Tupinambá episode:
"Among other reasons, he claimed that their God didn’t like the idea of him being eaten."
The following quote: "Staden immediately insisted that this in fact was the case: 'I told him yes, my God was angry, because he wanted to eat me.' "
'Their god' vs 'my god' is not hugely relevant to the review, but it put me in mind of the relative position of religious outlook in that episode. To Staden, their god might as well be *his* god, too, in essence. Auspicious slip-up?
In grad school we had a guest seminar from a researcher (sorry I forget who) who had developed techniques that were successful in helping paralyzed patients regain mobility. The technique involved having the patient think about moving their limb and simultaneously applying shocks to the gray matter in the spine, essentially training these regions to take on roles from the damaged motor cortex. This was the first time I heard that neural processing doesn't just occur in the brain. I don't have a biology background, do we have a good understanding of what the roles and capabilities of the extra-brain neurons are?
So if you cut yourself and develop a scar, you’ve “stored information” outside of your synapses.
While plausible, this is significantly less interesting of a phenomenon than the acquired memories discussed in the intro.
Anyways — I think the control group is out of control.
I think this review simultaneously
- extremely strawmaning positions of mainstream science (for example, no one thinking about actual neurons as bunch of weights it is an abstraction useful for large collectives of neurons)
- ignoring clear but maybe hard to formulate distinction between memories and state of system affected by previous events. If my arm get cut of fact that it’s still missing is not a memory mechanism. If my cells have epigenetic markers from malnutrition and now store more energy it’s not a memory it is a state of their metabolism. Of course it is a spectrum but I think we can’t ignore clear differences between information stored on hard drive and in the current processor state
- ignoring elephant in the room: mechanisms of transition from information stored in single cell to actual behaviour
- not providing reasonable questions about worms and sea slugs experiments - their learned behaviour is extremely simple and can be encoded by several bits. It is quite plausible that single protein amount affecting light sensitivity is enough to “encode” this behaviour
- not doubting memory from transplant observations. I see unnoticed information about donor leaking and reaction to it under stress much more plausible than actual memory transfer. I do believe that transplant can affect recipient behaviour in an unexpected ways because brain isn’t isolated system and some reactions can be casually connected to signals from other organs but I don’t believe in any information transfer
This post strongly reminded me of bioelectricity, i.e. the idea that non-brain cells also have the kind of electrical circuits as brain cells, just way fewer of them, and that they form a sort of low-level intelligence responsible for assembling cells into larger creatures. You can learn about it in Prof. Michael Levins Youtube Channel, for example here: https://www.youtube.com/watch?v=s-SL8qy-6kA
Perhaps such networks somehow interact with the proper intelligence in the brain?
The author can afford to be more precise with their definitions of memory. There is a loose definition of memory, where some arbitrary type of information is stored. Biochemistry is trivially implicated in this kind of information storage, as information is stored in rna/dna/etc. However, they conflate this loose definition of memory with "memory", the rememberance of complex visual and linguistic information (such as names and images). The problem is, they begin the review with anecdotes on the latter kind of memory, but all of their scientific evidence points towards the former, with no effort to bridge the two. I can quite easily believe that biochemistry stores some behavioral tendencies via epigenetics, but biochemistry routinely storing the names of people and places is much harder to believe (and evidence of this kind of memory being stored is not really included in this review).
> The most popular quantitative models of the brain, artificial neural networks (ANNs), assume that the strengths of the connections between neurons, or “weights”, completely determine how networks behave. These are the things that are assumed to change during learning, maybe through an algorithm like backpropagation, or maybe through something that looks more like Hebbian learning. If these models learn something, or store a memory, it has to be through changes in weights.
> In AI, too, weights are king. The term “weights” is even used as a synonym for “model parameters”. Weights are the numbers that completely characterize what a state-of-the-art model has learned through expensive training, and these days they can be ferociously protected trade secrets.
This is a popular oversimplification but technically incorrect - two main sets of parameters are learned during training. One is the connection weights, which determine how neurons influence each other, but there is a second, oft neglected set - the biases, which determine the activation threshold for a given neuron.
This article posits that mitochondria are the primary excitable unit the brain, not synapses. Might be a worthwhile addition to the discussion.
https://inference-review.com/article/the-excitable-mitochondria
Thank you. I have no view on the correctness of any of this nor am I competent to form one, but I greatly appreciated the ride and I did learn a lot, including some wholly new (albeit low probability!) hypothesis!
