Do You Need Concepts to See?

Imagine you’re looking at a wall painted a very specific shade of blue—not just “blue,” but that blue, the one you’ve never seen before and don’t have a name for. You can see it perfectly clearly. You could pick it out from a row of almost-identical blues. But you don’t have a word or a concept for it. You can’t think to yourself “ah yes, that’s blue #7,482.” You just… see it.

Here’s a weird question philosophers have been arguing about for decades: Does the fact that you can see that shade without having a concept for it tell us something important about how your mind works? Or is it just a boring fact about your vocabulary?

It turns out this question connects to bigger puzzles: Do animals see the world the same way we do? Could you have experiences without having any concepts at all? And what does it even mean to say that someone “represents” the world?

The Basic Idea

Let’s start simple. When you believe something—say, “my dog is hungry”—the content of that belief (what you actually believe) seems to depend on concepts you possess. You need the concept DOG and the concept HUNGRY to have that belief. If you didn’t know what dogs were, you couldn’t believe that your dog is hungry. That seems obvious.

Philosophers call this the conceptual constraint: when we describe what someone believes, we should use only concepts that person actually possesses. It would be weird to describe a toddler’s belief about her pet using technical concepts from biology that she doesn’t know.

But here’s the question: Does this same constraint apply to perception? When you see that unusual shade of blue, should the description of what you see be limited to concepts you possess? Or can your visual experience represent the world in ways that go beyond your concepts?

Philosophers who say “yes, perception goes beyond concepts” are called nonconceptualists. They think there’s such a thing as nonconceptual content—ways of representing the world that don’t require you to have the concepts used in describing them. Philosophers who say “no, perception is always shaped by concepts” are called conceptualists.

Nobody really knows who’s right. The debate is still very much alive.

The Fineness of Grain Argument

The most famous argument for nonconceptual content starts with something you’ve probably noticed yourself: your ability to see differences is much finer than your ability to name or remember them.

Think about faces. You can recognize hundreds of faces, but try describing someone’s face precisely enough for another person to pick them out of a crowd. You can’t—unless you’re a police sketch artist. Your perception is incredibly detailed, but your conceptual abilities are coarser.

Here’s a more precise version of the argument, first made by the philosopher Gareth Evans in the 1980s. Look at two color chips that are very slightly different shades of red—call them Red₁ and Red₂. You can see the difference between them. But unless you’re a paint expert, you probably don’t have separate concepts for Red₁ and Red₂. You just have the concept RED. So you see a difference that your concepts can’t capture. That means your visual experience represents more detail than your concepts can handle. Therefore, perceptual content is nonconceptual.

This seems plausible. But a philosopher named John McDowell had a clever response.

McDowell’s Response: Demonstrative Concepts

McDowell said: wait a minute. You do have a concept for that specific shade. When you see it, you can think “that shade.” The concept expressed by “that shade” is a demonstrative concept—a concept you form on the spot by pointing to something. It’s as fine-grained as your perception, because it just picks out whatever shade you’re looking at. So your conceptual capacities are fine-grained enough. The only reason you can’t describe the shade to someone else is that you can’t produce the sample for them to see.

This is a clever move. But nonconceptualists have a reply.

Suppose you see that unusual blue shade today, and then see it again next week. You probably won’t recognize it. You’ll think “hmm, that looks familiar” but you won’t know for sure it’s the same shade. Your perception of the shade was detailed, but your memory isn’t. Here’s the problem for McDowell: if your concept “that shade” is supposed to capture your experience, then the concept should be something you can use again. Concepts are supposed to be reusable. But you can’t reuse “that shade” once the sample is gone—you can’t pick the shade out again. So either the demonstrative concept isn’t a genuine concept (since it can’t be reused), or your perception really does outrun your concepts.

This part gets complicated, but here’s what the debate boils down to: Can there be a kind of concept that exists only in the moment you’re seeing something? McDowell thinks yes. His opponents think that if something is a concept, you should be able to think with it later, not just while staring at the sample.

Another Argument: Contradictory Experiences

Here’s a different reason to think perception is nonconceptual. Have you ever seen a waterfall, then looked away at a rocky cliff and had the weird experience of the rocks seeming to move upward? That’s the waterfall illusion (also called the motion aftereffect). After staring at moving water for a while, stationary objects look like they’re moving in the opposite direction.

Here’s the interesting thing: the rocks simultaneously seem to be moving and not moving. You know they’re stationary, but they look like they’re moving. Your experience has contradictory content.

Philosopher Tim Crane argued that this shows perception must be nonconceptual. Why? Because you can’t have a belief with contradictory content. You can’t believe “the rocks are both moving and not moving.” Beliefs have to be logically consistent. But your experience can be contradictory—the waterfall illusion proves it. So experiences must be a different kind of mental state than beliefs, with a different kind of content. And since beliefs have conceptual content, experiences must have nonconceptual content.

Some philosophers have challenged this. Maybe the experience isn’t really contradictory—maybe you just have two different experiences simultaneously, and the contradiction only appears if you try to put them into language. But Crane’s argument remains influential.

