Saturday, March 21, 2020

Exhibit A Suggesting Scientists Don't Understand How a Brain Could Store a Memory

Many a scientist claims that human memories are stored in brains. But when asked to explain how it is that a brain could retrieve a memory, scientists go "round and round" in circles, producing unsubstantial circumlocutions that fail to provide any confidence that scientists understand such a thing.  I discussed such explanatory shortfalls in my 2019 post “Exhibit A Suggesting Scientists Don't Know How a Brain Could Retrieve a Memory,” and my 2020 post  "Exhibit B Suggesting Scientists Don't Know How a Brain Could Retrieve a Memory."  When they attempt to explain how a brain could store a memory, scientists give the same kind of unsubstantial and empty discussion, the kind of discussion that should fail to convince anyone that they have a real understanding of how a brain could do such a thing. 

An example of such a thing is an article that appeared in the online site of the major British newspaper The Guardian. The article by neuroscientist Dean Burnett was entitled "What happens in your brain when you make a memory?" Burnett follows a kind of standard formula followed by writers on this topic.  The rules are rather as follows:

(1) Attempt to persuade readers that you understand memory by talking about the difference between long-term memory and short-term memory.   Whenever such discussion occurs, it actually does nothing showing any understanding of a neural basis for memory, for such a discussion can occur based on purely phenomenal observations about how well people perform on different memory tasks. 

(2) Attempt to persuade readers that you understand memory by talking about the difference between episodic memory and conceptual memory. This again is something that can be discussed without any reference to the brain, so any such discussion doesn't do anything to establish some understanding of a neural basis for memory. 
(3) Make frequent use of the word "encoding," without actually presenting any theory of encoding.   Neuroscientists love to use the word "encoding" when discussing memory acquistion, as if they had some understanding of some system of encoding or translation by which episodic or conceptual memories could be translated into neural states or synapse states.  They do not have any such understanding. No neuroscientist has ever presented a credible, coherent, detailed theory of memory encoding, of how conceptual knowledge or episodic experiences could ever be translated into neural states or synapse states.  Any attempt to do such thing would cause you to become entangled in an ocean of difficulties.  
(4) Mention one or two parts of the brain, usually exaggerating their significance.  I'll give an example of this in a moment. 
(5) Talk dogmatically about synapses, creating the impression that memories are stored in them, without discussing their enormous instability and unsuitability as a place for storing memories that might last for decades. 

 Burnett pretty much follows such a customary set of rules. He uses the word "encoding" or "encode" four times, but fails to present any substantive explanation or idea as to how any human episodic or conceptual information could ever be encoded, in the sense of being translated into neural states. Burnett claims,  "The hippocampus links all of the relevant information together and encodes it into a new memory by forming new synapses." He provides no evidence to back up this claim. There are some important reasons for thinking that the claim cannot possibly be correct. 

One reason is that studies have shown that people with very heavy hippocampus damage can have a normal ability to acquire conceptual and learned information.  The paper here discussed three subjects who had lost about half of the matter in their hippocampi.  We read the following:

"All three patients are not only competent in speech and language but have learned to read, write, and spell. ...With regard to the acquisition of factual knowledge, which is another hallmark of semantic memory, the vocabulary, information, and comprehension subtests of the VIQ scale are among the best indices available, and here, too, all three patients obtained scores within the normal range (Table 2). A remarkable feature of Beth’s and Jon’s stores of semantic memories is that they were accumulated after these patients had incurred the damage to their hippocampi."

The same thing was found by the study here. A group of 18 subjects were studied, subjects with severe hippocampus damage. Some 28% to 62% of the hippocampi of these subjects were damaged or destroyed. The subjects had episodic memory problems, but "relatively preserved intelligence, language abilities, and academic attainments."  We are told, "In all but one of our cases, the patients...attended mainstream schools."  Could patients with such heavy hippocampus damage have normal academic achievements if it were true that "the hippocampus links all of the relevant information together and encodes it into a new memory by forming new synapses"?  Not at all. In a similar vein, the study here involving 17 rhesus monkeys found that "monkeys with hippocampal lesions showed no deficits in learning and later recognizing new scenes."

