Daniel Dennett · The illusion of consciousness
Monterey, California, February 2003. On the TED screen, one photograph keeps flashing: an airliner parked on the ground. Picture, grey screen, picture, grey screen. The philosopher on stage asks twice, "How many still don't see it?" Then he gives it away:
How many engines on the wing of that Boeing?— Daniel Dennett, TED2003
The room laughs. The only difference between the two photos is the engine in the middle of the wing, there one moment and gone the next. "Right in the middle of the picture!" he adds, and then: "Thanks very much for your attention." It is a pun. Attention is exactly what had just failed.
The speaker is Daniel Dennett, a philosopher at Tufts University who spent his career arguing that consciousness can be explained by ordinary physical processes in the brain. His 21-minute talk has one stated goal: to shake your confidence that you are the authority on your own consciousness.
The argument runs like this. Everyone thinks they are an expert on consciousness, because "I am conscious, so I know." You feel you see a sharp, detailed world, yet the high-resolution patch of your eye is about the size of your thumbnail held at arm's length. Dennett once walked up to an eighteenth-century landscape to see the clothes of the people on a bridge, found only blobs of paint, and screamed. The brain does not send little painters to fill in the detail; it just makes you think the detail is there. He borrows a magic trick: a philosopher explains sawing the lady in half by saying the magician does not really saw her, he merely makes you think he does. And how does he do that? "Oh, that's not my department."
The finale is that flicker sequence. Psychologists call it change blindness: put a brief blank between two pictures and huge changes, a roof changing color, a shadow sliding, an engine vanishing, can go unseen for many seconds. Dennett's conclusion: science, working from the outside, can tell you things about your own consciousness that you would never dream of.
What he left out is that the original paper made a narrower claim. Ronald Rensink and colleagues, who ran the 1997 experiments with 10 observers per experiment, argued that seeing a change requires attention: give people a one-word cue about where the change is and they find it quickly, so the problem is not poor visibility. Eight years later Rensink and Daniel Simons wrote that using change blindness to show the brain stores little detail had been an "overly strong claim": lack of ability need not be caused by lack of representation.
What happened next? It partly stands. The phenomenon is rock solid. In a 2020 virtual reality study, almost a third of observers failed to notice when less than 5% of the display was in color. The half that says you are no authority on your own experience keeps gaining evidence. The half that says the detail simply isn't there remains open: philosopher Ned Block argues we are conscious of more than we can report, and a 2016 paper Dennett co-wrote softened the line itself, saying we see more than a handful of objects but "still see far less than we think we do." His larger claim, that consciousness is a bag of tricks, was still a minority view when he died in 2024.
In the last two minutes Dennett has mostly stopped talking. A photograph and its altered twin alternate on the screen, each followed by a grey blank. The first few are easy: a roof changes color and everybody sees it. Then they get harder. "What's on the roof of that barn?" "Is it a bridge or a dock?" One image has a large shadow sliding back and forth; for that one, he says, the median time for subjects was 15.5 seconds.
The last image is a parked airliner. He asks twice how many still don't see it, waits, and then:
How many engines on the wing of that Boeing?— Daniel Dennett, TED2003
Laughter. "Right in the middle of the picture!" And his next line is "Thanks very much for your attention." The pleasantry is a punchline: the audience has just shown on itself that where attention does not land, an entire jet engine can fail to exist.
You can disagree with the eighteen minutes of argument before this. The photographs don't ask for agreement. They happen to you.
Daniel Dennett (1942–2024) taught philosophy at Tufts University and wrote more than twenty books. He was an unusual kind of philosopher, one who cited eye-tracking experiments and robots rather than Kant; in this talk he uses the robotics and neuroscience talks given by other speakers that same week as evidence.
When he took the TED stage in 2003, his best-known book, Consciousness Explained, had been out for twelve years. Its claim was radical: there is no "Cartesian theater" in the brain, no place where all the senses come together and are shown to an inner audience. Our sense that consciousness is a continuous, high-definition movie is itself something the brain produces.
Those years were the moment experimental psychology began supplying him with ammunition. In 1997 Rensink, J. Kevin O'Regan and James Clark published the flicker-paradigm experiments on change blindness, and the phenomenon was soon, in Dennett's words, "being studied quite a bit." He told the audience he had predicted such results in the last page or two of his 1991 book. In the room were his fellow TED speakers: roboticist Rod Brooks, the neuroscience theorist Jeff Hawkins and AI pioneer Marvin Minsky, all of whom he named from the stage.
