Is the Brain Like a Muscle? Lessons from Debunking a Seductive, But Incorrect, Idea

We often reason through analogies. When something is confusing, we can try to tie it back to something we understand better.

Very often, this process it at the heart of scientific understanding. Charles Darwin formed his famous theory of evolution through natural selection by forming an analogy with artificial selection, the process of human intervention to create new breeds of dogs, cats and corn.

However, sometimes this process backfires. An analogy which seems productive ends up giving misleading impressions. The solar-system model of the atom (Bohr model), in which electrons orbit the nucleus, was such a theory. It predicted that electrons would be constantly accelerating, and thus should emit light continuously until they fell into the nucleus. A better analogy, with wave functions and probabilistic clouds, took root instead.

Given this double-edged sword of analogical reasoning, let’s turn to one about learning itself: is the brain like a muscle?

Measuring the Analogy

Obviously the brain is not literally a muscle. It’s made of neurons and glial cells, not muscle fibers. So on this level, the two can’t possibly be equated. But what about in its function?

Here it really depends on what features you want to compare. Muscles can change and improve, and so can the brain. Muscles seem to experience fatigue with use, and although the brain may turn out to be more complicated, I definitely feel tired after solving math problems all day, and I bet you do too.

However, since two different things can be compared in an unlimited number of ways, the strength of an analogy is never in the analogy itself—it’s what you use it for. And in this case, the analogy saying the brain is like a muscle has had a long history and a very specific implication.

Prior to the early part of the twentieth century, the popular educational dogma was formal discipline theory. This theory reasoned that the brain was possessive of different intellectual capabilities, broadly defined: reasoning, perception, affect, memory, etc.. Each of these, it was assumed, was like a muscle. If you flexed your reasoning faculty with a game of chess, it was assumed, this would help you with reasoning about mathematics, just as flexing your biceps by lifting weights can make you stronger in an arm wrestle.

This analogy led support to the widespread practice of teaching superficially useless subjects, on the idea that the content of education wasn’t what mattered, but its form. If you strained your brain on memorizing Latin, this would improve your memory muscles for all sorts of things later on. Effort, not content, mattered.

So was this analogy correct?

In this case, we know the answer by the work by the psychologist who sought to put it to the test. Edward Thorndike ran experiments (which have since led to over a century of robust replications) in 1906 that showed, contrary to the muscle analogy, that improving memory or reasoning in one domain did not necessarily benefit those faculties in different ones.

Thorndike, formulated a hypothesis, known as identical elements theory, that said the amount of transfer that would occur from learning one thing to another would be based on the amount of identical sensory or response elements it contains.

This was a valiant first explanation, but researchers now view it as too simplistic. For one, it would seem to prevent analogical reasoning at all (where surface characteristics are different, but deeper structure is the same), the same principle the muscle metaphor rests on. However, psychology was much less developed at the time and cognitive explanations had yet to take root in theories of behavior.

Despite the flawed alternative hypothesis, Thorndike was successful in debunking a popular myth of his time: that teaching students Latin and geometry would exercise the brain’s general-purpose reasoning and memory faculties, and thus make its students excellent reasoners and rememberers for everything else.

A Stubbornly Persistent Myth

Thinking about learning Latin for the sake of building mental muscles is so far from the current educational norm, that it can feel almost quaint. We see ourselves as above falling for such a simplistic error in our own ideas about learning, right?

Yet, it’s exactly a revival of this failed analogy that you see popping up in all sorts of places.

Critical thinking is often touted as the must-teach skill in high-schools and colleges. And while learning logic and syllogisms may be useful in some particular settings, the idea that this exercise trains us to be good thinkers generally was given a thrashing almost a century ago.

Similar are the cries for universal training in programming. While programming is definitely a useful skill (arguably more useful than the pre-calculus directed math we tend to teach) and there are many practical applications for programming, that’s not the only reason offered for it. Programmers have better thinking skills, it has been argued, and that teaching kids to code will help them think logically about the world. The muscle metaphor persists.

Brain-training games premise their entire application on the idea that one can improve cognition like a muscle, training on irrelevant mental tasks to improve brain strength. Some studies (often paid for by the brain training companies themselves) do even suggest that they may work on tests of general reasoning.

But there’s a sneaky trick here: very often the “games” that the brain trainer uses are extremely similar to the tests psychologists use to measure general properties like working memory. But nobody argued that these tests were unlearnable! Even an IQ test is something you could improve on if you studied for it deliberately. The only reason these tests work is because they’re not something people typically study for. If some subset of people practiced them deliberately, we wouldn’t expect them to predict things in a general way.

A final reincarnation of the brain-as-muscle metaphor I’ve seen is that bilingual people have greater creativity, reasoning, memory or something else. Learning new languages, therefore, has the much-sought-after general-purpose learning effects.