A study looked at memory performance in 140 patients who had undergone an operation called an amygdalohippocampectomy, which removes both the hippocampus and the amygdala. Table 1 of the study found that such an operation had no significant effect on nonverbal memory, causing a difference of less than 3%. Table 3 shows that most patients were unchanged in their verbal memory and nonverbal memory. More patients had a loss in memory than a gain, although about 13% had a gain in nonverbal memory. These results are not at all consistent with Burnett's claim that "the hippocampus links all of the relevant information together and encodes it into a new memory by forming new synapses."

There is a reason why it cannot be true that a new memory requires a formation of new synapses. The reason is that humans can form new memories instantly, but both the formation of a new synapse and the strengthening of a synapse require minutes of time.  If someone fires a bullet that passes near your head, you will instantly form a permanent new memory that you will remember the rest of your life.  The same thing will happen the moment you break your leg in a biking accident.  Claiming that memories require either the formation of new synapses or the strengthening of synapses is incompatible with a fact of human experience, that humans can form new memories instantly. 

If it were true that memories were stored by a strengthening of synapses, this would be a slow process. The only way in which a synapse can be strengthened is if proteins are added to it. We know that the synthesis of new proteins is a rather slow effect, requiring minutes of time. In addition, there would have to be some very complicated encoding going on if a memory was to be stored in synapses. The reality of newly-learned knowledge and new experience would somehow have to be encoded or translated into some brain state that would store this information. When we add up the time needed for this protein synthesis and the time needed for this encoding, we find that the theory of memory storage in brain synapses predicts that the acquisition of new memories should be a very slow affair, which can occur at only a tiny bandwidth, a speed which is like a mere trickle. But experiments show that we can actually acquire new memories at a speed more than 1000 times greater than such a tiny trickle.

One such experiment is the experiment described in the scientific paper “Visual long-term memory has a massive storage capacity for object details.” The experimenters showed some subjects 2500 images over the course of five and a half hours, and the subjects viewed each image for only three seconds. Then the subjects were tested in the following way described by the paper:

"Afterward, they were shown pairs of images and indicated which of the two they had seen. The previously viewed item could be paired with either an object from a novel category, an object of the same basic-level category, or the same object in a different state or pose. Performance in each of these conditions was remarkably high  (92%, 88%, and 87%, respectively), suggesting that participants successfully maintained detailed representations of thousands of images."

In this experiment, pairs like those shown below were used. A subject might be presented for 3 seconds with one of the two images in the pair, and then hours later be shown both images in the pair, and be asked which of the two was the one he saw.

Although the authors probably did not intend for their experiment to be any such thing, their experiment is a great experiment to disprove the prevailing dogma about memory storage in the brain. Let us imagine that memories were being stored in the brain by a process of synapse strengthening. Each time a memory was stored, it would involve the synthesis of new proteins (requiring minutes), and also the additional time (presumably requiring additional minutes) for an encoding effect in which knowledge or experienced was translated into neural states. If the brain stored memories in such a way, it could not possibly keep up with remembering most of thousands of images that appeared for only three seconds each.

There is another reason why it cannot be true that we remember things because "the hippocampus links all of the relevant information together and encodes it into a new memory by forming new synapses," as Burnett claims.  The reason is that synapses are too unstable to be a storage place for memories that can last for decades.  The proteins in synapses have short lifetimes, lasting for an average of no more than about two weeks.

A fairly recent paper on the lifetime of synapse proteins is the June 2018 paper “Local and global influences on protein turnover in neurons and glia.” The paper starts out by noting that one earlier 2010 study found that the average half-life of brain proteins was about 9 days, and that a 2013 study found that the average half-life of brain proteins was about 5 days. The study then notes in Figure 3 that the average half-life of a synapse protein is only about 5 days, and that all of the main types of brain proteins (such as nucleus, mitochondrion, etc.) have half-lives of less than 20 days. The synapses themselves do not last for more than a few years.  So synapses lack the stability that would have to exist if memories are to be stored for years.  Humans can reliably remember things for more than 50 years. Such a length of time is about 1000 times longer than the lifetime of proteins in synapses. 

