My original post was: A Theory of Laughter. This post is three updates I’ve made since then.

I’ll copy the key box from my original post for ease of reference:


  • (A) IF my hypothalamus & brainstem are getting some evidence that I’m in danger
    • (the “evidence” here would presumably be some of the same signals that, by themselves, would tend to cause physiological arousal / increase my heart rate / activate my sympathetic nervous system)
  • (B) AND my hypothalamus & brainstem are simultaneously getting stronger evidence that I’m safe
    • (the “evidence” here would presumably be some of the same signals that, by themselves, would tend to activate my parasympathetic nervous system)
  • (C) AND my hypothalamus & brainstem have evidence that I’m in a social situation
  • (D) THEN I will emit innate play signals (e.g. laughter in humans), and also I will feel more energetic (on the margin), and more safe, less worried, etc.

Summary / tl;dr

The first update is the most substantive one, the other two at the end are much shorter and more minor. 

  • In Section 1, I’ll propose that maybe a single conscious “thought” cannot trigger both Ingredients (A) & (B) in the pseudocode box above. I don’t think this much changes my original discussion of laughter-during-physical-play, but it does imply that conversational laughter (including humor) might need to entail a situation that can be thought of in two “frames”, one that evokes Ingredient (A) and the other Ingredient (B). I’ll clarify what I mean by “frames”, then go through a concrete example by analyzing an actual joke (I apologize in advance), and finally conclude with how this new insight brings me closer to reconciliation between my model and existing humor theories like “incongruity theory”.
  • In Section 2, I’ll suggest that maybe the words “danger” and “safe” in the box above should have been replaced by “high-stakes” and “low-stakes”. For example, asking your crush on a date is not “dangerous”, but still nerve-racking.
  • In Section 3, I note that when I put up my original post, there was a missing figure caption that would have explained and sourced the neuroscience-related diagram near the end. I fixed it a few days later, but by that point most people had already read it.

1. I think I now have a better reconciliation of my model with some existing theories of humor including “incongruity theory”

1.1 The new insight

Here’s a new idea I had since writing the original post: Maybe it’s rare or even impossible for a single “thought” to trigger both Ingredients (A) & (B). 

Maybe you’re thinking: “Pfft, how is that a new revelation? It’s trivially obvious! There is mutual inhibition between the sympathetic and parasympathetic nervous system. Your heart rate doesn’t simultaneously go up and down, right? So obviously one “thought” can’t trigger both those systems. Duh!!”

Well, I don’t think it’s quite as obvious as that. It’s possible for the laughter circuit inputs to be signals that are well upstream of wherever that mutual inhibition happens. That’s what I was originally thinking. But it could be either way, and I’m currently thinking that the boldface claim above is probably right after all.

If that’s the case, then there would be two main ways to get laughter.

First, in the context of tickling and other physical play (Section 4.1 of my original post), maybe Ingredient (B) comes from a “thought”, whereas Ingredient (A) does not come from a “thought” at all, but rather it comes pretty directly through brainstem reaction circuits. (Note that, in my terminology, a “thought” is more-or-less an activation pattern of the cortex specifically, not the whole brain.)

Second, in the context of humor and other conversational laughter (Sections 4.2-4.3 of my original post), maybe the only way to get both (A) & (B) signals to overlap, and hence to get laughter, is by rapid frame-switching. Let me explain that next.

1.2 What do I mean by “rapid frame-switching”?

Detour to explain “frame” in the “frame semantics” sense: If you haven’t heard of “frame semantics”, there’s a confusing wikipedia article and better YouTube lecture, or just read on, the basic idea is simple. A “frame” is kinda “how you’re thinking about something”, especially by analogy to other things. For example, you are familiar with a “commercial frame” involving a buyer, seller, item, and money. In the course of a conversation or thought process, the “commercial frame” will be sometimes invoked, with almost arbitrary items occupying those four slots—e.g. maybe the buyer is a robot, the seller is an alien, the item is a secret password, and the money is special moon rocks. That’s fine! As long as you’re thinking about the situation in the “commercial frame”, you will reliably have certain expectations and inferences, like “the robot desires the secret password”, “the robot will give the special moon rocks to the alien”, etc.

Likewise, you’re familiar with a “birthday party frame” involving a guest-of-honor, people they know, and often a cake, song, candles, presents, etc.

The “commercial frame” and “birthday party frame” are just two of the thousands or millions of frames that you’re extremely familiar with, to which we can add vastly more frames that you could create on the fly by analogizing to your past experience and knowledge.