While this case is more plausible than the above examples, because learning a language fluently is so extensive, it’s bound to impact your life in many diverse ways. It’s unlikely that any transfer effects from languages to non-linguistic domains will be robust enough to justify the immense cost of learning another language. If you want to learn another language for the purposes of learning another language, great, but the century-long history of failures of transfer in research should give one pause when suggesting it’s an easy fix to improve your intelligence.

Seductive Analogies and the Way Forward

Analogies are potent reasoning tools. They help us map a poorly-understood domain onto something we understand better. When they’re used well, this can be an enormous lever in reasoning, allowing us to think about something more clearly.

However, as the muscle-metaphor shows, analogies to scientific phenomenon can also backfire. Their potency, in this case, can also be their downside, as we get seduced into believing them even after evidence shows that they don’t apply.

Is this the final word on training your brain? Possibly not. It’s very difficult to prove that something is impossible in science. All we can show is that the things we’ve tried already don’t work. It may be the case that this next thing will do the work we need, and improve memory, reasoning, perception or some other cognitive faculty generally.

However, the research seems to suggest that this analogy may simply be wrong. We may not have a general-purpose reasoning ability that, like a bicep, gets improved whenever we use it. Instead, perhaps we have billions of neural circuits that form patterns conducive to solving some problems that can only extend, in a limited fashion to other intellectual tasks.

Or perhaps the circuit analogy is also wrong in some fundamental way, an analogy to computers that turns out to have undiscovered limitations. And when that happens, perhaps people will still insist on talking about the brain-as-circuits, and suggest experiments and techniques based on that analogy, a century after it has been debunked.

Read This Next
Should You Know Your IQ?
  • Michael Thiessen

    I think this is like most things, where the answer lies somewhere in between, a (linear?) combination of these descriptions and ideas. To suggest any is completely correct or completely wrong is typically an oversimplification.

    Light is not a wave, nor is it a particle. These are simply analogies that are used to describe how light behaves in various circumstances.

    In the same way, the brain is neither a muscle nor a circuit, but those analogies are useful for describing (and also predicting) certain phenomena.

    I think everyone has experienced that some skills and ideas are highly transferable between domains. Does this not provide some support for the brain-as-muscle explanation, at least in this specific case?

  • Kenneth Bruskiewicz

    If I remember correctly from Barbara Oakley’s course, the way she claimed transfer occurred was merely “practice the same skill in more circumstances and reflect on what they have in common”.

    Likewise, I remember the difference between beginner and master physics problem-solvers being outlined in Schoenfeld’s “Mathematical Problem Solving”. Some researchers did an experiment where they got beginners and experts to think out loud while they were solving a problem. The difference is that beginners only corresponded a method to a problem based on features which were very specific. Like they could only solve trajectory problems if the words in the problem were “ball, cannon…” etc. The experts identified the right method because they identified the right principle, such as energy conservation.

    Some concepts and skills are going to be more liable to transfer than others. By exposing yourself to many domains, /and reflecting on their meaning/, can we identify the patterns that distill themselves into rules. The ability to make rules seem to me as the crux of it. I don’t think that all domains are as good for this as others might be. It’s a property of the knowledge acquired rather that the brain itself (other than the brain being able to use such rules).

  • Roberto

    uhmmm… fairly long article that could be easily summarized in two sentences:
    1) the general belief is that the brain works like a muscle
    2) there is a lot of research that says it’s not true

  • Anonymous

    Fairly short comment that could be summarized with one word: stupid.

    The summary you are suggesting would be a great oversimplification of this article.

  • Max McGuire

    “Critical thinking is often touted as the must-teach skill in high-schools and colleges. And while learning logic and syllogisms may be useful in some particular settings, the idea that this exercise trains us to be good thinkers generally was given a thrashing almost a century ago.”
    That’s a bold claim without any support in the article. I completely disagree with this. Logical thinking is by definition “good thinking” so it is critically important that one study logic.
    Secondly, the article seems to indicate that “critical thinking skills” = logic. I agree that “critical thinking” has become a nebulous buzzword in education that doesn’t really mean anything anymore. The word logic is much more specific and I think it would be wildly incorrect to say that the study of logic is not important to becoming a “good thinker.”

  • Stanislaw Gadomski

    I recommend, interesting approach, interesting discussion.
    I also have an analogy, the brain as a hardware, the mind as software.
    Analogies are also loose and lame. But can they be inspirational? Maybe the mind has a backup copy that can be used after the final hardware failure?
    Knowledge of the brain has considerable progress and achievements (neurosurgery). Knowledge of the mind has not significant progress. Socrates’ motto is still valid.
    I also recommend available lectures by Tatiana Czernigovska, apparently a Sokratist.

