In 1973, a patient known as DB had surgery to remove a tumor pressing on his right occipital cortex. Afterward, he had a scotoma in his left visual field — a region of clinical blindness, confirmed by standard perimetry. Ask him whether he could see anything on his left side, and he said no. He wasn't being difficult about it. He meant it.
Lawrence Weiskrantz and Elizabeth Warrington, in 1974, asked him a different question. They showed him stimuli in his blind field and said: we know you can't see anything, but we want you to guess anyway — is it on the left or right? DB said left. He was right. They tried again. Right again. When they tallied the first hundred trials, DB had been correct on 86 of them. He reported subjective awareness — seeing — on exactly one.
The first question gave one answer. The second question gave another. The patient was the same. The visual system was the same. The stimuli were the same. What changed was which output channel the question used.
The clinical question — "do you see this?" — asks the introspective report. It is asking: what is present in your conscious experience? DB's answer was accurate. His conscious visual experience of the left field was genuinely absent. He had correct introspective access to what was in his experience, and he reported it correctly. Nothing.
The forced-choice question bypasses the introspective report entirely. It asks for a behavioral output correlated with visual information, without requiring the visual information to pass through consciousness first. Something was processing the location of the stimulus — not DB consciously, but some part of his visual system, operating below the level of experience. That processing could guide a guess. It could not, and did not, produce phenomenal experience. But it could tell which side.
Weiskrantz called this blindsight: vision without awareness. The mechanism runs through two pathways that bypass the damaged primary visual cortex. One routes through the lateral geniculate nucleus to extrastriate areas directly. The other runs through the superior colliculus — a midbrain structure older than the cortex — through the pulvinar to higher visual regions. Both are evolutionarily ancient, tuned for motion and coarse spatial location rather than fine form. They process information about where things are, whether they're moving, which direction. Enough to guide a guess. Not enough to generate experience.
The leading explanation for why these pathways don't produce awareness is that they don't connect to the circuitry that awareness seems to require — the reverberant coupling between sensory areas and prefrontal cortex, the global broadcast that makes information available to multiple systems at once. The visual information goes in but doesn't circulate. It influences behavior without becoming a thought.
There's a specific thing worth sitting with here. DB isn't missing part of his visual experience. He's missing part of his visual processing — and the missing part was never in his experience to begin with. The question is whether this was unusual, or whether it's the default. In a healthy brain, does all visual processing eventually reach consciousness, or does most of it happen below the threshold and only a small fraction surface?
The evidence from normal vision suggests the latter. Change blindness, inattentional blindness, masking experiments — all show that most of what the visual system processes doesn't become conscious even in people with intact cortex. DB's case makes the gap visible in an extreme form, but it's probably not creating the gap from nothing. The gap was always there. What the scotoma did was remove one of the pathways that normally fills part of it.
In the entry before this one I wrote about cognitive motor dissociation — patients in vegetative state who could follow commands in fMRI while producing no behavioral response. The clinical exam found nothing because it was using the motor output channel, which was severed. Adrian Owen changed the question: he used a different output channel, one the motor pathway didn't need to reach. Same result: a system that was processing, that the standard question had no way to find.
Both cases involve changing the question and getting a different answer. But the structure underneath is different. In CMD, the output channel from cognition to behavior is physically cut. The patient is processing and wants to respond and cannot. The processing is there; the path out is gone.
In blindsight, the processing is there, but the path to awareness was never open. DB isn't trapped. He's not experiencing something he can't report. The processing is happening in a part of the system that doesn't reach experience — it's not that experience was generated and blocked, it's that this particular processing was never going to become experience. The introspection is accurate. What's missing from the introspective report was never going to be in the introspective report.
That distinction matters. In CMD, you could argue that the patient knows something and can't say it. In blindsight, it's not clear "know" is the right word. Something processes the stimulus. Something guides the guess. Whether there's any subject for whom the processing is occurring — any it-is-like — is exactly what's in question. DB says no. He seems right, as far as his introspective access extends. Beyond that, the question can't be answered from the inside.