Back to laughter: Now, consider a scenario where there’s something that you can think about in multiple different frames. Suppose that thinking about the situation in one frame evokes Ingredient (A), and thinking about the situation in another frame evokes Ingredient (B). Suppose further you rapidly switch from one frame to the other, and maybe even flip back and forth a few times over the course of a second or two.

The downstream results would be a sources of (A) and (B) signals in the hypothalamus, which are not strictly simultaneous, but in close temporal proximity.

Now suppose further that those hypothalamus signals are a bit slow to decay. For example, whereas conscious thoughts can flip multiple times per second, I find introspectively that “feelings of mirth”, “feelings of excitement”, etc., cannot flip on and off at sub-second timescales. In that case, the (A) and (B) signals would wind up simultaneously active in the hypothalamus, thus triggering laughter, per the pseudocode box above and in my original post.

I should give a concrete example, shouldn’t I. Sigh. OK fine.

1.3 I’m really sorry everyone, but I’m gonna analyze a joke now

I don’t want to cherry-pick, so here’s “the funniest joke in the world” according to a survey:

Two hunters are out in the woods when one of them collapses. He doesn't seem to be breathing and his eyes are glazed. The other guy whips out his phone and calls the emergency services. He gasps: “My friend is dead! What can I do?” The operator says: “Calm down, I can help. First, let's make sure he's dead.” There is a silence, then a shot is heard. Back on the phone, the guy says: “OK, now what?”

I claim that this joke is forcing the listener to think about the situation in terms of (at least) two frames.

  • FRAME 1 is the “someone is gravely injured, and someone else with them is calling the emergency services” frame. In this frame, “make sure X is dead” is understood to mean “check whether X is dead or not”.
  • FRAME 2 is the “people are out hunting in the woods” frame. In this frame, “make sure X is dead” is understood to mean “ensure that X is dead”. (X would normally be the animal being hunted.)

Umm, I don’t personally find this joke to be very funny. But insofar as I maybe chuckled a bit, I think it’s because, when I first heard the joke, I experienced a period of transient confusion, during which my brain flipped back and forth a few times between these two frames. And during this transient period, whenever FRAME 1 was active, that thought was evoking Ingredient (A), and whenever FRAME 2 was active, that thought was evoking Ingredient (B). As above, the flipping back and forth would cause Ingredients (A) & (B) to be simultaneously present in my hypothalamus, triggering the laughter circuit.

1.4 Relation of my newly-updated model to things in the humor literature

When I think about my model in this way, it makes me more inclined than before to see a kernel of truth in certain theories of humor that I’ve read about. Specifically:

Incongruity theory: There’s a sense in which a situation X “should” evoke Ingredient (A), or Ingredient (B), but not both. Like, intuitively, there should be a “right answer”—either physiological arousal is appropriate to situation X, or it isn’t!!

So there’s something odd about a situation that calls for high physiological arousal when you think about it one way, but calls for low physiological arousal when you think about it in another way. It’s like an error in your brain. In fact, I think it’s literally an error, in the sense that some supervised learning algorithm(s) in your brain are getting an error signal while this is happening. Maybe our world-model is failing to capture something, maybe the situation is out-of-distribution for some reason—I think there are a lot of possible causes here. But I think the word “incongruous” is pointing roughly towards this type of situation.

(Incongruity is not sufficient for laughter, on my models. As above, it’s not enough to have two different incongruous frames for thinking about the situation; we also need one to trigger Ingredient (A) and the other to trigger Ingredient (B). However, even advocates of incongruity theory agree that incongruity is necessary but not sufficient for laughter—at least according to the Martin & Ford textbook.)

Relief theory: In the joke above, you’re switching quickly from a frame that evokes Ingredient (A) to a frame that evokes Ingredient (B). You could call that experience “relief”.

I think that’s basically the kernel of truth underlying “relief theory”.

(On my account, you’re also just as likely to laugh upon switching frames in the opposite, “anti-relief” direction, as far as I can tell. So it’s not a perfect match.)

Marvin Minsky’s “Jokes and the Logic of the Cognitive Unconscious” came up in the comments section of my original post. If I understand the article correctly (a big “if”!), Minsky was trying to relate humor to learning metacognitive rules of the type “Mental Operation X seems to be locally a good idea, but actually it isn’t”. He calls these learned rules “censors”.

I think that, on my models, Minsky is pointing to a real pattern that I can explain. As above, humor involves a situation that can be viewed in either of two frames, one triggering high arousal and the other not. When you’re flipping back and forth between them, your brain is simultaneously learning (or strengthening) a metacognitive pattern: If I see a situation that pattern-matches to this one, and I’m viewing it in FRAME 1, then I should flip it into FRAME 2, and vice-versa. Mechanistically, there’s nothing mysterious here—it’s just sequence learning, which your brain tends to do automatically. (Think of learning physical habits, or song lyrics.) So your brain can and will apply this new metacognitive pattern going forward.

