S1-212 – Knowing the name vs. knowing the thing

Three levels of understanding and why most of us live at the first

When Richard Feynman was a boy, his father took him for walks in the woods outside New York City and pointed out the birds. Other fathers named the birds. Feynman’s father did something different. He would say: “See that bird? It’s a Brown-throated Thrush. But in Portuguese it’s a Rufous-bellied Thrush. In Italian it’s a Reddish-breasted Thrush. In Chinese it’s a Chuen Tong. In Japanese it’s an Otera Ko Ko. In German it’s a Rotbrust. And so on. And when you’ve finished with all that,” he’d say, “you’ll know absolutely nothing whatsoever about the bird. You’ll only know about humans in different places, and what they call that bird. So let’s look at the bird.”

Feynman told this story many times. It was, he said, the most important lesson he ever received, more important than any of the physics he later mastered. And the lesson is deceptively simple: the name of a thing is a fact about language. The thing itself is something else entirely. Knowing that a bird is called a Brown-throated Thrush tells us what English speakers have agreed to call a particular kind of bird. It tells us nothing about how the bird navigates at night, what it eats, how it finds a mate, why it sings when it does, what biological machinery produces the song, or what evolutionary pressures produced the machinery. The name is a handle on the box. The name is not the contents.

This distinction, between knowing what something is called and knowing what something is, is the subject of this article. It sounds obvious when stated plainly. It is violated so consistently, in so many domains, that the violation has become one of the most consequential epistemic failures of contemporary life.

The first level: the name

The first level of understanding is the ability to name. The child who has learned the names of the planets can recite them in order. The student who has learned the names of the logical fallacies can identify them in arguments. The professional who has learned the terminology of their field can deploy it fluently in conversation. The politician who has learned the vocabulary of economics can discuss interest rates, inflation, and fiscal multipliers without understanding the mechanisms by which any of these phenomena operate.

Naming is not worthless. Names are the necessary scaffolding of thought and communication. Without agreed names for things, the accumulated knowledge of a discipline cannot be efficiently transmitted, applied, or built upon. The medical student who learns the names of the cranial nerves, the bones of the hand, the stages of cellular division, is not wasting their time: they are acquiring the vocabulary that will make everything that follows comprehensible. Names are the necessary first step toward understanding.

But names are only the first step. And the problem that this article is about is the consistent confusion of the first step with the destination, the treatment of fluent naming as though it were understanding of the thing named. This confusion is so widespread and so consequential that it has a name of its own in the educational psychology literature: the illusion of explanatory depth, discussed in article S1-202 in the context of chunking and the knowledge illusion. The name activates a chunk. The chunk carries the feeling of access to associated meaning. The feeling of access is mistaken for genuine understanding. The understanding is never there.

What reveals the confusion, in every domain, is the attempt to explain. The test is simple: can we explain what this concept means, in plain terms, to a curious person who has never encountered it before, without using the technical vocabulary of the domain? If the explanation collapses back into jargon as soon as the surface is pressed, if the definition of the term requires other terms that require other terms that eventually loop back to the original, then the name is doing the work that understanding should be doing. The chunk is being accessed rather than the structure beneath it. The box is being pointed to rather than opened.

The second level: the model

The second level of understanding is the ability to model. The person who understands at the second level can do more than name the phenomenon: they can describe the mechanism, the causal structure, the relationship between components, the conditions under which the phenomenon occurs and the conditions under which it does not. They can predict. They can explain not just what the thing is called but what it does, how it does it, and what would happen if some element of the system were changed.

The second level is where most of what is usually called education aims to deliver. The physics student who understands Newton’s second law not merely as F equals ma but as a claim about the relationship between force, mass, and acceleration, who can use it to predict the trajectory of a projectile, to analyze the forces on an inclined plane, to calculate the required thrust for a given acceleration, is operating at the second level. The economist who understands the Phillips curve not merely as a name for an inverse relationship between inflation and unemployment but as a claim about a specific mechanism, with specific conditions of validity and specific failure modes, is operating at the second level.

