There Is No Good Evidence for a Neural Hallmark of Conceptual Learning or Memory Storage

If memories were stored in brains, we would expect that when a person learned something, there would be some type of physical change in the brain that could be observed, although it might be a tiny subtle thing that was hard to identify.  We can call such a thing a neural hallmark of memory. But no neural hallmark of conceptual or episodic memory has ever been observed.  

Let us imagine two different experimental subjects, either animals or humans. Imagine that the brains of both are thoroughly scanned, in an attempt to determine the exact state of their brains.  Then imagine the first subject was immobilized in a black silent room for ten hours, and the second subject experienced an intense learning experience for ten hours.  If memories are stored in brains, it should be possible to detect some change in the brain that the second subject had that the first did not. 

No test like this has ever produced good evidence of a neural hallmark of conceptual learning or knowledge acquisition.  But there have been some experiments similar to that described above, and some have claimed to have found evidence of memory formation in what are called dendritic spine differences. 

Dendritic spines are little bumps that protrude out of dendrites in the brain. We see below a schematic visual depicting 24 dendritic spines:

 Visual cropped from this paper

The idea that long-term memories are stored in dendritic spines is untenable, for two reasons. The first is that it is known that the proteins that make up dendritic spines are very short-lived, having average lifetimes of only a few weeks.  So there is nothing stable inside a dendritic spine. The second reason is that dendritic spines themselves are unstable, and they generally last for much less than two years. 

Synapses often protrude out of bump-like structures on dendrites called dendritic spines. But those spines have lifetimes of less than 2 years.  Dendritic spines last no more than about a month in the hippocampus, and less than two years in the cortex. This study found that dendritic spines in the hippocampus last for only about 30 days. This study found that dendritic spines in the hippocampus have a turnover of about 40% each 4 days. This study found that a subgroup of dendritic spines in the cortex of mice brains (the more long-lasting subgroup) have a half-life of only 120 days. The wikipedia article on dendritic spines says, "Spine number is very variable and spines come and go; in a matter of hours, 10-20% of spines can spontaneously appear or disappear on the pyramidal cells of the cerebral cortex." A paper on dendritic spines in the neocortex says, "Spines that appear and persist are rare." While a 2009 paper tried to insinuate a link between dendritic spines and memory, its data showed how unstable dendritic spines are.  Speaking of dendritic spines in the cortex, the paper found that "most daily formed spines have an average lifetime of ~1.5 days and a small fraction have an average lifetime of ~1–2 months," and told us that the fraction of dendritic spines lasting for more than a year was less than 1 percent. A 2018 paper has a graph showing a 5-day "survival fraction" of only about 30% for dendritic spines in the cortex.  A 2014 paper found that only 3% of new spines in the cortex persist for more than 22 days. 

In the light of such facts, it is pretty ridiculous to be looking for signs of a physical hallmark of long-term memory by looking at dendritic spines, which typically have lifetimes only a thousandth as long as the maximum length of time that humans can remember things. But nonetheless some scientists have attempted to do such a thing. A few scientists have presented scientific papers showing "before and after" photos of dendritic spines, papers that try to insinuate that some physical hallmark of learning can be seen.  No one should be persuaded by such experiments,  which have glaring methodological flaws. 

Here is what we will read in a typical scientific paper describing such an experiment:

(1) We are told that there were two sets of rodents, one group that did not engage in learning, and another group that did engage in learning. 
(2) We are shown two photos of dendritic spines, microscopic little bumps that protrude from neural components called dendrites.  Each photo will show about ten of these dendritic spine bumps. One photo will be from the learning group, and one from the control group.
(3) Some captions on the photos will suggest that the learning group has more dendritic spines than the control group. 
(4) There will be some graph suggesting the same thing. 

You may realize such papers are very flawed after you consider the question: how did the scientists select these particular dendritic spines out of many millions or billions of dendritic spines in the brain of the animal being studied?  A huge number of dendritic spines appear and disappear every day in the brain of every human and rodent. So it could not at all be true that the scientists scanned the brains of their subjects, and found some special little group of ten or twenty dendritic spines that were the only ones that had changed. 

