Friday, March 19, 2021

Why a "Mechanical Memory" Theory Does Not Work

Neuroscientists rarely advance detailed explicit theories claiming that brains store memories in some very specific way. They are usually content to speak vaguely about such a topic, as they do when they claim not very specifically that memories may be stored through "synapse strengthening" without stating some specific idea about how memory storage could work in such a way.  I know why neuroscientists are so vague on this topic. It is because any attempt to postulate a detailed specific theory of memory storage in brains will have all kinds of glaring defects and credibility shortfalls (just as there would be glaring defects and credibility shortfalls in any specific detailed theory attempting to explain how Santa Claus could deliver toys to all the world's good children on Christmas Day or Christmas Eve). 

But very rarely an attempt will be made to advance a detailed explicit theory about brain memory storage. Let us look at one such recent attempt, and how it falls flat on its face.  The theory was advanced by Benjamin T. Goult of the University of Kent, in a paper entitled, "The Mechanical Basis of Memory – the MeshCODE Theory." 

Goult advances the theory that human memory information is stored in binary format. Binary is when information is stored as merely a sequence of ones and zeroes, such as 10110010101010110010101101111001. There are quite a few severe problems with such an idea, including the following:

Problem #1: Human experience and learning does not occur in binary format.  When we see things or hear things or feel things, there is not passing through our bodies anything like a stream of binary numbers such as 1100101010101010010101.  Auditory and visual perceptions occur in an analog form that is entirely different from the digital form of binary information. 

Problem #2: Whenever human experience or learning is capable of being translated into binary format, it requires translation schemes and encoding protocols that are not known to exist anywhere in the brain or body. Some things that humans learn or experience are capable of being translated into binary by means of translation schemes and encoding schemes. But such schemes are complicated. For example, visual information seen with the eye or a camera can be translated into binary through an RGB method in which each pixel is represented by three different numbers between 1 and 256: one number representing the red intensity, another number representing the green intensity, and another number representing the blue intensity. Then those three decimal numbers can be translated into binary format. But such a technique for converting analog visual information into digital binary information involves translation schemes and encoding schemes that are not known to be available anywhere in the brain or body.  Similarly, strings of text such as "my dog has fleas" can be translated into binary by a computer system that (a) has knowledge of the English alphabet; (b) has a table like the ASCII table that translates English letters into decimal numbers; (c) has a subroutine for converting such decimal numbers into binary. But no such things are known to exist in the human brain. Human minds are familiar with the English alphabet, but on the neuron level and synapse level we have no evidence of any familiarity with such an alphabet.  There is no reason why the brain's biochemistry  would have some encoding scheme based on something like the English alphabet, which has existed for less than 3000 years. Converting analog data such as sounds into binary requires two different conversions (from analog into decimal, and from decimal into binary). There is zero evidence that the human body has ever internally done either one of these types of conversions. 

Problem #3Much of human experience could never be translated into binary format.  Humans remember emotions, and there is no way to translate such emotions into binary format. Humans also remember things like pleasures, pains, tastes and smells, and there is no way to translate such things into binary format. Humans also remember learned physical skills, such as how to ride a bike, how to swim, how to dance and how to play a musical instrument. Such skills cannot be translated into binary format. 

Problem #4The human body is not known to have anything like any capability for writing learned information in binary format.  Scientists have not discovered any capability for writing learned information in any form to any part of the brain. 

Problem #5: The human body is not known to have anything like any capability for reading information in binary format.  Scientists have not discovered any capability for reading information in any form from the brain, with the exception of the DNA-reading capability found in brain cells and all other cells, which is not memory information. 

Goult's paper does nothing to address the first three of these problems. He does make a very clumsy attempt to address Problem #4, by speculating about how something known to exist in the brain might function as a system for storing binary information.  He mentions a protein called Talin, and (as we see in Figure 2 of his paper) he speculates that perhaps when some section of such a protein is folded, that stands for "0" and when the same section of such a protein is not folded, that stands for "1."  

This wildly imaginative speculation is about as silly as claiming that clouds might store binary information, because round clouds might stand for "0" and oval-shaped clouds might stand for "1."  Such clouds would not meet the essential characteristic of a binary storage system, that there be only two possible states. Since there would be 100 gradations between "round" and "oval" shapes, you could never store binary information in clouds. Similarly, sections of a protein molecule would have 100 or more possible states of folding. So it would never work to try to store binary information by using the shapes of particular sections of a protein molecule to stand for either 0 or 1. And if information were stored in such a way, there would be no way to read it as binary, as the body has no such thing as some mechanism for analyzing the shapes of sections of protein molecules. 