With a lot of respect, I think you’re missing the point a bit here Scott. I don’t think the author’s main argument is about memories of specific events or facts. They are taking a wider (and more useful, I believe) definition of memory as the state of a system that influences behaviour. That can include aspects of memory encoded in chemistry and biology. You might think of it as “vibe” memory rather than associative memory between concepts.
I agree if wouldn’t be useful to store associative memories like you describe in the heart because the brain is the organ that uses these types of memories in processing. However, processes like biological responses to stimuli e.g. fat storage when getting lots of calories , are processed in a distrusted way to some extent, and it would make sense for there to be state (memories) encoded throughout the body to manage that.
We do indeed see this kind of state (memory) passed from mother to child.
I think that one of the main takeaways of this article is that memory can be much more complex and diverse than things like episodic or associative memory.
I'm surprised the author didn't mention Roger Penrose's ( the physicist) proposal that the microtubules in neurons store/delete information & are the basis of consciousness.
> If even a quarter of them are true …
> But if any of these examples is at least kind of true …
These two phrases are dubious. I could list many historical instances in which there have been lots and lots of purported examples of a phenomenon, but actually the phenomenon doesn’t exist, and every last one of those purported examples turns out to be complete BS, no exceptions. This kind of thing happens all the time, unfortunately. Examples include UFO sightings, or Scott's old posts on parapsychology and 5-HTTLPR, among many others.
(Tbc, maybe in this case the phenomenon is real. Or maybe it isn’t. I’m just making a narrower claim that “hey look at how many striking examples I just listed!” is a weak argument in itself.)
Review of the Review
When psychologists refer to one model of neuronal learning as a "mere-casual mechanism", and contrast this with a hypothesized mechanism that isn't mere causal, they don't mean that they think memory operates "acausually" or without the interaction of molecular elements. They mean that causal functioning is direct and mechanical, and involves no symbolic information. We can define symbolic information as an arrangement of patterns elements in some system that represents conditions outside of that system. Information becomes symbolic when it can be rearranged from within the system to represent some external condition that doesn't currently exist. If the elements within the system can be rearranged to represent what conditions might look like at some point in the future, we can call that a prediction. If they can be rearranged such that they represent conditions as they might have existed at some point in the past, we can call that a memory.
A mere-causal system, by contrast, does not contain symbolic information, but only a direct mechanical response to conditions as they currently exist. A thermostat, for example contains only elements that react mechanically (the thermometer, the circuit to the heater, the set point mechanism) - they don't "represent" anything other than what they are, and can't be used to forecast or recreate anything. A thermostat doesn't predict anything or remember anything; it just adapts to changing conditions according to it's set point (as determined by a human brain, not the thermostat itself). "Mere causal" is a bad term for this - I would have called it "mechanical", or "bio-mechanical" in the case of bacteria (assuming bacteria cannot process symbolic information).
The immune system appears to be another case of a bio-mechanical response system. While immune cells possess the capacity to adapt to new intrusive pathogens, this seems to be a chemical structural adaptation. The immune system isn't predicting anything, and not remembering anything in the form of symbolic information either, anymore than the thermostat is (because elements within the immune cells are not being rearranged "independent of current conditions" - they get rearranged when they encounter a new invasive cell). Immune cells do not "think".
A brain is different. We know that the human brain is a prediction machine, constantly checking current conditions against a template of what was expected to occur a few micro-seconds to years ago, and therefore must contain a symbolic representation of these predictions. We also know they contain memories (albeit reconstructed to some degree) which symbolically represent conditions as they are thought to have been, again from a few micro-seconds to years ago. Therefore, somewhere in the brain there must be a capacity for "information processing", the ability to rearrange systemic elements independent of current conditions.
The question is where this information processing capability is located. One possibility is within each individual neuron. In that case, neurons would be storing information and manipulating it somehow to help produce predictions and memories. Another possibility is in patterns of connections between neurons. This would consist of the growth and decay of physical connections, or changes in the weights between them.
There is very strong evidence that patterns of neural connections process symbolic information. The brain could hardly operate otherwise. The question is then whether or not this is sufficient to explain what the brain can do, or is an additional mechanism needed. Since this would be a more complex model, it requires additional evidence in support of it. Does it exist?
