Sensorimotor Thought Experiment

theory
sensorimotor

#1

Hello all,

I wasn’t sure where exactly to put this. Just a brief thought experiment I thought I would share with respect to the sensorimotor aspects of the brain. I’m not at all confident in neuroscience or the current theory on the sensorimotor interface. But this thought experiment is interesting so I thought I’d share. It’s the “Unhappy Child” thought experiment.

Suppose a baby is born in the unfortunate circumstance of having only outgoing signals to the muscles from the brain, but no incoming signals from the muscles to the brain. The directions are with respect to the brain. So if the child were to move, they would have no direct indication of knowing they were moving. The only way to know is to use senses other than touch, like sight.

Here is the question: Will the child ever move?

A similar question was asked by a philosopher once upon a time (can’t remember who) but this draws a slightly different conclusion that is important for understanding how we move. I propose that the answer is no, the child will never move. With no previous experience “exercising” the brain activity (synaptic firing or what-have-you) used for movement, the child will remain oblivious to the fact that it could move.

If this is true, then it indicates that all movement is reactionary by nature, and that movement is dependent on first receiving input from the motor-related muscles and bits before it can give output.

This is not the answer, of course. While there is a definite answer, for now it is merely a philosophical question. As support for the idea that you need input from your muscles before you can use them, consider trying to wag your tail or wiggle your ears. For those of us not gifted with a tail or the ability to wiggle our ears, it is impossible to imagine moving them. We can see our ears, but without muscular input there’s no way for us to move them.

For this to stand as an analogy we would have to assume the ears have muscle, but the concept is really about anything we can’t move, whether it exists or not.

Perhaps the child will move, though. Then we can imagine that the movement would be random and may potentially result in exhaustion and death. That kid is quite unhappy. But we can support the idea that the child will move by considering a baby born into space - no gravity or physical contact, it simply comes into being. We would imagine the baby would move its arms around and learn of them, along with its head and legs and torso.

We might consider the fact that all muscles are constantly sending inputs to the brain, though, so we should stick to the baby born without “feedback” from its muscles.

If it were true that sensory input is required in order to be able to move, then this has obvious implications for a sensorimotor computer. The motors themselves must be able to act as inputs to the sensorimotor region in order for the program to use those motors, and it would use those motors by firing neurons and receiving feedback again.

Another consideration in the thought experiment is involuntary muscle movements such as those in the heart etc. I’m sure there are plenty more aspects I’m missing, as well. But I hope this helps people consider how the sensorimotor mechanisms behave a little better. If you have a different opinion, then by all means, have a go!

Sam


#2

Hey Sam, interesting thoughts.

If you imagine that many parts of the brain are initially developed as a “clean slate,” then you would expect the brain to be sending motor commands semi-randomly during early development. Thinking about the way neural tissue behaves, it certainly seems likely that young neural tissue would be spontaneously active before it develops activity-dependent associations.

Reinforcement learning systems in AI typically start this way, with random “motor babbling” in order to learn basic interactions with the environment. The behavior will only be random until the agent learns enough about its sensorimotor dynamics to start exploring in a more meaningful way. I wouldn’t say this strictly requires feedback from the motors themselves.

And in the actual brain of course, we have feedback not just in the form of touch, but also in the form of “efference copy” in which the motor axon splits and sends its signal both to the muscles and to the brain.


#3

Jake,

The idea of “motor babbling” is very interesting. It is a good resolution to the thought experiment, which has nothing against such a mechanism. I’m curious, would this be a natural result of the existence of neural tissue? In other words, why would so-called motor babbling be thermodynamically favorable with a “clean” brain? I’m treading well beyond my scope of understanding in neuroscience, so forgive me if the question seems odd or poorly worded. (In the coming months I’d like to begin learning more, I even picked up a book on synapses from a local bookstore!)

Sam


#4

Sam,

One idea is that neurons “want” to fire. The spatial pooler of Numenta’s HTM itself has this feature in the form of a “boredom” property (I think they call it “boosting”) where inactive columns of cells become more sensitive to their input over time, and this encourages competition between columns to ensure an even distribution of feature representations.

Many other systems have a form of boredom like this, for example SORN [1] which lowers the firing threshold of cells that are below their desired firing rate, and increases the threshold of cells that are above it. It’s a homeostatic mechanism that keeps the neurons at a useful activity level, and it’s based on biological evidence [2].

So you might consider this as the mechanism responsible for motor babbling, or you might just imagine that dynamical systems like neurons that are excitable (on the edge of stability) just tend to activate spontaneously due to the thermodynamically noisy nature of brain soup.

[1] Lazar, Andreea, Gordon Pipa, and Jochen Triesch. “SORN: a self-organizing recurrent neural network.” Frontiers in computational neuroscience 3 (2009). https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2773171/

[2] Turrigiano, Gina G., and Sacha B. Nelson. “Homeostatic plasticity in the developing nervous system.” Nature Reviews Neuroscience 5.2 (2004): 97-107. https://www.nature.com/nrn/journal/v5/n2/pdf/nrn1327.pdf


#5

Hello! I can recommend a great book: Norman Doidge - The Brain That Changes Itself (2007)

http://taisa.si/wp/wp-content/uploads/2013/12/The-Brain-That-Changes-Itself-by-Norman-Doidge-M.D..pdf

"… This theory was soon used to explain all movement.
Sherrington based his belief that reflexes
were the foundation of all movement on a deafferentation experiment
that he did with F. W. Mott.
They deafferented the sensory nerves in a monkey’s
arm, cutting them before they entered the
spinal cord, so no sensory signals could pass to the
monkey’s brain, and found that the monkey
stopped using the limb. This seemed strange, because
they had cut sensory nerves (which transmit
feeling), not the motor nerves from the brain to the muscles (which stimula
te movement). Sherrington understood why the monkeys couldn’t feel
but not why they couldn’t move. To solve this … "