Wednesday, October 7, 2020

Engrams Are Touted Like Phlogiston Was Once Touted

 Scientists were once very convinced that they had figured out how burning works.  They were convinced that things burn because inside them is a combustible element or material called phlogiston, and that during burning this combustible element is released. We now know that this once-cherished theory is entirely wrong.  Like the earlier scientists believing in an incorrect theory of phlogiston, many a neuroscientist believes in the dubious idea that there are engram cells that store memories.  There is no robust evidence for any such thing.  In the post here I discuss some of the very many reasons for rejecting such a theory of neural memory storage. In the post here I discuss some of the flaws in studies that claim to provide evidence for engrams. 

A recent MIT press release claims to have some new evidence for engrams, giving us the not-actually-correct headline "Neuroscientists discover a molecular mechanism that allows memories to form."  You might be impressed by hearing such an announcement from MIT, if you had not read my previous post entitled "Memory Experimenters Have Giant Claims but Low Statistical Power." In that post I examined many cases in which MIT had made impressive-sounding claims about memory research, which were based on studies that tended to be unconvincing because of their too-small study group sizes and low statistical power. It's the same old story in the latest study MIT is touting.  

Here are some phrases I quote from the paper, phrases indicating study group sizes or the number of animals showing some claimed effect:

"n = 3 mice"

"n = 30 mice"

"n = 15 mice"

"n = 3 biologically independent samples" 

"n = 4 mice"

"n = 4 mice"

"n = 4 mice"

"n = 4 mice"

"n = 4 mice"

Alas, we once again have from MIT a memory study that has failed to provide robust evidence. A general rule of thumb is that to get modestly persuasive results, you need to use at least 15 animals per study group.  In the latest MIT study, apparently either much smaller sizes were used for some study groups, or the claimed effects occurred in only a small fraction of the animals, such as 4 out of 15 or 4 out of 30.  In either case, the results are not compelling. My criticisms of such papers for using too-small study group sizes is partially based on the guideline in the paper "Effect size and statistical power in the rodent fear conditioning literature – A systematic review," which mentions an "estimated sample size to achieve 80% power considering typical effect sizes and variances (15 animals per group)," and says that only 12% of neuroscience experiments involving rodents and fear met such a standard. 

To help understand why results involving only four mice are not convincing, let us imagine a large group of 1000 astrologers scanning birth and death data, eagerly looking for spooky correlations.  They might look for things such as this:

  • A match between a father's month of death and his son's month of birth
  • A match between a father's month of death and his son's month of death
  • A match between a father's month of birth and his son's month of birth
  • A match between a father's month of birth and his son's month of death
  • A match between a mother's month of death and her son's month of birth
  • A match between a mother's month of death and her son's month of death
  • A match between a mother's month of birth and her son's month of birth
  • A match between a mother's month of birth and her son's month of death

Now, if one of the astrologers were to show such a match (or a similar correlation), with only a sample size of four, this would be very unconvincing evidence. For it is not very unlikely that four such matches might occur by chance, particularly if there were many astrologers searching for such a match. If the ratio of matches was 4 out of 15 or 4 out of 30, that also would not be convincing, and not very unlikely to occur by chance. But if the sample size was much larger, showing something like 15 out of 15 such matches, that would be compelling evidence for a real effect, being something very unlikely to occur by chance.  Similarly, experimental results in neuroscience papers should not persuade us when only four animals were used, or when 4 out of 15 or 4 out of 30 animals had some claimed effect. There is too big a chance that such results may be mere false alarms, the kind of matches or correlations that might be showing up merely by chance. When thousands of experimental neuroscientists are busily doing experiments and busily scanning data eagerly looking for correlations that can be interpreted as engram evidence, we would expect that very many false alarms would be popping up, particularly when too-small sample sizes were used such as only  four animals, or when low-percentage effects were claimed, such as 4 out of 15 or 4 out of 30. 

Once again, in the Marco paper we have a neuroscience study using mouse zapping.  Typically a study claiming engram evidence will shock a mouse,  and then later send some burst of energy or light to some cells where the scientists think the memory is stored. A claim will be made that this caused the mouse to freeze (in other word, not move) because the burst or energy of light has activated the fearful memory.  Such a methodology is laughable.  For one thing, it is hard to accurately measure the degree of freezing (non-movement) in a mouse, and judgments of a degree of freezing tend to be subjective. A measurement of heart rate (looking for a sudden spike) is a fairly reliable way to measure whether a fearful memory is being recalled, but such a technique is not used in such neuroscience studies. Also, if freezing behavior (non-movement) occurs, we have no way of knowing whether this is caused by a recall of a fearful memory, or whether it is an effect produced by the very burst of energy or light sent into the mouse's brain. It is known that there are many areas of a mouse's brain that if zapped will cause the mouse to show freezing behavior.   (The Marco paper uses the same unreliable technique of judging fear by trying to measure freezing behavior of mice, rather than the reliable technique of measuring heart rate spikes.)  One of quite a few reasons why trying to measure freezing behavior in mice is not a reliable way of determining fear is that fear typically produces in animals the opposite of freezing behavior: a fleeing behavior.  Over my long life I have very many times seen a mouse around my living quarters, but never, ever saw a mouse freeze when I walked near it (the mice always fled instead). 