He opens with a joke at his own expense. At a party, when he says he is a professor, eyes glaze over. At a professors' party, when he says philosophy, eyes glaze over. At a philosophers' party, when he says he works on consciousness:
their eyes don't glaze over -- their lips curl into a snarl.— Daniel Dennett, TED2003
His colleagues think the very idea of explaining consciousness is arrogance. Then comes the diagnosis behind the whole talk: minds are hard to change on this subject because everyone thinks they are an expert. People have strong opinions about video games or climate change but don't call that expertise; about consciousness, they feel that simply by being conscious they know all about it. So he announces the job for the day:
I want to shake your confidence that you know your own innermost minds -- that you are, yourselves, authoritative about your own consciousness.— Daniel Dennett, TED2003
He lays two foundations. The first is material: we are made of about 100 trillion cells (a 2016 re-estimate by Sender and colleagues puts human cells at about 30 trillion, off by more than a factor of three, though it does not affect the point), and not one of them is conscious. How does an army of unconscious little robots add up to color, ideas and memories? The second is attitude. He reads from Lee Siegel, a religion professor and magician: people ask whether his book is about "real magic," but "real magic" means the kind that isn't real, while the magic that can actually be done doesn't count. Many people feel the same way about consciousness, Dennett says: if it can be explained as a bag of tricks, it isn't real consciousness.
Then the line that got the biggest laugh, the philosopher's explanation of sawing the lady in half: the magician doesn't really saw her, "He merely makes you think that he does." And how? "Oh, that's not my department, I'm sorry." It is self-mockery and also the template for every example that follows.
First example: an animation in which every square changes color and almost nobody notices. He explains that the eye saccades three or four times a second and that the high-resolution fovea covers about a thumbnail at arm's length. "You're getting in a lot less information than you think." Second example, from his own life. In a North Carolina museum hangs a painting by Bellotto, which he took for a Canaletto, a painter famous for fine detail. He saw people on a bridge from across the hall and walked over to make out their clothes:
And as I got closer and closer, I actually screamed.— Daniel Dennett, TED2003
Up close there were no people, only artfully placed blobs of paint. Does the brain send little painters to fill in the missing detail? His answer is the core of the talk:
The brain just makes you think that it's got the detail there. You think the detail's there, but it isn't there.— Daniel Dennett, TED2003
The third example shifts from seeing to knowing how you see. Two 3D shapes: is one a rotated copy of the other? Easy. Then he asks: did you rotate the left one in your mind, or the right one? "How do you know that's what you did?" His point is that even when you are the subject of the experiment, you don't know how you do it; you only know that you end up with certain beliefs. To learn what is really going on, "you have to go backstage and ask the magician."
Along the way comes a rare self-correction. Discussing a Necker cube illusion, he says he had first claimed the brain does no filling-in there and took that to be simply true, "But Rob Van Lier has recently shown that it isn't." A man arguing that the brain doesn't fill in detail concedes, on stage, a case where it really does.
Only then come the change-blindness photos. He walks through the procedure: the original picture for 240 milliseconds, blank, the altered one for 240 milliseconds, blank, and so on until the subject presses the button. A few images later, the Boeing. He closes by stating the conclusion outright:
in fact, you're not the authority on your own consciousness that you think you are.— Daniel Dennett, TED2003
Third-person science, working from the outside, can tell you things about your own mind. That is the flip side of the confidence he set out to shake.
What the experiment actually showed. The change blindness on stage comes from Rensink, O'Regan and Clark's 1997 paper in Psychological Science, "To See or Not to See." Each experiment had just 10 naive observers and used the same set of 48 photos of real scenes. Each image was shown for 240 ms, with a blank of 80 ms between images. On stage Dennett said the blank lasted "about a quarter of a second": he had the picture duration right and the blank more than three times too long.
The results were genuinely striking. When the change was to something people don't mention when describing the scene, observers needed on average 17.1 alternations (10.9 seconds) to identify it, and for some images more than 80 alternations (50 seconds). Changes to the objects people do mention were found much faster, after 7.3 alternations (4.7 seconds). Take the blanks out and the same changes were spotted in 1.4 alternations (0.9 seconds). The "2.9 seconds" and "15.5 seconds" Dennett quoted are for particular images; I could not find them in the paper's text, so they stand as his report.
What matters is how the paper explained the effect. In one experiment a word shown before each trial named the part that would change, and identification sped up dramatically. So the problem was not visibility but attention. The paper's central proposal is that you see a change in an object only when you give it focused attention; the global flicker swamps the local motion signal that would normally pull attention to the spot. The abstract does also make a bolder claim, that the results support the idea that "observers never form a complete, detailed representation of their surroundings." That is the layer Dennett built on.