Contrary to Goult's speculations, protein molecules are totally unsuitable for storing binary sequences.

protein molecule

No binary storage capability in something like this

Let's imagine some protein molecule in which particular sections of the molecule would always toggle between two states (contrary to the evidence that no such two-state toggle could exist, and that such sections could have innumerable different shapes).  Then how much binary information could be stored in such a protein molecule? No more than a few bits. 

But what if you wanted to store a decent chunk of information, such as, say, the famous line, "Four score and seven years ago our fathers brought forth, upon this continent, a new nation, conceived in liberty, and dedicated to the proposition that all men are created equal"?  That requires the following binary sequence:

 01000110 01101111 01110101 01110010 00100000 01110011 01100011 01101111 01110010 01100101 00100000 01100001 01101110 01100100 00100000 01110011 01100101 01110110 01100101 01101110 00100000 01111001 01100101 01100001 01110010 01110011 00100000 01100001 01100111 01101111 00100000 01101111 01110101 01110010 00100000 01100110 01100001 01110100 01101000 01100101 01110010 01110011 00100000 01100010 01110010 01101111 01110101 01100111 01101000 01110100 00100000 01100110 01101111 01110010 01110100 01101000 00101100 00100000 01110101 01110000 01101111 01101110 00100000 01110100 01101000 01101001 01110011 00100000 01100011 01101111 01101110 01110100 01101001 01101110 01100101 01101110 01110100 00101100 00100000 01100001 00100000 01101110 01100101 01110111 00100000 01101110 01100001 01110100 01101001 01101111 01101110 00101100 00100000 01100011 01101111 01101110 01100011 01100101 01101001 01110110 01100101 01100100 00100000 01101001 01101110 00100000 01101100 01101001 01100010 01100101 01110010 01110100 01111001 00101100 00100000 01100001 01101110 01100100 00100000 01100100 01100101 01100100 01101001 01100011 01100001 01110100 01100101 01100100 00100000 01110100 01101111 00100000 01110100 01101000 01100101 00100000 01110000 01110010 01101111 01110000 01101111 01110011 01101001 01110100 01101001 01101111 01101110 00100000 01110100 01101000 01100001 01110100 00100000 01100001 01101100 01101100 00100000 01101101 01100101 01101110 00100000 01100001 01110010 01100101 00100000 01100011 01110010 01100101 01100001 01110100 01100101 01100100 00100000 01100101 01110001 01110101 01100001 01101100 00101110

But there would be no way to store that in a Talin molecule under Goult's speculation. Under his speculation, each Talin molecule could store no more than about 13 of these digits. So storing a binary sequence like the one above would require many Talin molecules.  But Talin molecules do not exist in any linear sequence in the brain. Instead they are scattered in three dimensional space.  There would be no way to trace any sequence such as the one above in the brain. There would be innumerable routes between the different Talin molecules scattered throughout three-dimensional space, not a single linear route. Similarly, if I pour a jumbo box of Alpha Bits cereal (each piece of which is a letter) into a bucket of thick mud, and shake the thick mud, then the Alpha Bits letters would be scattered in a three dimensional way, and there would be no way to recognize a particular path from one letter to the next letter. The resulting mess could always be read in a million different ways, depending on how the path was traced in three-dimensional space. 

A DNA molecule is a one-dimensional thing. It has a very clear beginning and end, and once you are at one point in the sequence, there is always a very clear "next token" and a very clear "previous token." A DNA molecule is a physical structure that allows linear reading.  Talin molecules scattered in different positions in three-dimensional space (among very many other protein molecules) could never be a system allowing information to be read in any kind of regular, linear way.  

Were binary information to be stored according to Goult's speculation, there would be no way to read it. Reading such information would require some shape recognizer or fold shape recognizer that could traverse Talin molecules to analyze what shapes particular sections had. No such thing exists. 

What Goult has imagined is that protein folding could be used to store binary information. Protein folding is a mysterious thing, and we don't know how it happens. It is known that protein folding is relatively slow. For a new protein molecule to assume its characteristic three-dimensional shape requires between 50 seconds and 3000 seconds. Such a process is way, way to slow to be an explanation for human memory acquisition, which can occur instantly. 

Then there is the question of protein molecule lifetimes, which Goult ignores. Protein molecules in synapses have only short lifetimes averaging less  than two weeks.  According to the paper here, the half-life of the Talin molecule is only about 18 hours. Synapse proteins such as Talin therefore have lifetimes 1000 times too short to explain human memories, which can survive for 50 years of more. This factor alone is a decisive reason for rejecting Goult's theory altogether, along with every other claim that long-term memories are stored in synapses.  

Trying to lessen the probem of instant memory retrieval, Goult mentions several times the idea of indexes in the brain, which would make retrieval faster. He fails to tell us the reality here, that there is zero evidence for any kind of indexing in the brain. In fact, we know of the strongest reason why indexing should be impossible in the brain. It is that the brain is absolutely lacking in any type of coordinate system or position notation system or addressing system.  

