Thomas Metzinger thinks you don’t exist, and he doesn’t mean you in particular, he means nobody does. On his account there is no self, only a self-model — a representation the brain builds of itself and then takes for the real thing. We never catch the substitution because there is no way to step outside the model and inspect it as a model. We are always already inside it, and that inside, the only place any of us has ever been, is what he calls the tunnel.
The Ego Tunnel (2009) is his attempt to make this case to people who aren’t professional philosophers. I’m reading it slowly, a chapter at a time, and writing each one up partly to find out whether the argument holds once you stop nodding along and start leaning on it. This is the first of those write-ups, and it covers Chapter 2, which turns out to be mostly Metzinger refusing to give you what you came for.
Note
On how these get written: I read the chapter and think it through, a good deal of it in a long back-and-forth with an LLM, and then I lean on the LLM as an editor to turn that thinking into something readable, because I’ve never written much and I haven’t got the patience to grind my own sentences into shape. The ideas, the connections, and the judgments are mine, the wrong ones included. The prose has had help. I’d rather say that plainly than pretend otherwise.
A tour of problems, not answers
The chapter is called “A Tour of the Tunnel,” so you settle in expecting a description of the tunnel. Instead Metzinger spends almost the whole time walking through the problems that any theory of consciousness has to survive, ruling out the shortcuts one by one. It reads less like being shown a building than like watching a man clear a patch of ground so thoroughly that whatever he eventually plants there seems like the only thing that could have grown.
The first problem is that a good part of what we want explained cannot be said at all. Think of what a piece of music does to you, then try to write that down and hand it to someone who has never heard it. You can’t, and the failure isn’t one of effort, it’s structural. Metzinger calls this ineffability, and he draws the uncomfortable conclusion straight away: if some of conscious experience can never be put into words, then introspection-plus-description has a built-in ceiling as a method. You can report that you had an experience, but you can’t transmit the experience. A science resting only on what people manage to articulate is quietly missing data it can never reach by that route.
This bites harder than it first looks. The standard opening move in a lot of writing about the mind is the one he’s undercutting: tell me in careful detail what it is like, and we’ll reason from there. His point is that the careful detail leaks on the way out, and an honest theory has to account for the leak rather than carry on as if everything important made it onto the page.
Why you can’t just delete the problem
The next shortcut he closes is the eliminativist one, associated with Paul Churchland. It says our ordinary vocabulary for the mind — beliefs, desires, the redness of red — is bad folk theory that neuroscience will eventually replace, the way “phlogiston” gave way to oxygen, so that what you took for an experience of colour was never a colour experience at all but a brain state you’d been misdescribing your whole life. Churchland, to his credit, is sunny about this, and Metzinger lets him say it in his own words:
I suggest, then, that those of us who prize the flux and content of our subjective phenomenological experience need not view the advance of materialist neuroscience with fear and foreboding. Quite the contrary. The genuine arrival of a materialist kinematics and dynamics for psychological states and cognitive processes will constitute not a gloom in which our inner life is suppressed or eclipsed, but rather a dawning, in which its marvelous intricacies are finally revealed — most notably, if we apply [it] ourselves, in direct self-conscious introspection.
Metzinger doesn’t buy it, and his reason is the load-bearing idea of the whole book. Introspection can never reach the machinery, only the model. When you turn your attention inward, what you meet is the brain’s representation of itself, never the process building that representation. Sharpening your introspection doesn’t help, because the thing you’re straining to see is the thing doing the seeing. The walls of the tunnel aren’t hidden because we’ve been lazy; they’re invisible by construction. So Churchland’s optimism is aimed at a door that doesn’t open, and his deletion is too blunt besides — it throws away the thing that needed explaining and calls that an explanation.
The question he won’t let you ask
Here Metzinger gets genuinely contrary. The most natural question about consciousness is what is it for, what payoff made it worth building, and he refuses to let you ask it in that form. He’s right to. Evolution doesn’t aim at anything; it’s a blind process of variation and selection working on whatever happens to be lying around. Much of what it produces isn’t an adaptation at all, and even the genuine adaptations aren’t optimal, since selection can only tinker with what already exists. Consciousness might carry real survival value, or it might be a by-product that came along for the ride and never paid rent. Nothing about how it arose guarantees it is for anything.
