The Thousandfold Shortfall of the Neural Reductionists

My previous three posts constitute quite an astonishing thing: a kind of neurological case for something like a human soul. Each of the posts presented an argument for the claim that your brain cannot be storing all your memories. One argument was based on the apparent impossibility of the brain ever naturally developing all of the many encoding protocols it would need to store the many different things humans store as memories. The second argument was based on the apparent impossibility of explaining how the human brain could ever be able to instantly recall memories, if memories are stored in particular locations in the brain, because there would be no way for the brain to know or figure out where a memory was stored. The third argument was based on the fact that humans can remember 50-year old memories, but scientists have no plausible explanation for how the human brain can store memories for longer than a single year.

No doubt some counterarguments arose in the minds of some readers, particularly any readers prone toward neural reductionism (the idea that human mental experiences can be entirely explained by the brain). In this post I will rebut such possible counterarguments.

The most obvious counterargument that one could give would be based on the fact of strokes and Alzheimer's disease. It can be argued that such afflictions show that very long-term memories are stored in the brain. But such an objection can be easily answered.

One way of answering such an objection is to point out that the evidence on brain damage and memory problems is very mixed. The physician John Lorber documented many cases of people whose brain tissue had been largely destroyed by disease, and found many cases of people whose memory and intelligence was still good. The operation called a hemispherectomy is sometimes performed on children, and involves removal of half of their brain. An article in Scientific American tells us, “Unbelievably, the surgery has no apparent effect on personality or memory.”

An even better way of answering such a counterargument is to point out that we cannot tell whether Alzheimer's patients or strokes have actually suffered a loss of memories. For it might be that such patients merely experience a difficulty in retrieving memories.

Imagine you are used to visiting cnn.com to get the news each morning. But one day you turn on your computer and find you can no longer access any information at cnn.com. Does this prove that the information stored at cnn.com has been lost? It certainly does not. The problem could merely be an inability for you to retrieve information at cnn.com, perhaps because of a bad internet connection. Similarly, if I write the story of my life, and place it on my bookshelf, I may one day go blind and be unable to access that information. But the information is still there on my bookshelf.

In the same vein, the memories of people with Alzheimer's may be perfectly intact, but such persons may be merely experiencing some difficulty in retrieving their memories. There are, in fact, reports of incidents called terminal lucidity, in which people suffering from memory loss or dementia suddenly regained their memories shortly before dying. Such reports tend to support the idea that memory problems such as Alzheimer's involve difficulties in retrieving memories rather than the actual destruction of memories stored in the brain. 

There is actually a way in Alzheimer's may argue against the idea that your memories are all stored in your brain.  A doctor reports the following:

One of the big challenges we face with Alzheimer's is that brain cell destruction begins years or even decades before symptoms emerge. A person whose disease process starts at age 50 might have memory loss at 75, but by the time we see the signs, the patient has lost 40 to 50 percent of their brain cells.

If your brain cells were the only place your memories were stored, why would you not notice memory loss until 40% or 50% of your brain cells were gone?
 

Another way of rebutting my argument about very-long term memory (and our inability to explain it) is to do an internet search for “long term memory,” and then find some biological component I didn't mention (perhaps some kind of protein), a component that is described as “playing a role in long-term memory.” You might then argue that I failed to mention that component, and that maybe that is the secret to very long-term memory.

But such an approach would be fallacious. Very confusingly, scientists use the term “long-term memory” for any memory lasting longer than a day. So anything at all that affects memories lasting longer than a day may be described as something that “plays a role” in long-term memory. In general, such things do nothing to answer the problem of very long-term memory, the problem of how memories can last for years or decades. For example, PKM protein molecules have been described as “playing a role in long-term memory.” But having lifetimes of less than a month, such molecules cannot explain how the brain could store memories for years.

Very rarely, it is suggested that maybe epigenetics can explain long-term memory. Epigenetics involves the effect of methyl molecules that can attach themselves to a DNA molecule. To use a rough analogy, we can imagine a DNA molecule as a protein recipe book, and we can imagine these methyl molecules as being strips of black electrical tape that block out certain parts of that book. Another analogy is that they are like “off” switches for parts of DNA. As these methyl molecules are very simple molecules, they are quite unsuitable for information storage use. Also, the methyl molecules attached to a DNA molecule are not connected to each other. So the methyl molecules involved in epigenetics lack the two characteristics of synapses that caused scientist to suspect their involvement in memory: the fact that they can have different “weights,” and the fact that they are connected to each other.

Another objection that can be made is to point out that some scientists have come up with theories trying to explain how information could be maintained long-term in synapses. Such theories are sometimes called theories of “synaptic plasticity maintenance” or “synaptic plasticity persistence,” although given that they involve attempts to show how unstable synapses could preserve information indefinitely, it might be better to call them “magic maintenance” theories. Far from debunking my objection that molecular turnover should make it impossible for synapses to maintain memories, such theories actually show that such an objection is quite weighty, and one that scientists have taken very seriously.

I can characterize such theories with an analogy. Let us imagine a young boy whose mother died a year ago. He comes to a beach where his mother used to swim, and he sees written in the wet sand a message of love.