In the MIT press release, we are told the scientists shocked some genetically modified mice, and that the mice then began to produce some protein marker. We have no way of knowing whether the production of such a protein marker had anything to do with an alleged formation of a memory in the brain. Organisms such as mice are forming new memories all the time, and also producing new proteins all the time. The formation of the protein could have been merely the result of the electrical shocking, not the formation of a new memory.  Or the protein could have formed simply because proteins are constantly forming in the brain, which replaces its proteins at a rate of about 3% per day (as discussed below). Electrically shocking an organism probably produces many a brain effect that has nothing to do with memory formation.  We can compare the brain during electrical shocking to a pin ball machine that lights up in many places at certain times. 

The MIT press release gives a quote by the post-doc researcher Marco that gives us a hint that he may be a bit on the wrong track. We read this:

“ 'The formation and preservation of memory is a very delicate and coordinated event that spreads over hours and days, and might be even months — we don’t know for sure,' Marco says. 'During this process, there are a few waves of gene expression and protein synthesis that make the connections between the neurons stronger and faster.' ”
It is utterly false that the formation of a memory requires "hours and days, and might be even months." To the contrary, we know that  a human being can form permanent new memories instantly.  If someone sexually assaults you or puts a gun in your mouth, you will instantly form a permanent memory of that event that will probably last the rest of your life.  But protein synthesis requires many minutes. The fact that humans can form permanent new memories instantly is one of the strongest reasons for rejecting all claims that memories are formed when engrams (new cells or new cell proteins) are produced.  The formation of neural engrams would necessarily take a length of time sufficient to prevent the instantaneous formation of permanent new memories. 

The ability of humans to form new memories in only three seconds was shown by a scientific experiment discussed in this post. 

We would take much, much longer to acquire new memories if the theory of engrams (neural memory storage) was correct.  Discussing the rate of translation (something that must occur during the synthesis of a new protein), the source here states, "It was found that the rate is quite constant across proteins and is about 6 amino acids per second."  A article agrees, citing a speed of 6 to 9 amino acids per second. The average eukaryotic protein has a length of about 472 amino acids, according to this source.  Dividing 472 by 6, we are left with the conclusion that the synthesis of a new protein must take many minutes.  We cannot be forming new memories by some "engram creation" requiring the synthesis of new proteins, because we can acquire new memories instantly. 

engrams debunked
The 2018 paper here gives us a reason for rejecting all claims that memories are stored in brains. The paper finds that proteins in the human brain are replaced at a rate of about 3% to 4% per day. Unlike very many neuroscientists, who seem very skilled at ignoring the implications of their own findings, the authors actually seem to have a clue about the implications of their research. We read the following:

"Here we show that brain tissue turns over much faster at a rate of 3–4% per day. This would imply complete renewal of brain tissue proteins well within 4–5 weeks. From a physiological viewpoint this is astounding, as it provides us with a much greater framework for the capacity of brain tissue to recondition. Moreover, from a philosophical perspective these observations are even more surprising. If rapid protein turnover of brain tissue implies that all organic material is renewed, then all data internalized in that tissue are also prone to renewal. These findings spark (even) more debate on the interpretation and (long-term) storage of data in neural matter, the capacity of humans to consciously or unconsciously process data, and the (organic) basis of our own personality and ego." 

The authors rightly seem to be hesitating about whether there actually is an organic basis for our personality and ego Given a protein replacement rate of 3% per day in the brain, we would not be able to remember things for more than about 35 days if our memories were created as brain engrams.  

Postscript: This month the Science Daily site (which so often has hyped headlines not matching any robust research) has been showing a headline of "New Player in Long Term Memory."  The article is about a paper that suffers from the same problems as the paper discussed above.  The paper provides no real evidence for any physical effect in the brain causing memory consolidation.  Examining the paper, I find the same old problems that are found again and again and again in papers of this type, such as the following:

(1) Too-small study group sizes, with several being less than 8 animals per study group (15 is the minimum for a moderately reliable result).
(2) A study involving only mice, not humans.
(3) A use of an unreliable method for judging fear in animals (trying to measure the amount of time a mouse is "frozen" in fear), rather than use of a reliable fear-detection method such as measuring heart rate spikes. 
(4) Citations to other papers that suffered from the same type of problems.

Looking further at the Marco paper (which is behind a paywall, but kindly provided to me by a scientist), I see other methodological problems with it. For one thing, mouse brains were studied hours  after some foot-shocking of mice,  which means there wasn't any real-time matching between a memory creation event and something happening in a brain.  The paper also informs us that "blinding was not applied in the behavioral studies (CFC) and imaging acquisition because animals and samples need to be controlled by treatment or conditions."  Blinding is a very important procedural precaution to prevent biased data acquisition and biased analysis, and we should be suspicious of experimental studies that fail to thoroughly implement blinding protocols.  The paper also makes no claim to be a pre-registered study. When a study does not pre-register a hypothesis to be tested, the scientists running the study are free to go on a "fishing expedition" looking in countless places for some type of association or correlation; and in such cases there is a large chance of false alarms occurring. 

No comments:

Post a Comment