Fusion and Equivocation in Korzybski's General Semantics

by abramdemski3 min read21st Dec 202012 comments


IntrospectionNoticingFallaciesCognitive FusionMeditationBucket ErrorsDistinctions

I recently tried to figure out Korzybski's General Semantics again. I think I gained some insight into what Korzybski and followers were so excited about, and thought I'd try to write a brief note about it.

My "aha" moment came from reading the wikipedia entry on General Semantics, which includes the following:

General semantics postulates that most people "identify," or fail to differentiate the serial stages or "levels" within their own neuro-evaluative processing. "Most people," Korzybski wrote, "identify in value levels I, II, III, and IV and react as if our verbalizations about the first three levels were 'it.' Whatever we may say something 'is' obviously is not the 'something' on the silent levels."[11]


Although producing saliva constitutes an appropriate response when lemon juice drips onto the tongue, a person has inappropriately identified when an imagined lemon or the word "l–e–m–o–n" triggers a salivation response.

This seems like a very Buddhist use of "identification". It reminds me very much of Kaj's explanation of insight meditation. Korzybski is trying to get us to "respond appropriately" by differentiating rather than identifying. He writes that differentiating "introduces a most beneficial delay" between stimulus and response. Identification and differentiation seems like exactly what Kaj is talking about with "fusion" and "defusion".

In traditional argumentation terms, this mistake is "equivocation" -- treating two things as the same when they are importantly different. Equivocating between two things forces us to respond in the same way to both. For example, if no firm distinction is made between "liberal immigration policy" and "democratic values", then one must accept or reject both at once 

  • The grecio-roman (ie "western") rhetorical tradition focuses on explicit reasoning, and treats equivocation as an error in explicit reasoning. It is countered by precision in language, including taking care about explicitly defining terms, and careful analysis of the logic of arguments.
  • The Buddhist tradition focuses instead on the internal, subjective, phenomenological version. "Identification" here has to do with insufficiently fine-grained perception. It is remedied by introspection and training the senses. This allows one to see more clearly by focused attention, which allows one to differentiate (de-identify) subtly different mental phenomenon, such as that between hearing a sound and the lightning-fast inference which produces an interpretation of that sound (which is normally equivocated with the sound itself, at a basic level). This focus on nonverbal processing makes it seem in many ways like the opposite of the grecio-roman tradition. A Buddhist might naturally say that the focus on rhetoric traps the rhetorician in their mental models, whereas the focus on perception frees one from such mental models. A rhetorician might naturally accuse Buddhists of abandoning symbolic reasoning, a tool which has proves extremely powerful (e.g. modern mathematics).
  • Korzybski wants to do both: train explicit reasoning, while also recognizing "the silent level" (the nonverbal stuff). He uses the terms identification/differentiation to point to the general case.
  • Modern psychotherapy has picked up these ideas under the heading of fusion/defusion, mostly focusing on the more preverbal stuff that Buddhism cares about.
  • CFAR uses the concept of bucket errors for the same thing, although there the emphasis is more on why "irrational" behavior might actually be helpful/protective if you're making a bucket error, rather than a heavy-handed emphasis on getting rid of bucket errors in all cases.

In order to point people at the general idea of making distinctions, these traditions naturally point at some important distinctions people are neglecting. Korzybski famously touts the map/territory distinction as one of the main things people often equivocate, pointing out that (even if you explicitly endorse the existence of a map/territory distinction) it's easy to make mistakes by thinking an element of your map necessarily corresponds to some reality in the territory. Much was made of this in Eliezer's Sequences. 

But Korzybski really wanted to point to a lot more distinctions than this, as evidenced by the reference to "levels I, II, III, and IV" mentioned in the quote above. Level I refers to the territory ("happenings" as he called it), and level IV refers to explicit verbal description; the intervening levels refer to various levels of intermediate processing between the two. But there were many more distinctions as well. Again quoting from the wikipedia page:

Suppose you recognize one student—call her Anna—from a prior course in which Anna either excelled or did poorly. Again, you escape identification by your indexed awareness that Annathis term, this course is different from Annathat term, that course. Not identifying, you both expand and sharpen your apprehension of "students" with an awareness rooted in fresh silent-level observations.

Other traditions have pointed to other distinctions as their central examples. Buddhism has a number of favorite equivocations to point at, including the tangled mess of equivocation that is "the self". 

Each important distinction has a number of important implications. Pulling things apart allows us to address each differently. I talked about a number of illustrative examples in Separation of Concerns. The more things we correctly differentiate, the more freedom we have to optimize.

I think a lot of the content of Nonviolent Communication (NVC) is a set of important de-identifications:

  • You are encouraged to explicitly label your own experience. This helps de-fuse your experience from the experience of others, so that you don't implicitly assume everyone has experienced things the way you have. For example, you don't assume your roommate knows you find their music annoying (which can lead to the further assumption that they are annoying you on purpose).
  • You are encouraged to explicitly label wants and needs, and separate these from requests. This helps de-fuse your desires from the desires of others. For example, rather than say "you need to be quiet", NVC would encourage you to explain how the noise makes you feel, state that you (not them) has a need for quiet, and only then make a request. This facilitates group problem-solving, rather than an argument about whether the other person "needs" to be quiet.
  • You are encouraged to avoid "should", as it tends to encourage equivocation between all of the above-mentioned things.