  • aldiper

    I wonder if what Scott means by “..the idea that this exercise trains us to be good thinkers generally was given a thrashing almost a century ago.”, is that you can teach people logic and empirical verification etc., but they’ll still believe what they want.

    I’ve known people with university degrees, with experience of critical thinking, who still believed, based on “the evidence” that homeopathy was more than just the placebo effect.

    (Clarification on this point would be great Scott, thanks)

  • Sebastian Peter

    As a Computer Science student I have to say that my studies DID actually help me develop structural thinking, which I now use in various other areas in life. I agree that it wasn’t the programming per se, but more breaking down problems and finding organized solutions. As you studied CS as well, Scott, maybe you can relate.

    I feel like in your post here you discuss more the GENERAL assumption that logical thinking can be learned through learning programming, which I agree isn’t necessarily true.

  • Sebastian Peter

    I think the difference here is that these kind of studies are very short-term oriented. I think it’s obvious that a couple of hours of learning skillset x doesn’t immediately improve related skillset y.
    Learning a skill for a longer time though makes you more or less embody the change, such as in the case of learning languages as Scott mentioned.

  • Faisal Anderson

    Great article as always Scott!

    It’s difficult to fit anything to a solid analogy, but I think “brain as a muscle” is especially abstract. I think the brain is much like any part of your body and is adaptive to it’s environment, but it doesn’t necessarily grow in a linear fashion, it can also laterally expand (developing more neuron connections to link ideas and resources.) So I think the analogy is at it’s least incredibly limiting in promoting the extreme flexibility and functionality of the brain, and at it’s worst easily exploited by brain training companies to market products.

  • Nick Wright

    Can I suggest that it is the ability to achieve intellectual work: focus, concentration, absorption into memory, recall, etc. that can be likened to a muscle? And if we structure our time well (another generic skill!) we can plan for the exercises that will increase our ability to work.

  • anonymous

    you are a too dispersed guy
    !
    good at everything good at nothing !
    Three things to be good at is still too much !!
    the idea being that if you have done one thing correct and full, which is not my case yet, something called an evolution like in pokemon should happen but not like the highest you would want !
    A firend of mine encourages all of us to do our best , because he says if we do so, if we could complete the learning of One book, we will have complete one part of the “world program” that is running, and will understaned the world closer to the way he does.

  • Aanand Srinivas

    Is it worthwhile to learn Euclidean geometry in middle and high school?

    I agree that improving critical thinking in general seems to be difficult and there is no evidence that it transfers well across domains. So learning Euclidean Geometry / solving logic puzzles may not make me a “better thinker”.

    Are there other reasons to learn geometry then? In my experience, geometry has been a convenient way to learn the vocabulary commonly used to reason. What does it mean when people say something has been proved? What do assumptions mean? What is an axiom? What is a postulate? What does a typical deductive proof look like? Once I had this vocabulary and mental model, I could use it in totally disconnected areas of knowledge. It would be hard/impossible to teach this vocabulary in the abstract and geometry seems to be the current domain of choice.

    I feel like geometry is analogous to the storybooks we read as kids that develop our general vocabulary. Geometry is a story that develops a special case of that vocabulary focussed on deductive reasoning. Maybe, we can say it “builds reasoning vocabulary and communication” rather than “builds the reasoning muscle”. What are your thoughts?

  • bobango

    Let’s look at Latin, since that is one of the example subjects brought up here. I took five years of Latin, including learning to read Caesar and others in the original. I don’t think that learning Latin will make it easier for you to learn Russian, other than that you may acquire more of a disciplined approach to learning than you had before. So, the general analogy of the brain as a muscle doesn’t work here. However, I DO believe that learning Latin improved my pattern recognition competencies. Pattern recognition is a procedural skill that functions over and above the particular subject one is studying. Like other skills, it can be improved with practice. So, it turns out that I am competent at other pattern recognition challenges, such as duplicate bridge. Both learning Latin and learning to be a good bridge players requires thousands of hours of practice. In my opinion and experience, they both draw on a procedural skill that underlies and compliments both.

  • Scott Young

    What does “procedural skill” mean though? Is the brain actually organized so there is such a thing as a general “procedural skills”? That’s exactly the aspect of the metaphor I’m calling into question.

    I won’t go so far as to say that *everything* is isolated and the benefits of learning something are solely contained within the subject you learned it in. There’s clearly some benefit of analogical transfer, albeit it might only occur at extremely high time investment.

    However, it may be likely that what is being “learned” in either case simply isn’t a general ability, like a muscle, but some set of patterns. If those patterns transfer somewhat to a new domain, perhaps transfer is possible. If those patterns don’t transfer to the new domain, they may not.