Think of how an index works in a book. The index has lines that link topics with page numbers that represent exact locations in the book. But the brain is like a city in which none of the streets have names, and none of the houses have house numbers (or a book in which none of the pages are numbered).  Lacking any such addressing system, there is no way in which a brain could ever have an indexing system.  That's one of many reasons why instant memory retrieval cannot be reading information stored in brains.  Finding a memory stored in a brain would be as slow as finding an index card in a swimming pool that was a disorganized heap of index cards.  

Goult tells us, "Synapses are the perfect system for optimised cell signalling between connected cells, and there are approximately 100 trillion synapses in the brain."  The claim that synapses are "the perfect system for optimised cell signalling between connected cells" is pretty much the opposite of the truth. To the contrary, it is well known that synapses transmit signals with low reliability.  A particular signal will have a probability of less than .5 (and as low as .1) of transmitting successfully across a chemical synapse; and a brain signal would need to cross countless such unreliable synapses to move a tiny distance in the brain.  One expert tells us that a signal passing through a synapse "makes it across the synapse with a probability like one half, or even less." This is a very major reason for thinking that when humans recall with 100% accuracy large bodies of information (as people do such as stage actors who play Hamlet), they cannot possibly be retrieving information stored in or around synapses, as Goult imagines. An analogous situation is some person in a very noisy cafeteria, giving a message to the person next to him (who has only a 50% chance of hearing the message right), and then saying, "Keep the message passing on." If the message has to pass through 100 people in the cafeteria, from one to another, with each one having only a 50% chance of passing the message on accurately, we have pretty much the perfect recipe for unreliable signalling. 

A second reason why synapses are quite the opposite of being "the perfect system for optimised cell signalling between connected cells" is that chemical synapses are a very serious signal slowing factor. Each jump across the gap of a synaptic junction causes what is called a synaptic delay, of between .5 milliseconds and sometimes as much as 2 to 4 milliseconds. The problem is that a huge number of these synaptic junctions must be traversed each time a brain signal crosses every centimeter. The cumulative effect of such synaptic delays should make brains way too slow to account for instant human recall and very fast human calculation speed by many savants.  The problem is discussed in great detail in this post. 

There is no observational evidence to substantiate Goult's theory. No one has detected any binary information stored in any Talin molecule in the brain. No one has detected any binary information stored anywhere in the human body. There is genetic information in DNA molecules, but that information is not binary information. 

We know what binary information would look like if it were stored in the body. There would be a very long continuous sequence of physical items that could have only two possible states. It would be an arrangement nothing at all like what Goult has imagined.  An example might be a long molecule with only two elements, existing in a long string-like sequence. For example, the molecule might have a composition with a very long sequence like this: COOOCCCOCOOOCCCCOCOOCCOCCOCOCOCCOCOCOCOCOCCOCOCOCCOCOCOCOC. 

Under such a system, the C's (carbon atoms) might stand for 1, and the O's (oxygen atoms) might stand for 0.  We see no such sequences in any molecules in the body.  Carbohydrates are combinations of three types of atoms (carbon, oxygen and hydrogen), not two. Protein molecules are made from twenty different amino acids, and each such amino acid is built from at least four different atoms (nitrogen, hydrogen, oxygen and carbon).  Goult speculating about binary information stored in protein molecules is like someone speculating that clouds store advertising messages.  Just as clouds bear no resemblance to a system for storing advertising messages, protein molecules bear no resemblance to a system for storing binary information. 

Some of the things I have mentioned here are "show stoppers" not merely for Goult's scheme but also for any and all attempts to imagine the brain permanately storing information in binary format or any other material format. 

I have argued at length in various posts on this blog (such as this one) that the concept of an engram (an alleged place in the brain where a memory is stored) has no robust observational basis.  It is interesting that Goult's paper is part of a group of five papers by different authors, and one of those papers suggests abandoning the use of the term "engram," replacing it with "more neutral" terms such as cell assembly (supporting memory).  Besides discussing numerous ways in which current neuroscientists are using language in dubious and objectionable ways, the authors (Hardt and Sossin) state, "Stated succinctly, the term engram may reflect more wishful thinking than how memory and brain actually relate." 

2 comments:

  1. I've heard that mirror neurons are responsible for empathy is this true or another baseless claim.

    ReplyDelete
  2. There are no mirror neurons, and it is based on the same old nonsense, they put an encephalogram on a baby, animal or adult etc, it imitates someone and they see that a part of the encephalogram turns on randomly and they say that those are neurons mirror, obviously they don't exist

    ReplyDelete