in Filosofi, Retorik & argumentationsteori

Why ”hate speech is violence” is a logical fallacy

I usually blog in Swedish, but let’s make an English contribution for one’s sake.

George Lakoff, professor of psychology, has written the article Why Hate Speech is Not Free Speech.

The argument is quite simple, and it’s that bad words create bad things in the brain, exactly the same way as physical violence does. So, bad words must be the same as violence. And from this, the normative claim ”(at least some) hate speech should not be allowed” is generally concluded.

Here’s some of Lakoffs argument:

Like violence, hate speech can also be a physical imposition on the freedom of others. That is because language has a psychological effect imposed physically — on the neural system, with long-term crippling effects.

Here is the reason:

All thought is carried out by neural circuitry — it does not float in air. Language neurally activates thought. Language can thus change brains, both for the better and the worse. Hate speech changes the brains of those hated for the worse, creating toxic stress, fear and distrust — all physical, all in one’s neural circuitry active every day. This internal harm can be even more severe than an attack with a fist. It imposes on the freedom to think and therefore act free of fear, threats, and distrust. It imposes on one’s ability to think and act like a fully free citizen for a long time.

That’s why hate speech imposes on the freedom of those targeted by the hate. Since being free in a free society requires not imposing on the freedom of others, hate speech does not fall under the category of free speech.

Hate speech can also change the brains of those with mild prejudice, moving it towards hate and threatening action. When hate is physically in your brain, then you think hate and feel hate, you are moved to act to carry out what you physically, in your neural system, think and feel.

That is why hate speech in not “mere” speech. And since it imposes on the freedom of others, it is not an instance of freedom.

Several others have made the same point, for example When Is Speech Violence? by Lisa Feldman Barrett, also professor of psychology:

The answer might seem obvious: Physical violence is physically damaging; verbal statements aren’t. “Sticks and stones can break my bones, but words will never hurt me.”

But scientifically speaking, it’s not that simple. Words can have a powerful effect on your nervous system. Certain types of adversity, even those involving no physical contact, can make you sick, alter your brain — even kill neurons — and shorten your life.


If words can cause stress, and if prolonged stress can cause physical harm, then it seems that speech — at least certain types of speech — can be a form of violence.

I have stated the argument here in more formal terms that permits logical scrutiny:

  1. Violence creates physical stress.
  2. Hate speech creates physical stress.
  3. Therefore, hate speech is violence.

This seems like a reasonable and airtight argument, so what’s wrong with it?

Well, there are several problems with it. I could say that the distinction between what a person says and what a person does cannot be conflated into the same category on the mere basis that both affect the brain. I could point out that classical conditioning might create the same effect on the brain, or that we should make a Humean is-ought distinction in question of disallowing hate speech on scientific grounds. Others have written responses along these lines.

However, I don’t need any fancy counter-argument. In fact, since the argument is based on invalid deductive logic I only have to point out the major flaw, which is completely devastating for the argument. The good thing with invalid arguments is that we can just discard them immediately and carry on with our lives.

The argument is a fallacy of undistributed middle. If we restate the above argument with symbols, we get something like this:

  1. A are Q.
  2. B are Q.
  3. Therefore, A are B.

But this conclusion does not follow from the premises. In fact, the middle term (Q) is undistributed among the other terms (A and B). That means that the argument only specifies that Q is connected to both A and B, but it actually says nothing about whether A and B are connected. This can easily be demonstrated with a reductio ad absurdum argument:

  1. Cats have four legs.
  2. My dog has four legs.
  3. Therefore, my dog is a cat.

This argument is based on the same invalid deduction, and one does not need any degree in formal logic to see that this is not a very compelling argument. In fact, this is precisely the same argument scheme that the hate speech is violence argument uses. But it’s just bad reasoning.

One would think that two professors in psychology would not make a basic deductive fallacy. But when politics is involved, humans tend to be motivated reasoners.

By the way, the ”my dog is a cat” is from Yes Minister! Thanks to Jake Errey who corrected me on the right categorical syllogism fallacy, and was kind enough to create a Twitter account to inform me. 

Gillade du artikeln?
Få senaste artiklarna direkt i din inkorg. Jag skickar aldrig spam.

Skriv en kommentar


  1. A discussion on Twitter (Swedish) came up, with this question:

    Shouldn’t it be that A causes Q rather than A are Q?

    Well, yes. I decided to ignore it when I wrote the text, but I can comment on it now as it came up. I made it a little bit simpler in the above argument to facilitate understanding with the Wikipedia page. However, the problem is still the same since the middle term is not distributed among the premises. That’s because it is a categorical syllogism, which means that the argument is used to link things together. Hence, it doesn’t really matter whether there is a material implication or not.

    We can simply restate the argument as follows:

    1. A creates Q.
    2. B creates Q.
    3. Therefore, A is B.

    Still, this argument is invalid and Q is not distributed among the premises.

    The same problem is still present regardless of whether we say some or all (which I also decided to ignore in my text):

    1. Some A creates Q.
    2. Some B creates Q.
    3. Therefore, some A are B.

    Still invalid.

    1. All A creates Q.
    2. All B creates Q.
    3. Therefore, all A’s are B.

    Nope, still invalid.

    How to turn it into a better argument

    But how would we go about if we’d like to turn the fallacy into a valid argument? We have to distribute the middle term. This can, for example, be done like this:

    1. Violence creates physical stress.
    2. Hate speech creates physical stress.
    3. That which causes physical stress is violence.
    4. Therefore, hate speech is violence.

    Now the term ”physical stress” is distributed among the premises and thus connected. And then someone would have to defend that claim (i.e., premise 3) in order for the conclusion to hold. It is not enough to assert the claim. For example, someone may be classically conditioned into feeling physical stress when confronted with a soda can. Therefore, proponents of this argument must, in some way, defend the claim that soda cans are violence. But then we’re back again to the Humean is-ought problem: should we really infer the normative claim what should and should not be allowed from a mere description of physical properties?

    Sam Harris wrote about this in his book, i.e. that science can tell us what is right and wrong. But he didn’t really spell out his assumption, that ”things that causes harm is bad”. Then it’s just a matter of putting people in a fMRI scanner and deduce what is right and wrong from neural activity. However, the ”things that causes harm is bad” assumption precedes science, and thus science in and of itself has nothing to say whether something is good or bad. Science can only tell us what causes good and bad things when we first have decided what is good and bad. So bringing science to the table is a waste of time in order to establish a moral claim.

    That’s why many scientists often make really bad philosophers. Scientists seldom understand where a normative claim begins, and where a descriptive claim ends. And it’s even worse in areas of science that’s explicitly committed to (but not limited to) social justice.

    But then again, we could just as easily as before refute the better argument with another reductio ad absurdum:

    1. Cats have four legs.
    2. My dog has four legs.
    3. That which have four legs is a cat.
    4. Therefore, my dog is a cat.

    It’s a valid argument, but nonetheless untrue. Good luck defending premise 3 in either the cat argument or the violence argument.