  • Scott Young

    I think it’s useful to learn geometry to the extent that you want to be able to solve problems that need geometry. Geometry isn’t a useless subject. But if the argument is being made that it has some powers to go beyond geometry itself, the evidence there (from actual educational practice) is pretty weak. In fact, most of the transfer doesn’t even let people use the geometry lessons they learned in school in real life! So the situation is even more pessimistic than I’m illustrating here.

  • Scott Young

    I think there’s probably a number of skills that vary in terms of their generality. The question is which intellectual skills are structured this way, and which are really just abstract terms we use to lump a lot of things that exist only as fairly specific skills in the brain.

    If I were writing a neural net to recognize faces, I might say it has “good pattern recognition ability” but also notice that it can’t also recognize correct chess moves without further training. In this situation, “pattern recognition ability” isn’t a general property of the network but a rather specific skill that it’s been trained on. The muscle metaphor suggests many properties of the human brain are quite general, but the research on transfer suggests this is much more limited than is commonly presumed.

  • Scott Young

    There’s probably some aspect of metacognition that, once learned to an extreme degree in one domain, show some aspects of transfer. However, these are probably coming more from heuristics that work in both domains, things like “think through all the consequences of a change” “visualize the outcome” “break the problem into parts” etc.. I don’t believe they transfer more than that–so the content of the subject will probably need to be relearned, to the extent that it is different.

    Note, sometimes this metacognitive transfer can be negative! https://www.scotthyoung.com/blog/2014/10/14/assumptions/

  • Scott Young

    Or, if you’re like me, you believe brain = mind and are just different languages for describing the same process, just as this text = aligned magnetic fields on a hard disc in some server farm somewhere.

  • Scott Young

    I’ll try to dig into the research when I have time, but as someone who has studied logic, it’s my personal opinion that it doesn’t make you a better thinker.

    Here’s my reasoning:

    1. Transfer is weak. Most people who learn abstract subjects in school aren’t able to apply it outside of that, even when the problem is *directly* transferable. This may not be an attack on an ideal education in logic, but it should give one pause if training in logic is being suggested as an educational policy.

    2. Almost no real-life problems directly involve logic, as taught. Programming uses a lot more logic than does the real decisions where one could benefit from more rational thinking such as politics, philosophy, religion, science, etc.. I’m inclined to say the Rationalist community’s focus on cognitive biases and pattern recognition in popular types of arguments are more useful, but even then there’s room for doubt. The amount of situations where having a good mastery of prepositional logic will aid in real-life debate are minuscule.

  • Scott Young

    Everyone has experienced that some skills are highly transferable, but is that actually true?

    My point is that analogies can’t be considered on their own. They always need to be considered in what context you’re applying them.

    If I made the argument that “the brain is a muscle because you can improve your abilities through training.” That would be completely unobjectionable, so the analogy is a good one.

    The argument “the brain contains faculties the way your arm contains muscles, training one of those faculties will improve it for related tasks” is a lot more dubious.

    The research on transfer is quite interesting. On the one hand, it’s clear that transfer does occur at least some of the time. Otherwise analogies would be impossible and everything would be neatly contained in the box you learned it. It’s clear that something like transfer is possible for many subjects. But the question is how does that transfer occur and what limits it. Why isn’t it a general by-product of education rather than the special case that shows up after extreme involvement of practice?

    My own sense is that this is because the way memories, skills and knowledge are stored in the brain is actually quite specific, and not very general. Skills are transferrable to the extent that parts of these mastered patterns do actually copy over to the new situation, and that one knows to do this for that skill. So if one spends decades mastering Chess and moves onto Go, the benefit may come in the form of metacognition–knowing how to learn it, how to manage one’s time and composure under stress. But since the actual patterns of Go are completely different from Chess, these will probably have to be learned from scratch.

  • Stanislaw Gadomski

    The distinction is necessary. If you play a computer game and you have to find the key to doors allowing you to advance, you will not find it in the processor, even having most precise X-rays. You need to search it in the virtual reality of the game scene. Similarly the problems of the mind are in the other realm as these of the brain. I believe that without brain there is no consciousness / mind, Contrary to all religious beliefs. But my opinion is only a belief too. Before discovery of electromagnetic waves there was no reason to believe in their existence,

  • MC Hakim

    very nice

  • bobango

    My understanding is that memory is organized so that learning of cognitive content and procedural knowledge are processed in different modules of the brain. Read Artificial Minds by Stan Franklin. He is distinguished professor at the University of Memphis and designs models of the brain. They are based on the known neurophysiological features of the brain and simulate what those physical modules do. I understand the point that you are trying to make. I certainly don’t think of the brain as a muscle. However, for procedural knowledge, I think there is pretty good evidence of some general competencies that can be applied to more than one venue. I can dig up some more detailed sources if that is useful for this discussion.

AS SEEN IN