Animals and Babies

Here’s a third motivation for nonconceptual content, and one that many philosophers find the most compelling. Think about a dog chasing a frisbee. The dog clearly perceives the frisbee—its arc, its speed, its location. The dog can track it through the air and catch it. But does the dog have the concept FRISBEE? Does it have concepts at all?

If you think concepts require language (and some philosophers do), then the dog can’t have concepts. But the dog obviously has perceptual experiences. So the dog’s perceptual content must be nonconceptual.

The same goes for human babies. A six-month-old can track objects, recognize faces, and get surprised when things don’t behave as expected (in experiments, babies stare longer at impossible events, like a ball passing through a solid wall). But six-month-olds don’t have concepts yet—or at least, they don’t have the rich kind of concepts adults have. So their perception must be nonconceptual.

Here’s the punch: if we share basic visual systems with dogs and babies, then our own perception probably also has nonconceptual content. We may add conceptual understanding on top, but the raw perceptual layer is nonconceptual.

This argument doesn’t prove the point, but it makes the nonconceptualist position look very natural.

What Makes Content “Conceptual” Anyway?

You might have noticed that this whole debate depends on what you think a “concept” is. Philosophers disagree about this too.

At one extreme, a concept might just be the ability to tell things apart. If you can distinguish red things from non-red things, you have the concept RED. On this view, dogs and babies have lots of concepts. Nonconceptual content would be almost nonexistent.

At the other extreme, having a concept might require being able to justify your judgments and integrate them into a system of reasoning. On this view, dogs and babies don’t have concepts at all. Nonconceptual content would be everywhere.

Most philosophers fall somewhere in between. They think concepts require more than just discrimination, but less than full-blown linguistic justification. But exactly where you draw the line determines whether you think perception is conceptual or nonconceptual.

This is one reason the debate is still alive. Philosophers aren’t just arguing about perception—they’re arguing about what concepts are, and that’s a messy question.

The Two Flavors of Nonconceptualism

There’s an important distinction that philosophers have noticed. When someone says “perception is nonconceptual,” they might mean one of two things:

State nonconceptualism: You can be in a perceptual state without having the concepts needed to describe that state’s content. (This is the claim about the conditions for having the experience.)

Content nonconceptualism: Perceptual content is a fundamentally different kind of thing than conceptual content. (This is the claim about what the content itself is like.)

These sound similar, but they’re different. You could accept state nonconceptualism without accepting content nonconceptualism. Maybe perceptions and beliefs have the same kind of content, but perceptions just happen not to require concepts.

Some philosophers (like Robert Stalnaker) hold this view. They think all content—for both perception and belief—should be understood in terms of possible worlds (the set of ways the world could be for the content to be true). This kind of content isn’t conceptual at all, but it’s the same kind for everything. The difference between perception and belief isn’t in their content but in other features.

Other philosophers think this state/content distinction is artificial. They argue that if perceptions don’t require concepts, that must be because their content is a different type. You can’t separate the two.

There’s no consensus here either.

How to Specify Nonconceptual Content

Suppose you accept that perception has nonconceptual content. How do you describe it? What are the “units” of nonconceptual content?

Philosopher Christopher Peacocke proposed an elegant answer. He said the nonconceptual content of a visual experience is a scenario—a way of filling out the space around you. For every direction you could look, the scenario specifies what surface would be there, what color, what orientation, what texture. The content is correct if the world actually has those surfaces in those places.

This captures something important about perception: it’s all about spatial layout. When you open your eyes, you don’t just get a list of features (“red, round, moving left”). You get a three-dimensional world spread out around you.

But scenario content isn’t enough to capture everything. Remember the square-diamond example? Look at a tilted square: you can see it as a tilted square or as an upright diamond. The scenario content—the actual spatial layout—is identical in both cases. Yet the experiences feel different. So there must be another layer of nonconceptual content that captures how objects are organized.

Peacocke called this proto-propositional content. It involves objects and properties, but not concepts. You perceive the square as a square or as a diamond, and these are different ways of organizing the same visual information. Neither requires you to have the concept SQUARE or DIAMOND.

Other philosophers have different proposals. Some think nonconceptual content should be understood in terms of what the organism can do—the actions the environment affords. Some think it should be understood in terms of biological functions—what our visual systems evolved to detect. None of these proposals has won universal acceptance.

Can Nonconceptual Content Stand Alone?

Here’s a radical question: Could a creature have nonconceptual content without any concepts at all? Philosophers call this the Autonomy Thesis—the idea that nonconceptual content can be completely independent of conceptual content.

This matters for explaining animal and infant behavior. If you want to say that a baby represents the world without having any concepts, you need the autonomy thesis to be true.

Peacocke originally argued against this. He thought that having genuine spatial experiences required some minimal conceptual abilities—in particular, the ability to reidentify locations and track your own position. This, he thought, required at least a rudimentary concept of yourself as a subject.

But other philosophers pushed back. José Luis Bermúdez argued that you could have a nonconceptual point of view—a way of locating yourself in space—without having a concept of yourself. This would be enough to have spatial experiences. The information is there in your perception and proprioception; you don’t need a concept to use it.

Eventually Peacocke changed his mind and accepted the autonomy thesis. But many philosophers remain skeptical.