What has actually happened in such experiments is that the scientists have simply randomly or arbitrarily selected a group of about ten or a hundred dendritic spines out of millions or billions they could have selected.  Perhaps this was done in a truly random way, using some random selection technique, or perhaps the scientists scanned many dendritic spines looking for a set that would show the alleged "memory storage" effect they were trying to show.  It's usually hard to tell from the way the papers are worded how the tiny set of "study spines" was selected.  Usually there will be no explanation of why this tiny set of about 10 or 100 dendritic spines is being photographed or carefully studied, rather than any 10 or 100 others of millions or billions of dendritic spines that could have been chosen.  In one paper we are told that the set of dendritic spines photographed was a "representative sample."  But we have no idea whether such a sample is truly representative, any more than we would know that ten randomly selected New York residents on the street are "representative samples" of New York residents. 

I can give an analogy explaining how bogus and bunk such a methodology is.  Imagine your hypothesis was that your memories are stored in the flowers of Central Park. You might do an experiment with this protocol: (1) you photograph some groups of Central Park flowers the first week of April; (2) you learn a lot on the second week of April; (3) you could go back to Central Park to photograph the same groups of flowers on the third week of April. Looking for a group of flowers that would support your hypothesis, you would probably have little trouble finding a few flowers that had grown nicely during the second week of April. You could then  publish "before and after" photos of such flowers to try to back up your claim that  your memories are being stored in flowers.  This would, of course, be a completely bunk methodology.  You would have no reason for suspecting that your memories were actually stored in the particular group of flowers you had photographed. 

Similarly, the scientists who perform experiments like the one I have described have no reason for believing that any memories acquired during testing are stored in the tiny group of dendritic spines they are showing in their papers.  

There are a few studies suggesting a little bit of a neural hallmark during muscular exertion activities (such as maze training), but such studies may be showing a bit of a kind of "muscle memory" effect that should not be confused with a physical sign of conceptual learning or knowledge acquisition.  My leg muscles also increase if I do enough activities involving walking, but that does not show that memories are stored in my legs.  We know that nerves are connected to muscles, so even if a brain is not storing learned knowledge, we might expect that parts of brains related to muscle activity might bulk up a tiny bit during novel muscle activities.  But if such a "muscle memory" exists, it is not a real memory, for the hallmark of a real memory is that it can be retrieved by a motionless person. 

In 1979 a scientific paper by Huffenlocher reached these conclusions:

1. Synaptic density was constant throughout adult life (age 16 to 72 years), with a density of about 1100 million synapses per cubic millimeter.
2. There was only a slight decrease in old age, with density decreasing to about 900 million synapses per cubic millimeter.
3. “Synaptic density increased during infancy, reaching a maximum at age 1--2 years which was about 50% above the adult mean.”

So according to the paper, the density of synapses sharply decreases as you grow up, in contrast to claims that learning or knowledge acquisition produces synapse strengthening. 

Some have claimed that a hallmark of knowledge acquisition can be found in London taxi drivers. To become a London cab driver, you have to memorize a great deal of geographical information. A study followed London cab drivers for 4 years, taking MRI scans of their brains.

But the study did not find that such cab drivers have bigger brains, or brains more dense with synapses. The study has been misrepresented in some leading press organs. The National Geographic misreported the findings in a post entitled “The Bigger Brains of London Cab Drivers.” Scientific American also inaccurately told us, “Taxi Drivers' Brains Grow to Navigate London's Streets.” 

But when we actually look at a scientific paper stating the results, the paper says no such thing. The study found no notable difference outside of the hippocampus, a tiny region of the brain. Even in that area, the study says “the analysis revealed no difference in the overall volume of the hippocampi between taxi drivers and controls.” The study's unremarkable results are shown in the graph below. 

The anterior part of the left half of the hippocampus was about 25% smaller for taxi drivers (100 versus 80), but the posterior part of the right half of the hippocampus was slightly larger (about 77 versus 67).  Overall, the hippocampus of the taxi drivers was about the same as for the controls who were not taxi drivers, as we can see from the graph above, in which the dark bars have about the same area as the lighter bars. So clearly the paper provides no support for the claim that these London cab drivers had bigger brains, or brains more dense with synapses.