The gap between the first and second levels is large and is often invisible from the first level. The person who knows only the name typically does not know that there is a second level: the chunk presents itself as a complete representation, as article S1-202 described. The person who has begun to understand the model typically knows that the model is incomplete; they can see the edges of what they understand and glimpse the territory beyond. This is the valley of despair that article S1-205 described: the ascent from the first level to the second is the moment at which the fluency of the name gives way to the difficulty of the mechanism, and confidence drops as the true extent of the territory comes into view. As article S1-205 also notes, the dramatic quantitative version of the Dunning-Kruger effect does not survive statistical scrutiny, but this particular experience, the loss of confidence that accompanies the first real encounter with a domain’s complexity, is familiar to anyone who has genuinely entered a field of study, whatever the fate of the original chart.

The transition from the first to the second level is also where the map-territory distinction of article S1-203 becomes practically urgent. At the first level, the name is the map. At the second level, the model is the map: a better map, one that can generate predictions, that can be tested against observation, that can be revised when the predictions fail. But it is still a map. The second level carries its own characteristic failure mode: the expert who has internalized a model so thoroughly that they mistake it for the territory, who forgets that the model was built from a specific dataset, tested in specific conditions, calibrated to specific domains, and who applies it beyond those conditions with a confidence that the model’s actual scope does not warrant.

The third level: understanding

The third level is the hardest to describe precisely, because it is the level that cannot be fully captured in explicit terms, which is, in a specific sense, what distinguishes it from the second level. The third level is the understanding that knows its own limits, that can see where the model runs out and the territory continues, that can feel the difference between what the model genuinely explains and what it merely accommodates. It is what Feynman meant when he said that he could live with not knowing: that the ability to hold open questions without forcing premature closure was more valuable than any specific piece of knowledge.

The third level is not merely more model. It is a different relationship to the model. The physicist at the third level of understanding does not merely know more equations than the physicist at the second level. They have a sense (partly explicit, partly tacit, partly embodied in years of experience with specific kinds of problems) of which equations apply, which approximations are valid, which regime the current problem is in, and where the mathematics is tracking something real and where it is tracking its own internal structure. This sense is what Polanyi called tacit knowledge: we know more than we can tell, and the more we know at the third level, the more clearly we see the limits of what we can tell.

The third level is also what Peterson’s observation about acting out beliefs without understanding them addresses from a different direction. In his lectures, Peterson observes that a person can act in accordance with a framework, can live by a set of values, can navigate a social world, can be guided by a narrative, without being able to articulate the framework explicitly. The acting out precedes the understanding. The child who has internalized a moral framework through observation, imitation, and correction is at the third level with respect to that framework (their behavior is guided by something real and complex) long before they reach the second level of being able to articulate what the framework is. This is not a failure of understanding. It is a different kind of understanding, one that the emphasis on explicit articulation systematically undervalues.

The Mayan astronomers: predictive power without causal understanding

The history of science provides a striking illustration of the gap between the second and third levels that is directly relevant to the series’ account of what a good model is. The Mayan astronomical system, developed over centuries of careful observation, could predict celestial events (solstices, equinoxes, solar and lunar eclipses, the movements of Venus) with extraordinary accuracy. The tables and calendars produced by Mayan astronomers were better predictive instruments, for the phenomena they covered, than anything available in Europe at the same period.

But the Mayan system had no causal model of why the celestial bodies moved as they did. It had pattern, the observation that phenomena recurred at specific intervals, and the mathematical tools to project those patterns forward in time. What it lacked was mechanism: an account of the forces, relationships, and physical processes that produced the patterns. It was, in the terms of article S1-102, a model with strong predictive power and weak explanatory power: reliable within its domain, unable to generalize to phenomena outside that domain, and unable to diagnose its own failure modes because it had no access to the causal structure that its predictions tracked.