Why This Matters (Besides Being Interesting)

You might wonder: who cares whether perception is conceptual or nonconceptual? This feels like a technical dispute among philosophers who spend too much time staring at color chips.

But the debate has real consequences.

First, it matters for epistemology—how we know things. Many philosophers think that perception justifies our beliefs about the world. But how can a nonconceptual experience justify a conceptual belief? If your experience represents the world in a fundamentally different format than your beliefs, it’s not obvious how the experience can serve as evidence for the belief. This is a live problem.

Second, it matters for cognitive science. If you think subpersonal processes (like early visual processing) have nonconceptual content, you need a theory of what that content is. And you need to explain how nonconceptual subpersonal content relates to conceptual personal-level thought.

Third, it matters for animal cognition. If animals don’t have concepts, we can still explain their intelligent behavior by attributing nonconceptual representations to them. But then we need to say what nonconceptual representation is, and how it differs from conceptual thought.

Fourth, it matters for self-consciousness. Some philosophers think there are primitive, nonconceptual forms of self-awareness that don’t require a concept of self. If they’re right, then self-consciousness might be more widespread in the animal kingdom than we think.

What’s Still Unresolved

The debate about nonconceptual content isn’t going away anytime soon. Here are some of the biggest unanswered questions:

What exactly is a concept? Until philosophers agree on this, it’s hard to settle whether perception is conceptual or nonconceptual.
Can perception justify belief if it has nonconceptual content? This is the deepest challenge. Even if perception is nonconceptual, we need an account of how it can serve as a reason for belief.
Is there one kind of nonconceptual content or many? The kind of content that explains animal behavior might be very different from the kind that explains fine-grained color perception. These might be different phenomena wearing the same label.
Does nonconceptual content extend beyond vision? Recent work has explored nonconceptual content in smell, emotion, and even numerical cognition (your ability to roughly estimate quantities without counting). How far does the notion stretch?

Nobody has the final answers. But that’s part of what makes the debate interesting—it forces you to think carefully about what perception is, what concepts are, and how they relate.

Appendices

Key Terms

Term	What it does in this debate
Nonconceptual content	A way of representing the world that doesn’t require the person to possess the concepts used in describing it
Conceptual content	Content that the subject must have concepts for, like the content of beliefs and thoughts
Conceptual constraint	The idea that descriptions of someone’s mental content should only use concepts that person actually possesses
Demonstrative concept	A concept like “that shade” that picks out something by pointing to it in the moment
Fineness of grain	The idea that perception can represent more fine-grained differences than our conceptual repertoire can capture
Scenario content	Peacocke’s proposal for nonconceptual content: a specification of how space around the perceiver is filled
Autonomy thesis	The claim that nonconceptual content can exist in a creature that has no concepts at all
State vs. content nonconceptualism	A distinction between the claim that perceptual states don’t require concepts and the claim that perceptual content is a different kind of thing

Key People

Gareth Evans (1946–1980): A British philosopher who died young but introduced the idea of nonconceptual content into modern philosophy. He thought the information from our senses is nonconceptual until our thinking system processes it.
John McDowell (b. 1942): A South African-born philosopher who defends conceptualism—the view that perceptual content is always conceptual. He argues that demonstrative concepts like “that shade” are fine-grained enough to capture perception.
Christopher Peacocke (b. 1950): A British philosopher who developed detailed theories of nonconceptual content, including the idea of scenario content. He changed his mind about whether nonconceptual content can be autonomous.
Tim Crane (b. 1962): A British philosopher who argued that perceptual illusions with contradictory content (like the waterfall illusion) show that perception must be nonconceptual.
José Luis Bermúdez (b. 1963): A Spanish-born philosopher who argues for autonomous nonconceptual content and uses it to explain animal cognition and nonconceptual self-consciousness.

Things to Think About

The demonstrative concept “that shade” seems like a good way to capture fine-grained perception. But can you really think with that concept later, when the sample is gone? If not, is it really a concept? What would it take for it to be a genuine concept?
If animals have nonconceptual content, does that mean they experience the world differently than we do? Or do we share the same basic perceptual world, with concepts just adding a layer on top? How would you test this?
Could there be a creature that has only nonconceptual content—no concepts at all—and still navigates the world successfully? What would that creature’s mental life be like?
The waterfall illusion seems to show that perception can have contradictory content. But maybe the experience isn’t really contradictory—maybe it’s just confusing. How could you tell the difference between an experience that’s genuinely contradictory and one that’s just hard to describe?

Where This Shows Up

Artificial intelligence: When AI systems process visual information without having anything like human concepts, are they operating with nonconceptual content? The debate shows up in discussions of whether AI really “understands” what it sees.
Animal cognition research: Scientists who study animal intelligence need some way to describe what animals represent without assuming they have human-like concepts. The notion of nonconceptual content provides one framework.
Child development: Developmental psychologists study what infants can perceive before they acquire language. The concept/nonconcept distinction helps structure their theories.
Virtual reality and illusions: Understanding how perception can present contradictory information (like in the waterfall illusion) has practical applications for designing VR experiences that feel real but aren’t.