First, let’s not forget that evolution is driven by chance, does not pursue a goal… It is incorrect to assume that evolution had to invent consciousness — in principle it could have been a useless by-product. No necessity was involved.
This shuts off the reverse-engineering strategy, where you guess what consciousness does and reason backwards to what it must be. With that door closed, Metzinger reels off, almost wearily, the long list of functions people have proposed for it: motivation, social coordination, memory retrieval, planning, conflict resolution, one-shot learning, the assembly of a single integrated picture of the world. The length of the list is the argument. When everyone’s favourite function is different, that’s a fair sign the function-first approach isn’t converging on anything, and that the question to settle first is not what consciousness does but what kind of thing it is.
The one functional idea he likes
For all the refusal, he doesn’t walk away empty-handed. The one account he takes seriously is Bernard Baars’s global workspace theory, and it’s worth stating carefully, because it’s the first thing in the chapter that feels like an actual mechanism. The bulk of what your brain does is handled by fast, specialised, unconscious machinery. Consciousness is held in reserve for information that has to be made broadly available, precisely because the brain doesn’t yet know which subsystem will need it next. Will you have to speak about this, act on it, hold it against a memory, weigh it against an alternative? When the next move is uncertain, the information gets posted where every subsystem can see it, and that broadcast is what consciousness is. Metzinger puts the criterion in Baars’s own framing:
conscious information is that subset of active information in the brain that requires monitoring because it’s not clear which of your mental capacities you will need to access this information next… Part of Baars’s idea is that you become conscious of something only when you don’t know which of the tools in your mental toolbox you’ll have to use next.
The analogy that finally made it click for me comes from computing: conscious information looks a lot like RAM, data pulled into a general-purpose space exactly because its destination isn’t fixed yet. The unconscious specialists are more like cache or dedicated circuits, fast because each only ever does one thing. RAM is slow and expensive by comparison, which maps onto a real and slightly humbling fact about consciousness — it’s metabolically costly, serial, and sluggish next to all that automatic processing, the kind of price you’d only pay for flexibility. And once you have that picture, the long list of proposed functions reads differently. Memory, planning, social reasoning, conflict resolution aren’t rival theories of what consciousness is. They’re the various programs that might request the data. Consciousness is the loading dock, and the functions are the trucks.
The self as a virtual organ
The best single idea in the chapter, and one I hadn’t run into before, is the virtual organ. Metzinger needs a category for something physically real that does real work and yet isn’t a fixed structure you could point to on a dissection table. The example he reaches for is the immune response. Nobody doubts it exists, and yet it has no address; it’s assembled on demand out of cells that are constantly replaced, it does its job, and then it stands down. It’s an organ that exists as a process rather than a place.
Virtual organs are like physical organs in that they fulfill a specific function… Unlike a liver or a heart, they are realized transiently.
Consciousness is a virtual organ in this sense, and so is the self. Both are real, both are made of nothing but neurons firing in concert, and neither sits anywhere in particular. This is the move that lets Metzinger have it both ways without cheating. The self is neither an illusion you ought to stop believing in nor a soul-shaped object tucked away in the skull. It’s a transient assembly, present while it’s running and gone when it isn’t, which is why it can switch off in dreamless sleep, under anaesthesia, or in deep meditative states without the person dying with it. The organ was never the building. It was the weather inside it.
I kept thinking, all through this section, of Ghost in the Shell. The anxiety underneath that whole story is whether the Major’s “ghost” — the thing that makes her her — can survive having her body swapped out piece by piece for hardware. Metzinger’s answer would be that the question is confused at the root. The ghost was never welded to the substrate in the first place; it was always a virtual organ, always assembled on the fly. The ghost is real, and yet it isn’t something she could lose by changing the parts, because it was never one of the parts.
Where this is going
So by the halfway mark of the chapter, the ground is cleared and two things are left standing on it. Consciousness is the brain’s general-purpose workspace, the place information goes when its eventual use hasn’t been decided. And the self that seems to sit at the centre of it is a virtual organ — genuinely present, genuinely physical, genuinely temporary. That sets up the harder turn. Once you can run a model of a self, you can run models of other selves, and once you can do that, you can reach into someone else’s model and change it. That’s where Metzinger says something quietly alarming about what a lie actually is, and where the question of who, exactly, is home inside the tunnel starts to press. But that’s Part 2.