Without providing any evidence for such a claim, Burnett teaches the widely taught idea that memories migrate from one part of the brain to another. He states the following:

"Newer memories, once consolidated, appear to reside in the hippocampus for a while. But as more memories are formed, the neurons that represent a specific memory migrate further into the cortex."

We have no understanding of how a neuron could represent a memory, no evidence that memories are written to any part of the brain, and no understanding of how any such thing as a writing of a memory could occur in neurons and synapses. We also have zero understanding of how a written memory could migrate from one place in a brain to another place, nor do we have any direct evidence that any such migration occurs.  But we do have an extremely strong reason for thinking that accurate memories could not possibly migrate from a hippocampus into the cortex. The reason has to do with the very low reliability of signal transmission in the cortex. 

A scientific paper states, "Several recent studies have documented the unreliability of central nervous system synapses: typically, a postsynaptic response is produced less than half of the time when a presynaptic nerve impulse arrives at a synapse." Another scientific paper says, "In the cortex, individual synapses seem to be extremely unreliable: the probability of transmitter release in response to a single action potential can be as low as 0.1 or lower." 

Another paper concurs by also saying that there are two problems (unreliable synaptic transmission and a randomness in the signal strength when the transmission occurs):

"On average most synapses respond to only less than half of the presynaptic spikes, and if they respond, the amplitude of the postsynaptic current varies. This high degree of unreliability has been puzzling as it impairs information transmission."

So the transmission of information into the cortex must be extremely unreliable.  To imagine how unreliable such a transmission would be, with only a 10% chance of a nerve signal transmitting, imagine that you are trying to send an email to someone, but your email provider is so unreliable that there is only a 10% chance that any character that you type will be accurately transmitted.  You might send your friend an email saying, "Hi Joe, what do you say we have dinner at that new steak place that opened on 42nd Street?"   But the email your friend got would be unreadable gibberish, something like "Hwdsd ondSt?"  That's the type of information scrambling that would occur if memories were to migrate from the hippocampus into the cortex, given a cortex where there is only a 10% chance of any action potential (or nerve signal) transmitting.  

So if memories were migrating into our cortex, we would never be able to remember things accurately.  But humans have an astonishing capability for memorizing vast amounts of information with 100% accuracy. It is a fact that some Muslims accurately memorize every word of their holy book.  We also know that actors can accurately memorize each of the 1569 lines of the role of Hamlet, and that Wagnerian tenors can accurately memorize both the notes and the words of the extremely long parts of Siegfried and Tristan (the role of Siegried requires someone to sing on stage for most of four hours). 

Once we carefully ponder all the reasons for rejecting its main claims, and also carefully ponder the lack of any discussion of robust evidence for a brain storage of memory,  we can see that an article such as the Guardian article is a kind of Exhibit A that modern neuroscientists have no real understanding of how a brain could do any such thing as store a memory.  Nature never told us that brains store memories. It was merely neuroscientists who made such a claim, without good evidence. 

The article by Burnett is not a detailed scientific paper, but if we look at a typical scientific paper attempting to present evidence for memory storage in a brain, you will not find any robust evidence. A recent example is the 2019 paper "Changes of Synaptic Structures Associated with Learning, Memory and Diseases."  The paper fails to provide any solid evidence that synapse states have any causal relation with memory acquisition.  No clear message comes from findings such as "motor learning rapidly increases the formation and elimination of spines of L5 PyrNs in the mouse  primary motor cortex (M1), leading to a transient increase in spine number, which over days returns to the baseline," combined with other statements such as "another study showed that spine dynamics on L2/3 PyrNs are not affected by motor learning."  Anyone looking to find a relation between one effect and some other physical factor (in a small number of tries) will have perhaps a 25% chance of finding what looks like a correlation purely by chance.  For example, if I try to look for a relation between stock market declines and rainfall, I'll have perhaps a 25% chance of finding such an effect if I test on four random days. So we would expect that neuroscientists hoping to find some correlation between synapse activity and learning would find such a correlation in a certain fraction of the times they tried, purely by chance, even if synapses are not a storage place of learned information. Nowhere in this paper is there anything like an explanation of how a brain could store a memory, and when the paper authors confess that "the stability of memory and the dynamism of synapses remain to be reconciled," they basically admit that they have no answer to the objection that synapses are too unstable to be storing memories that last for decades.