I think the evidence is consistent with a low level within neuron effect, which could then go on and affect macro-behavior, esp. in more primitive organisms. But it seems clear that the majority of the information stored and processed in the human brain is going on between cells, and that if intracell processing is going on, it works closely with intercellular processes. Strictly speaking, then, synaptic weights are not sufficient to explain all information processing, but do account for most of it.
Wait, if this was true, wouldn't it be very easy to test with blood transplants? Set up an experiment where people get transplanted with e.g. the blood of the nobel-prize-winning physicist, and some placebo blood, and check if they get better at solving physics problems?
(Or use a Navy SEAL's blood and test endurance against tests, or a drowning survivor's blood and test for hydrophobia, etc.)
In the essay and comment section, it seems part of the problem was that behavioral changes were characterized as memories. Food and sexual preferences may be better characterized as a novel reaction to a stimulus. Say, I lift a 30-pound dumbbell over my head daily, but I develop pain and on subsequent days I can no longer do so. I try, but can’t. The inability isn’t memory, but to an observer, it may appear to be.
Food and music preferences change without a heart transplant, sometimes suddenly. Behavior and mood are variable, and do not consistently correlate with any particular memory.
"If synaptic weights turn over on a time scale of weeks, maybe there’s something slower that turns over on a time scale of months. And maybe there’s something else that’s slower than that."
sounds like spaced repetition 🤔
It would be interesting to analyze what various drugs that impact memory are doing and how they might impact this. Psychedelics in particular. If they increase neuroplasticity, why is it often so hard to remember trips? Whereas if they are disrupting the flow of memories from short term to long term storage, it makes sense
I lack the relevant background to be able to tell if this is plausible or if this is more like the biology version of Velikovskyism, so I plead epistemic learned helplessness ( https://slatestarcodex.com/2019/06/03/repost-epistemic-learned-helplessness/ ) and just shrug my shoulders and move on. Probably a lot of the other readers should too.
Memory is stored in 1) morphic fields 2) with scalar waves 3) in metabolism as you can read in my paper https://thinsia.gumroad.com/l/life
Interesting review on Synaptic Plasticity! It's fascinating how our brains constantly adapt. Reminds me of the evolving strategies you need in many io games – always learning and adjusting to stay competitive. Some of those io games even mirror the complex networks our neurons create! Keep exploring these connections, both in science and those addictive browser games. https://iogamesonl.com/
It blows my mind that the stuff about memories from the things you eat made it in here. Did we just skip past digestion breaking down pretty much everything that would give you a memory? It makes the rest of the article seem incredibly credulous.
And the “if even 25% of these stories are true” argument is so bad.
I njoyed reading this review, even as someones who studies neuroscience in uni and knew most of the facts.
My main issue is that:
a) there is a distinction to be made between cognitive and none cognitive memory. I don't think I need to go into much detial to explain that immune system memory is not the same as an episodic memory of me as a child tasting strawberries for the first time, or even semantic memory of knowing a fact. Though there might be grey areas, I think its a distinction that is super easy to build a classifier for.
b) the review doesn't make a super compelling case as to why wouldn't the objective truth be something like: cognitive memory is govenred mostly by synaptic weights updating, while non cognitive memory is goverend by all sorts of mechanisms the we evolutionary had before having a brain (and cognition, without getting into the weeds of what exactly is cognition).
That's what most biologists think today, at least in my experience. And sure, at the begining the author talks about organ transplants and memory, but more of as a hook to the point of the review than actually zooming into the evidence and what this could entail.
I was a student almost a decade ago, and back then we were taught:
1. cognitive memory is probably mostly connections between neurons, but neurons have all sorts of complex modulating behaviors that we don't yet understand
2. there are, of course, other forms of "memory" in an organism
Both of these weren't controversial, and this review didn't really change my priors on that too much.
> there are two common arguments against ‘counting’ immune memory as a cognitive-like form of memory ... the Error Argument... the Mere Causal Argument.
If I blow off your leg with explosives, it will still be gone tomorrow. Is that "memory"? So I propose a third argument, the "Not Cognitive" argument: any form of memory that isn't accessed in the mind in a vaguely memory-like way isn't "memory". So while maybe your mind registers some leg pain, the continued absence of the limb doesn't count as "memory". Immune memory can be called "memory" informally, but in a discussion about how the mind works it seems like a red herring. (meanwhile the Error Argument seems to imply that computer memory isn't memory because the computer doesn't misremember.)