Like the map/territory distinction, these distinctions can be gateway drugs to the general "make distinctions" thing.


12 comments, sorted by Highlighting new comments since Today at 8:26 AM
New Comment

One thing to remember when talking  about distinction/defusion is that it's not a free operation: if you distinguish two things that you previously considered the same, you need to store at least a bit of information more than before. That is something that demands effort and energy. Sometimes, you need to store a lot more bits. You cannot simply become superintelligent by defusing everything in sight.

Sometimes, making a distinction is important, but some other times, erasing distinctions is more important. Rationality is about creating and erasing distinctions to achieve a more truthful or more useful model.

This is also why I vowed to never object that something is "more complicated" if I cannot offer a better model, because it's always very easy to inject distinctions, the harder part is to make those distinctions matter.

Yeah. Probably the reason why e.g. the experience of a raw sound and the interpretation of that sound are fused together by default, is that normally it's only the interpretation we care about, and we need to be able to react to it quickly if it carries any urgent information. I made a similar observation in the essay that Abram is referencing:

Cognitive fusion isn’t necessarily a bad thing. If you suddenly notice a car driving towards you at a high speed, you don’t want to get stuck pondering about how the feeling of danger is actually a mental construct produced by your brain. You want to get out of the way as fast as possible, with minimal mental clutter interfering with your actions. Likewise, if you are doing programming or math, you want to become at least partially fused together with your understanding of the domain, taking its axioms as objective facts so that you can focus on figuring out how to work with those axioms and get your desired results. [...]

Cognitive fusion trades flexibility for focus. You will be strongly driven and capable of focusing on just the thing that’s your in mind, at the cost of being less likely to notice when that thing is actually wrong.

Yeah, totally. I think I want to defend something like being capable of drawing as many distinctions as possible (while, of course, focusing more on the more important distinctions).

One of the most distinction-heavy people I know is also one of the hardest to understand. Actually, I think the two people I know who are best at distinctions are also the two most communication-bottlenecked people I know.


if you distinguish two things that you previously considered the same, you need to store at least a bit of information more than before

Not literally. It depends on the probability of the two things. At 50/50, it's 1 bit. The further it gets from that, the more we can use efficient encodings to average less than 1 bit per instance, approaching zero.

In our community there are people who make a lot of distinction in the uncertainty via putting different probabilities on different claims but fail to distinguish levels of abstration. 

In plenty of cases it's not necessary to do fine distinctions, in others it's essential for reasoning clearly. It's worth to be able to go to fine distinctions where the occasion needs clear reasoning. 

You can think of growth mindset as a deidentification, basically identical to that example of Anna the student except done by Anna about herself rather than by her teacher. "Yet" is a wedge that gets you to separate your concept of you-so-far from your concept of future you. "I'm bad at X" sneaks in an equivocation to imply "now and always."

Ah, good example!

For me, this illustrates how obviously useful and defensible a fusion can seem -- "I am bad at math" can seem like just an empirical fact, and generalization from the past to the future seems like a very defensible heuristic. Nonetheless, using "Yet!" to drive a wedge as you describe turns out to be quite useful.

Does anyone have a good explanation of the differences between the four different levels?

Sure. I think we could add a lot more detail (and subtract a few mistaken neurological notions from his system), but his basic idea still makes sense today:

  1. The world outside the nervous system.
  2. The nervous stimulation event. EG, light hitting the retina. This includes "immediate physical-chemical-electro-coloidal" impact of the stimulation, but I'm not sure where he'd draw the boundary.
  3. Broader but still preverbal reactions. Thinking, feeling, etc. This is (at least in part) what Focusing is trying to access.
  4. Linguistic, symbolic processing. "I see a chair", "I feel hurt", etc.

He refers to 1-3 together as "the silent level" and places emphasis on trying to properly distinguish, and access, the silent level.

I'm not sure whether 4 included the internal monologue, or only actual speech. If not, it seems like Korzybski must not have thought in words. (Note how "thinking" is placed as part of the silent level, in #3.)

The internal narrator is only one form of thought.

One meditation technique is to quickly label each passing thought (it's called "noting" I believe). At some point you can begin to label the narrator process itself and see it separate from your other thinking processes ("voice" I call it, though it becomes wordless at that point).

[Edit: nevermind the Focusing link actually mentions the labeling. Though I recall Focusing was more about depth of analysis, not fast, high frequency labeling]

I haven't read Korzybski, but in general all of this sounds correct to me.

Much mas made of this in Eliezer's Sequences. 

"was" or "more". (I know the red underlining in the editor won't show up in the final version, so it's "mas" that is a typo (in English).