In this case, the carelessness of our major science news media is remarkable. They've created a “London cab drivers have bigger brains” myth that is not accurate.  The supposedly bigger part (an area of about the size of a jelly bean) is only about 1/500 of the size of the brain. Give me any two randomly chosen set of people, and give me the freedom to make comparisons of 500 parts of their brains, and I will probably be able to find some little part that differs in size by 25% or more, purely because of random variations. This is no strong evidence for anything. 

Another scientific study claimed to find evidence that people called pandits who had memorized Sanskrit scriptures had brains different from normal people.  In a Scientific American story, we have a claim which initially sounds (to the casual reader) like evidence that memorization changes the brain. A scientist says, "Numerous regions in the brains of the pandits were dramatically larger than those of controls, with over 10 percent more grey matter across both cerebral hemispheres, and substantial increases in cortical thickness." But when we take a close look at the study, we find no robust evidence for any brain change caused by memorization. 

The study is what is called a whole-brain study. This means that the authors had complete freedom to check hundreds of tiny regions of the brain, looking for any differences between their 20 pandits who memorized scriptures, and a group of 20 controls.  The problem with that is that a scientist may simply find deviations that we would expect to exist by chance in tiny little brain regions, and then cite these as evidence of a brain effect of memorization.   Note that the scientist did not claim that the brains of the pandits who memorized scriptures had 10 percent more grey matter than ordinary people. He merely claimed that in "numerous regions" there was a 10% difference.  If I take 20 random people, scan their brains, and compare them to 20 other random people whose brains I scanned, I will (purely by chance) probably be able to find quite a few little regions in which the first group had more grey matter than the second (as well as quite a few regions in which the first group had less gray matter).  In the paper, we read that these pandits who memorized scriptures "showed less GM [grey matter] than controls in a large cluster (62% of subcortical template GM) encompassing the more anterior portions of the hippocampus bilaterally and bilateral regions of the amygdala, caudate, nucleus accumbens, putamen and, thalamus."  So the study found in some regions of their brains, these pandits who memorized scriptures had less gray matter than ordinary people, and that in other regions of their brains, they had more grey matter.  That is basically what we would expect to find by chance, and provides no good evidence for anything. 

On page 23 a technical paper tells us how many subjects we would need to have when doing this kind of "whole brain" study using brain scanning:

"With a plausible population correlation of 0.5, a 1000-voxel whole-brain analysis would require 83 subjects to achieve 80% power. A sample size of 83 is five times greater than the average used in the studies we surveyed: collecting this much data in an fMRI experiment is an enormous expense that is not attempted by any except a few major collaborative networks."

How many subjects did the whole-brain analysis study of Sanskrit pandits use? Only 21.  Meaning that it used only one-fourth of the subjects needed for a moderately convincing result, using the approach used.  The results are therefore not robust evidence for anything.  See here for a fuller explanation of why the Sanskrit pandits study provides no good evidence of a neural effect of memorization. 

A certain class of studies called environmental enrichment studies compares groups of animals, one raised in an environment that is not stimulating, and another raised in an environment that is stimulating (such as one with toys and exercise wheels for rats).  It is claimed that the animals raised in the "enriched environment" may have slightly denser or larger areas in certain parts of the brain. But we have no idea whether this is caused by simply increased muscular activity rather than anything pertaining to memory. A review of the topic says, "Several studies have even suggested that physical activity is the sole contributor to the neurogenic and neurotropic effects of environmental enrichment." 

As for the claim so often made that "neurons that fire together wire together," a dogma originated by Hebb,  there is no robust evidence for this dogma of neuroscience that synapses that are more often used become stronger than synapses that are not used.  To get good evidence for this claim, a neuroscientist would need two things: (1) some method of measuring how often synapses fire, and (2) some method of measuring how much they are strengthening.  But unlike the situation with a car (which comes up with an odometer that allows you to precisely know how much it is being used), there is no way of monitoring precisely in vivo how often a neuron or synapse is firing; and it is also extraordinarily difficult to tell whether or not a synapse has strengthened during some period of time. So we don't have any adequate way of testing the Hebbian dogmas that synapses are strengthened more strongly when used more often. And since we could never tell whether a synapse strengthening was caused by mere physical activity rather than memory storage, something similar to muscles increasing in size after greater physical activity, a claimed example of synapse strengthening couldn't be cited as a neural hallmark of conceptual learning.  