Engaging in wishful thinking, the boy decides that the letters were written by his mother who died a year ago. When his father points out that this couldn't be, because letters in wet beach sand don't last, the boy comes up with an elaborate theory. Perhaps, the boy reasons, his mother hired some people to constantly preserve the letters she wrote in the sand. Perhaps, the boy thinks, whenever the words are overwritten by the high tide, some person hired by his mother makes sure that the letters are rewritten just as his mother wrote them.

Such a theory – basically a reality denial mechanism – would be very much like the attempts that have been made to advance theories by which synapses could preserve memories even though the protein molecules in synapses are recreated every few days. Both involve complex speculations hoping to get us to believe that some information that should be very short-lived is really very long-lived. The synapse theories in question – these “magic maintenance” theories – are not any more believable than the little boy's theory about the sand.

One such theory is a very sketchy theory of “bistable synaptic weight states,” which just offers the most fragmentary suggestion of how some special and very improbable “bistable distribution” setup might help a little --not very much, and not enough to account for memories lasting for years. But as this 2015 scientific paper states, “Despite extensive investigation, empirical evidence of a bistable distribution of two distinct synaptic weight states has not, in fact, been obtained.” The theory in question relies on an assumption of unobserved biochemical positive feedback loops, but one scientific paper notes that “modeling suggests that stochastic fluctuations of macromolecule numbers within a small volume such as a spine head are likely to destabilize steady states of biochemical positive feedback loops,” making them unsuitable for anything that can explain long-term memory.

Another such theory is a “cluster dynamics” theory advanced in this paper. The author makes a totally unwarranted ad-hoc speculative assumption. Speaking of the insertion of a new protein – where a new protein would appear when a brain cell replaces a protein, as it constantly does – the paper says, “The primary effect of this implementation is that the insertion probability at a site with many neighbors (within a cluster or on its boundary) is orders of magnitude higher than for a site with a small number of neighbors.” There is no reason for thinking that there should be any difference in these probabilities. He then attempts to show that this imagined effect can work to preserve the information in synapses, by giving us a “simplest possible case” example of a square shape that is preserved on a grid, despite turnover of its parts. But such an effect would not work in cases much more complicated than this “simplest possible” case, and would not work to preserve information more complicated than a square.

I tried created a spreadsheet that would use this type of effect. After a great deal of trial and error, I was able to find an “intensity of neighbors” formula that would set things up so that the lower set of numbers in the photo below (marked Generation 2) recreates the data in the first set of numbers (marked Generation 1). The formula used to create Generation 2 used the same idea in the “cluster dynamics” theory – the value of an item in the second generation depends on how high are the nearby values in the earlier generation. When I tried using some square-shaped data in the Generation 1 area, the effect worked okay, and the Generation 2 area looked like the Generation 1 area. 

But as soon as I tried using some data in the Generation 1 area that was not square-shaped (such as X-shaped data or E-shaped data), the effect no longer worked. Here are the results when you try some E-shaped data. The resulting Generation 2 does not look like an E-shape.

Simple as this experiment was, it captures the fatal flaw of this “cluster dynamics” theory. While it might work to explain a preservation of the simplest possible data (square shapes), it will not work to preserve any data more complicated, such as E-shapes and X-shapes. Of course, our memories involve memories of things infinitely more complicated than an X shape.

All of these “synaptic plasticity maintenance” theories have one thing in common: they have already been ruled out by experiments with long-term potentiation (LTP) and synapses. Such experiments have shown that LTP very rapidly degrades, and that synapses do not magically maintain their information states over long periods.

Another objection that could be made is one along these lines: sure, scientists don't have an explanation now for very long-term memory, but they'll find out someday how the brain is storing it. My answer to this is: no, they won't, because there are good reasons for thinking there is no place where it could be stored in the brain. The only candidates for places where very long-term memory could be stored in the brain are: synapses, DNA in nerve cells, protein molecules surrounding the DNA, other nerve cell proteins, or what is called the perineuronal net. My previous post made a good case that none of these are plausible storage places for very long-term memory. If you've ruled out all the places where something could be in some unit of space, then you have to conclude it's not in that space. For example, if you're looking for your car keys and you've ruled out all the places they might be in your bedroom, you should conclude they are not in your bedroom (rather than telling yourself: I'll find them some day hidden in my bedroom).

What is very astonishing is the difference between the longest memory that can be plausibly explained by neuroscience, and the longest memory that humans can hold. As discussed in my previous post, the best “candidate explanation” for memory is LTP (long-term potentiation), but that decays in a matter of days. The protein molecules between synapses also last no longer than two weeks. Even if you were to claim that LTP is a good mechanism for memory (something quite questionable, since it doesn't correlate highly with memory), then about the longest memory that science can currently explain (without resorting to baroque speculation) is a memory lasting about two weeks. Because humans can hold memories for 50 years, they are displaying a memory capacity that is about 1000 times longer than what can be plausibly explained by neuroscience. The explanatory gap here is like the gap between Jupiter and Saturn. How long will it be until we wake up, smell the coffee, and realize that to explain our minds we must go beyond the brain?