The distinction matters for this article because the Mayan astronomical system is the limit case of second-level understanding: a model so refined that it can generate accurate predictions without any access to the third level of understanding why the predictions are accurate. The astronomical tables are second-level knowledge without third-level understanding: a map that works without knowing why it works, and that therefore cannot tell us where its working ends.¹

This is the situation of many domain practitioners who have developed genuine predictive expertise through experience but who cannot articulate the basis of their predictions. The experienced clinician who can predict patient outcomes more accurately than any available formal model, but who cannot specify the variables on which the predictions are based. The chess grandmaster who can evaluate positions without being able to say how. The experienced teacher who knows which students are struggling before any assessment reveals it. In each case, the predictive performance is real and valuable. And in each case, the inability to access the causal basis of the prediction is also real, and it is the source of the characteristic failure mode: the expert who cannot recognize when they have left the domain of their expertise, because they cannot see the edges of the pattern that underlies their predictions.

The Feynman test

Feynman developed a practical method for distinguishing the levels of understanding that is simple enough to apply to any belief, any concept, any claimed area of expertise. It has since become widely known as the Feynman Technique, though it is less a technique than a diagnostic.²

The procedure is this. Take a concept one believes one understands. Write down an explanation of it as though it were being explained to a curious person with no background in the field: a 12-year-old, or a highly intelligent adult who has simply never encountered the subject. Do not use the technical vocabulary of the domain. Use plain language, concrete examples, and analogies. The point at which the explanation stalls, where it cannot continue without resorting to jargon, where the analogy breaks down, where the concrete example does not exist, is the limit of the understanding. That limit is not the end of the map one thought one had. It is the edge of the territory one has genuinely explored.

The test is uncomfortable because it consistently reveals that the limit is closer to the surface than the feeling of understanding suggested. The student who thought they understood supply and demand discovers, when they try to explain it without using the words “supply,” “demand,” “price,” or “equilibrium,” that the mechanism they thought they had is not there in the form they believed. The professional who thought they understood the regulatory framework that governs their industry discovers, when they try to explain why the framework has the features it does, why this requirement exists, what problem that provision was designed to solve, that what they knew was the name of the framework and a set of procedural rules, not an understanding of the structure.

The Feynman test is not a counsel of paralysis. It does not require that one achieve the third level of understanding before acting on a belief, making a decision, or expressing a view. Almost all action is based on second-level understanding at best, and much of it on first-level understanding: naming and association rather than mechanism. The point is not to refuse to act until understanding is complete. It is to know, as precisely as possible, which level of understanding the action is based on, and to hold the action with the corresponding degree of confidence: high confidence where the mechanism is understood and tested, lower confidence where the model is known to be approximate or domain-limited, and explicit acknowledgment of uncertainty where the name is all there is.

Peterson’s fourth mode: enacted knowledge

Jordan Peterson’s contribution to this framework introduces a fourth mode of understanding that does not fit neatly into the three levels described above, and that is important enough to examine separately.³ Peterson’s observation, developed in Maps of Meaning and across many lectures, is that there is a form of understanding that is prior to all three levels: the understanding that is enacted, that is lived, that is expressed in behavior and orientation before it is available for explicit articulation.

Peterson’s example is typically religious or ethical. A person can act in accordance with a deep ethical framework, can respond with genuine care, genuine courage, genuine fairness in difficult situations, before they can articulate what the framework requires, why it requires it, or how they arrived at it. The behavior is guided by something real and complex. The articulation of that something, the production of a second-level model of the ethical framework, comes later, if it comes at all. And Peterson’s argument is that the enacted understanding is not inferior to the articulated model. In some respects it is superior: it is more complete, more contextually sensitive, more responsive to the specific features of the situation, than any explicit model of the same framework could be.

This fourth mode complicates the simple hierarchy of name, model, and understanding in an important way. It suggests that explicit articulation, the criterion that the Feynman test uses to identify the limit of understanding, is not the only form of genuine understanding. It is one form. There is another form that is prior to it, that the Feynman test cannot access and that explicit modeling cannot fully capture. The person who acts wisely in a difficult situation, without being able to say why, has a form of understanding that the person who can articulate every relevant principle but who acts unwisely does not.

The implication for this series is specific. The Conscious Look is not primarily an exercise in explicit articulation. It is a practice: a way of holding one’s beliefs and models in relation to the world that produces behavior different from the behavior of someone who simply acts on unreflective conviction. The distinction between the three levels of explicit understanding and the fourth mode of enacted understanding is the distinction between the theory of The Conscious Look and its practice. The theory is useful. The practice is what matters.