"I predicted it," with a discount. Dennett said he predicted such experiments at the end of his 1991 book, and that Rensink's work was "in some degree inspired" by that suggestion. I could not read that page of the book, so I cannot check it. What I can check is that the 1997 paper's own introduction lists two earlier lines of work: brief-display experiments from the 1970s, and experiments from 1979 onward that swapped images during an eye movement. The phenomenon had precedents; Dennett's contribution was seeing how much it mattered for theories of consciousness.
The longer argument. The magic metaphor on stage is a whole theory in his books. Consciousness Explained compares consciousness to the user illusion of a computer interface: icons and folders correspond only abstractly to what goes on inside the machine, yet they let us operate it well. What we find by introspection is similar, a metaphorical presentation of neural events that is useful but does not show the underlying truth. In 2016 the philosopher Keith Frankish named this approach "illusionism" and called Dennett its pre-eminent defender. The line about the brain making you think the detail is there is the visual version of a bigger claim: the way consciousness seems is itself an impression the brain produces, something to be explained rather than taken as evidence.
His own later version. Thirteen years after the talk, Dennett co-wrote "What is the Bandwidth of Perceptual Experience?" in Trends in Cognitive Sciences with vision scientist Michael Cohen and neuroscientist Nancy Kanwisher. This version is more careful: we see more than "the handful of objects" allowed by prominent models of attention and working memory, but "we still see far less than we think we do." The extra comes from ensemble statistics. For large unattended regions the brain keeps a summary rather than item-by-item detail, such as "this area is mostly red" or "that crowd is moving roughly left." Compared with "the detail isn't there," this is a real concession: not nothing, but a summary.
PARTLY STANDS · as of 2026-10
Change blindness as a phenomenon is rock solid, and the claim that you are no authority on your own experience keeps gaining support. But the people who ran the experiment called the inference to "the detail isn't there" an overly strong claim, and the larger idea that consciousness is a bag of tricks is still a minority position, with no decisive evidence on either side.
2005: the experimenters withdraw the strongest layer. Simons and Rensink published a review in Trends in Cognitive Sciences, "Change blindness: past, present, and future," then in a reply to the philosopher Alva Noë put it bluntly. They still found the "sparse representations" view appealing, but had "initially made the overly strong claim that change blindness supports the conclusion of sparse representations"; the inference is "logically flawed: lack of ability need not be caused by lack of representation." You may miss the vanished engine because your brain stored it but never compared it. This does not refute Dennett, but it downgrades his loudest line on stage from "the detail isn't there" to "the detail may be there but goes unused."
From 2011: the overflow camp pushes back. NYU philosopher Ned Block, in "Perceptual consciousness overflows cognitive access" (Trends in Cognitive Sciences), used iconic-memory experiments descending from George Sperling's 1960 partial-report studies to argue that when we look at a complex scene "we are conscious of more than we can report or think about." The 2016 paper Dennett co-wrote is a reply, finding no clear empirical evidence for overflow. In 2017 four researchers including Giulio Tononi and Christof Koch struck back, arguing that "this postulated poverty of experiential content is illusory" and that binary forced-choice methods underestimate what people see. There is no verdict yet.
2018: the philosophical showdown. The British philosopher Galen Strawson, in "The Consciousness Deniers" in the New York Review of Books, attacked the view as "the silliest claim ever made" and wrote that if Dennett is right, "no one has ever really suffered." Dennett replied: "I don't deny the existence of consciousness; of course, consciousness exists; it just isn't what most people think it is." In the same letter he reached for the lady being sawn in half again, urging people to exhaust "the more mundane possibilities" first, and wrote that while we know many things about the contents of our experience, "we have no privileged insight into how this is possible, and we have no immunity to error on this score." Fifteen years on, the two core claims from the stage are unchanged. Strawson answered that this is still "denial of the existence of consciousness."
2020: the experimental side's best showing. Michael Cohen, Thomas Botch and Caroline Robertson reported in PNAS that observers wearing eye-tracking VR headsets explored 360° real-world scenes while everything outside their gaze was drained to grey. "In the most extreme case, almost a third of observers failed to notice when less than 5% of the visual display was presented in color." Their conclusion: "our intuitive sense of a rich, colorful visual world is largely incorrect." Their discussion cites Consciousness Explained for the view that the intuition of rich perception is "simply misguided." It is Dennett's Bellotto, upgraded.
Where it stands. Dennett died on April 19, 2024, at 82. His two claims have had different fates. "Introspection is unreliable, and outside science can tell you about your own consciousness" has change blindness and the VR study behind it and largely stands. "The way consciousness seems is itself an illusion" was still described by Frankish in 2016 as "a minority position," and no experiment has settled it, because change blindness fits "the detail isn't there" and "the detail is there but not accessed" equally well. The most vivid line from the stage is the one that has been proven least.