This article quotes the SPM hypothesis as
> Activity-dependent synaptic plasticity is ... both necessary and sufficient for the information storage underlying the type of memory mediated by the brain area in which that plasticity is observed.
But soon afterward it starts treating SPM as a strawman that ALL organisms on Earth use and need synapses for memory, so that we can "falsify" it by creating doubt that insects rely exclusively on synaptic memory. This is silly! What scientist has ever claimed that all organisms rely exclusively on synaptic memory? I think it would be rare to find a scientist saying this about all organisms with brains, but the author looks all the way down to single-celled organisms as if this isn't banal and unrelated to the subject of interest.
No, almost everyone talking about SPM is talking about humans. I expect other mammals to be studied mainly due to a certain social pressure not to do human vivisection, not because we care that much about how rats' brains work.
In section VI it is asserted that synaptic weight changes are insufficient for learning and memory, but the text prior to the sentence "And there are other reasons to think synapses are insufficient..." shows merely that other forms of memory may/do exist in some insects. I would point out that this doesn't even disconfirm SPM in the insects themselves. Synaptic storage can potentially be sufficient, or even necessary, even if we know for a fact that memories are also stored elsewhere, because redundancies exist in nature. Granted, if synaptic storage *is* sufficient and necessary then it would be surprising for evolution to preserve other mechanisms in these insects.
But it's possible for a mechanism to be sufficient and necessary and *gimpy*: a mechanism can *work* on its own but be suboptimal, just we can function with one eye but it's suboptimal. We can also function with only two hands, which as a father of two toddlers, I think is also very much less than ideal, but it's all evolution could manage. If a primary mechanism is sufficient, necessary and gimpy, there is selection pressure to preserve a secondary mechanism.
After that sentence, it is asked "How can we remember things for years with synapses that turn over on a time scale of weeks or less?"
This reminds me that computer DRAM can potentially remember things for years using memory cells that degrade within one second or so. Which, in turn, reminds me of a potential biological analogue, because I've been having an odd experience lately. I'm 45 years old, and just this year I noticed something: occasionally, just as I wake up, when my eyes are closed, I see a sequence of images roughly 0.22 seconds apart, with a ~0.05 second fade between the images. There are no thoughts or feelings associated with them; I see only the images themselves. Sometimes they are color, but they seem more often closer to monochrome. Every image tends to be visually similar to the next one in the sequence, with some overlap between the two. Some images depict something clearly, like a face, but more typically they depict parts of objects. And I have seen two kinds of image sequences: sometimes, the images are primarily foveatic (clear images in the center of my view) while other times they are primarily peripheral (blurry images in peripheral vision). It hasn't happened for a week or two now, but IIRC, in the most recent instance I tried thinking about a specific object to see if the object would appear in my view. It did not. These experiences, of course, led me to hypothesize that during sleep my brain could be systematically refreshing visual memories sort of like how DRAM does, and that a glitch was letting me see the process consciously.
The author lists a series of papers challenging elements of SPM, but oddly doesn't explain what any of the papers propose as an alternative. Maybe they agree with his CPM (or Martin et al 2000's hypothesis which evidently is not called CPM: "I’m just the only one calling it this") ― but he doesn't say so.
> What are these molecular and intracellular processes I allude to? Is learning and memory stored in RNA, as McConnell thought and Glanzman et al.’s experiments suggest? Is it stored in post-translational modifications to proteins, like Crick thought? Is it stored via epigenetic mechanisms, or in the stable states of gene regulatory networks, or in the action of transcription factors? The truth is that no one really knows. But I think there’s evidence for all of these...
Well, as I said, it would be surprising for evolution to preserve other mechanisms if one is necessary and sufficient. Is it possible? Sure. Is it worth investigating? Maybe. Do I trust this particular author to give me a clear-headed, unbiased view of the matter? Not at all. The vibes, the "reasoning", and the rhetorical approach I see in this article feel similar to the articles I have read by global warming dismissives.
"What causes global warming? Is it the sun? cosmic rays? Natural internal variability? Volcanoes? Increasing humidity? The truth is that no one really knows, but I think there’s evidence for all of these!" Well, I can tell you why each of these does not cause global warming and how we know it's mostly greenhouse gases and mainly CO₂. I don't know much about SPM, but the vibes he gives off? The vibes are very reminiscent. He's just not thinking clearly.