OK, that’s my argument that “humor events” involve learning or strengthening what Minsky calls “censors”. It’s somewhat of an incidental side-effect on my account, as opposed to the main story, but it’s there nonetheless.

What about the converse—does learning or strengthening “censors” inevitably lead to laughter? I don’t think so. Maybe Minsky wasn’t claiming that anyway, but if he was, then I disagree. For example, I think deliberate practice in any domain involves learning lots of “censors”, but it’s not funny (more discussion here).

2. Maybe I should have globally replaced “danger” with “high-stakes” and “safe” with “low-stakes”?

In my original post, I frequently wanted to talk about the set of situations that evoke physiological arousal (i.e., increased heart rate etc.). One way to talk about those situations is directly, i.e. using the exact words in the previous sentence. That’s fine for the “brain-level” discussions that filled most of the post. But the post also had “evolution-level” discussions, and for those, it’s not ideal to talk about “situations that evoke physiological arousal”, because that would just raise a new question of why some situations evoke physiological arousal and others don’t.

So I also often used terms like “danger”, or “dangerous situations”. Unfortunately, those terms aren’t great. Not all dangerous situations evoke physiological arousal—e.g. a situation might be dangerous but not feel dangerous to the person in it. And conversely, not all physiological arousal is caused by evidence of danger—e.g. asking your crush on a date is not “dangerous”, and if my infant is screaming it’s not “dangerous” to me. (At least, not in the normal sense of the word “danger”.)

Maybe “high-stakes” and “low-stakes” would have been better than “dangerous” and “safe”. Still not perfect.

3. Due to a copy-paste error, for the first several days the article was posted, the caption to the final figure was missing, oops sorry.

It was mostly just references to neuroscience papers—you probably would have ignored it anyway. But if you’re interested, find it in Section 5.1 of my original post.

(I have added “check that the captions are all there” to my blog post publishing checklist, so this mistake shouldn’t happen again 😇)

Thanks especially to @Unnamed and @Carl Feynman for excellent and thought-provoking discussion in the comments section of my original post. :)

New to LessWrong?

New Comment
3 comments, sorted by Click to highlight new comments since: Today at 10:19 AM

I think this all makes sense. Just a couple of thoughts to add, which I think are consistent with what you've said:

Laughter from tickling seems like an incentive to play-fight. I think humor is an evolutionariy outgrowth of that to incentivize intellectual sparring. Like physical play-fighting, this sharpens skills.  Both can also demonstrate skills to improve one's position in a hierarchy, or desirability as a mate. There are theories about humor as a mating display; I'd like to call it stotting or pronking. Those technically are about signaling fitness to predators, not competitors or mates, but maybe we could broaden the term for the sake of humor?

From this perspective, humor is a serious business.

The danger/safety juxtaposition or switch might actually be created by the moment of confusion in parsing a joke.

I have some memories of being an adolescent, not getting the joke, and feeling distinctly in danger. Was I the butt of the joke? Or was I proving I was in the out-group by not getting it?

Even outside of social situations, confusion could be evolutionarily dangerous. Am I lost? Did I misinterpret the clues about where I could find food?

The click as you get the joke brings you back to safety. 

As a side note, this perspective implies that humor doesn't need to extend accidentally from the same principles that create laughter from play-fighting, since it's been actively adapted to serve a purpose from that starting point.

We have yet to touch on the topic of timing: comedians who perform in front of an audience often say that timing is important.

There can be a delay of a few seconds between punchline and the start of the laughter, and once the laughter begins, it usually gets loud very quickly. This behavior suggests that there is a (social?) cost to laughing at something most of the audience choose not to laugh at and also probably a cost to not laughing when most are laughing.

I think there are many jokes that do not depend on low or high stakes but work purely on the unexpected situation conjured in the joke. Good examples are puns and play of words, which are quite common in Germany (though many puns involve readings that have different levels of danger/stake that's not required). 

I see two readings out:

  • The ambiguity between the two readings and the associated flaw in the model is what is detected. Arousal is downstream of this detection.
  • The ambiguity is detected in terms of how much the steering system predictions fluctuate during a short time. All frames will necessarily predict at least slightly different ground truth and maybe if the cosine (or whatever other vector difference measure) is large enough, it counts. Fluctuation between arousal and non-arousal would drive the cosine high even if other dimensions wouldn't change much.