A small number of studies have claimed to show evidence of synapse strengthening after learning, but fail to do that in a convincing manner. Such studies typically involve tracking a small number of synapses in some animal.  But there are billions of synapses in every mammal, and we have no way of knowing whether the small number of synapses studied had any connection with a learning experience.  We can compare such studies to a study trying to prove that flowers wilt in Central Park when the Yankees lose, by showing us pictures of a few flowers that showed such a wilting.  To prove the idea of synapse strengthening during learning, you would need to compare the total synapse strength of one group of subjects that had learning and a control group of subjects that did not. But we have no method allowing a scientist to measure the total strength of synapses in an organism. 

The 2019 study here is the latest example of an unconvincing study trying to show some evidence of memories being stored in a brain.  There are two big reasons why the study shows nothing of the sort:

(1) The study uses a technique in which animals are trained to fear some stimulus, and are then subjected to a brain "cell reactivation" that can be roughly described as a brain zapping.  The animals supposedly froze more often when this brain zapping happening, and the study interpreted this behavior as evidence of an artificially produced memory recall of a fear memory. But such a technique does nothing to show that a memory is being recalled, because it is well known that there are many parts of a mouse brain that will cause freezing behavior when artificially stimulated.  The freezing behavior is probably a result of the strange stimulus, and not actual evidence of memory recall.  If you were walking along, you would also freeze if someone turned on some brain-zapping chip implanted in your brain. 

(2) The study is using sample sizes so small that there is a very high chance of a false alarm.  The number of animals per study group was only 10 to 12. But 15 animals per study group is the minimum needed for a modestly convincing result, and a neuroscientist has stated that to get a decent statistical power of .5, animal studies should be using at least 31 animals per study group. 

The second problem is one that is epidemic in modern neuroscience.  Neuroscientists are well aware that the sample sizes typically used in neuroscience studies (the number of animals per study group) are so low that there must be a very high chance of false alarms in very many or most of their experimental studies; but they continue year after year producing such unreliable studies.  There is a "publication quota" expectation that provides a strong incentive for such professional malpractice. 

In considering matters such as these, I like to remember a particular rule:

The rule of well-funded and highly motivated research communities: almost any large well-funded research community eagerly desiring to prove some particular claim can be expected to  occasionally produce superficially persuasive evidence in support of such a claim, even if the claim is untrue.  

We can consider an example of this rule, one involving astrology, the claim that the stars and planets exert a mysterious occult influence on the destiny of humans. Let us imagine that instead of there being merely a handful of poorly funded astrology researchers in the United States, there were instead 10,000 or more very well-funded astrology researchers, with billions of dollars in research grants to use to try to support their belief in astrology, by doing things like crunching statistics in various ways with computers.  It would then occur that we would occasionally read in the press stories presenting superficially persuasive evidence for astrology.  Such evidence probably would not stand up well to very close scrutiny, but it would be sufficient to give some talking points to astrology supporters. 

Similarly, if there was a large community of 10,000 ardent fairy researchers who were funded with billions of dollars, we would probably occasionally see superficially persuasive papers offering evidence for fairies. For example, with such an army of researchers, and so much money to spend, there might be occasional infrared heat signature studies suggesting anomalous little blobs of heat floating about that might be interpreted as fairies.  The researchers would be helped by the research rule that says, "Torture the data sufficiently, and it will confess to almost anything." 

And so it is for the 10,000 or more US neuroscientists funded with billions of dollars of research money (more than 5 billion dollars each year, according to this site).  Such scientists are able to occasionally produce studies providing superficially persuasive evidence for the dogmas the neuroscientists want to believe in, such as the idea that there is a physical hallmark of conceptual learning in the brain. Such evidence does not hold up well to very close scrutiny, but it is at least sufficient to provide some talking points for the neuroscientists.  Such evidence is actually no greater than the evidence we would expect to be produced for an untrue claim, given the "rule of well-funded and highly motivated research communities" cited above.