The diagnostic question

The diagnostic question for this article is the one that the Feynman test operationalizes. Applied to any belief held with confidence, it asks: can I explain the mechanism, not the name, not the vocabulary, but the actual causal structure, the actual process, the actual relationship between components, in plain terms, without technical vocabulary, to a person who has never encountered the subject?

If the answer is yes, and the explanation survives cross-examination, the belief has at least second-level grounding. If the answer is no, if the explanation collapses back into jargon, if the analogy breaks down immediately, if the mechanism is simply not there, then the belief is at the first level: name and association without mechanism. Holding it with the confidence appropriate to a second-level understanding is an epistemic error. It is not a devastating error, since we cannot operate without first-level beliefs, and many first-level beliefs are approximately correct for most practical purposes. But it is an error that matters when the belief is consequential: when it informs policy, guides medical decisions, shapes political commitments, or provides the basis for overriding the preferences of people who hold different beliefs.

The specific version of this question that Feynman applied to himself is worth noting. Feynman did not merely apply the test to claims about physics. He applied it to the claim of understanding itself: to the question of what it means to know something, what counts as an explanation, what would have to be true for an explanation to be satisfying. The recursion is not infinite, but it goes one level deeper than most people take it. The Conscious Look, applied to knowledge itself, asks not only “do I understand this?” but “what would it mean for me to understand it, and am I sure I know what understanding is?”

Notes

¹ The Mayan astronomical achievement is relevant to the series’ account of what a good model is, developed in article S1-102, in a specific way. Article S1-102 identified four criteria for a good model: explanatory power, predictive power, grounding in deeper principles, and knowledge of limits. The Mayan system had strong predictive power and essentially no explanatory power. It was not grounded in any account of the physical mechanisms producing the celestial motions it tracked, and it could not know its own limits because it had no access to the causal structure within which those limits would be visible. It is the clearest historical example of what article S1-102 called borrowed reliability: a model whose predictive success is inherited from a deeper regularity that the model does not represent, and that therefore fails silently when conditions change in ways that the deeper regularity would predict but the surface pattern does not register.

² The Feynman Technique as a pedagogical method has been widely discussed and adapted since Feynman’s death, primarily in self-improvement and educational contexts. The attribution of a specific four-step technique to Feynman is in some respects a posthumous formalization of practices he described informally in various places. The core principle, that the attempt to explain a concept in plain terms without technical vocabulary reveals the limits of understanding, is clearly Feynman’s own, and appears in several places in his published writing and recorded lectures. The clearest statement is in the Caltech commencement address of 1974, in which Feynman described “cargo cult science,” the performance of the surface features of scientific practice without the substance, and argued that honest acknowledgment of what one does not know is the first requirement of genuine scientific thinking.

³ Peterson, J. B. (1999). Maps of Meaning: The Architecture of Belief. Routledge. The concept of enacted or embodied understanding that Peterson develops draws on several traditions: the phenomenological tradition of Heidegger and Merleau-Ponty, in which being-in-the-world is prior to explicit representation; the ethological tradition of Lorenz and Tinbergen, in which behavioral patterns precede their cognitive representation; and the Jungian tradition of archetypes as structures that organize experience and behavior prior to conscious articulation. The synthesis Peterson achieves is distinctive and not uncontroversial, but the specific claim that enacted understanding precedes and often exceeds explicit understanding is well-supported across all three traditions and is independently established in the cognitive science literature on embodied cognition and tacit knowledge.

S1-212 – Knowing the name vs. knowing the thing

Three levels of understanding and why most of us live at the first

The first level: the name

The second level: the model

The third level: understanding

The Mayan astronomers: predictive power without causal understanding

The Feynman test

Peterson’s fourth mode: enacted knowledge

The diagnostic question

Further reading

Notes

Leave a reply Cancel reply

S1-212 – Knowing the name vs. knowing the thing

Three levels of understanding and why most of us live at the first

The first level: the name

The second level: the model

The third level: understanding

The Mayan astronomers: predictive power without causal understanding

The Feynman test

Peterson’s fourth mode: enacted knowledge

The diagnostic question

Further reading

Notes

Share this:

Leave a reply Cancel reply