Blindsight: Seeing Without Knowing You See

Type 1 blindsight is defined by a flat contradiction between what a patient reports and what a patient does: no awareness of a visual stimulus in the blind field, and above-chance performance on forced-choice tasks about that same stimulus. The founding case is patient DB, described by Weiskrantz and colleagues in 1974. His right striate cortex had been surgically removed to treat an arteriovenous malformation, leaving him blind in the left visual field. In the laboratory he could locate visual targets in that field by pointing, discriminate between an X and an O, and detect motion, while consistently reporting that he could not see anything and was, as he put it, just guessing.

The performance is real, and it is not vision restored. Accuracy in blindsight runs from roughly 60 to 90 percent depending on the task, against near-100 percent in the same patient's sighted field. Whatever survives the loss of the primary visual cortex is not normal seeing at reduced volume. It is a narrower, cruder channel, and every claim below has to be read against that ceiling.

Patient GY is the most extensively studied blindsight subject in the literature. He sustained damage to his left primary visual cortex in childhood, and shows above-chance performance detecting, localizing, and discriminating stimuli in his right visual field. His performance improves with high contrast, fast motion, and large stimuli, which is itself a useful description of the channel: it answers to things that are strong, moving and obvious, and not to much else.

There is no established epidemiological count of blindsight patients anywhere in the world. The great majority of the foundational literature, including the pathway anatomy, the Type 1 and Type 2 distinction, and the amygdala and motion findings set out in the sections below, rests on intensive, decades-long study of a small number of named individuals. Above all two: DB, studied from the mid-1970s, and GY, whose damage dates from childhood and who has been studied for over three decades. When a sentence in this article begins blindsight patients, read it as a finding from a handful of people, because that is what it is.

The evidence is not confined to those two names, though the cohorts stay small. Sahraie and colleagues, with Weiskrantz himself among the nine authors, tested 10 cortically blind patients in 2003 and found a narrowly tuned residual spatial-frequency channel in 8 of the 10. That is real evidence that residual visual function in a blind field is not a peculiarity of the two most famous cases. It is also still 10 patients. The honest phrasing is documented in a range of patients beyond the two canonical cases, in cohorts numbering in the low tens, and it is neither common nor a population in the epidemiological sense.

The split follows a known division of labor in the visual system. The dorsal stream, the where and how pathway, is more accessible in blindsight than the ventral stream, the what pathway. Patients are consistently better at detecting where something is and whether it is moving than at identifying what shape or color it was. Put crudely, the blind field can tell you that something happened over there without being able to tell you what it was.

Four capacities are empirically demonstrated across multiple patients and studies. Localization: pointing to the position of a visual target (Weiskrantz, 1986). Motion, which is the strongest of them: detecting and discriminating the direction of moving stimuli, suggesting involvement of area V5/MT (Azzopardi and Cowey, 1998). Orientation: above-chance discrimination of grating orientation (Stoerig and Cowey, 1997). And wavelength: some capacity for color discrimination (Stoerig and Cowey, 1992), which is the item on the list that sits least comfortably beside the dorsal-stream picture just described.

The fifth capacity is the one that tends to stop people. Facial expressions of emotion, fear and happiness, presented in the blind field can influence behavior and produce measurable responses in the amygdala, even though the patient reports no awareness of the face at all (de Gelder and colleagues, 1999; Morris and colleagues, 2001). Something in the head registered the expression on a face the person will tell you, honestly, that they never saw.

And here is the boundary of the list. The claim that blindsight patients can see normally is false. Performance is far below normal vision, above chance but typically well below ceiling, limited to simple discriminations, and it does not support complex visual cognition. There is no reading in the blind field and no face recognition in it, and that last one is worth separating from the amygdala finding above: an emotional expression presented in a blind field can influence behavior and produce a measurable amygdala response, which is a reaction to the expression and not recognition of whose face it was. Any popular account describing blindsight as intact vision routed around awareness is describing something that has never been observed in anyone.

Blindsight is mediated by visual pathways that bypass V1. The principal one is the retino-tectal pathway: retina to the superior colliculus in the midbrain, on to the pulvinar in the thalamus, and from there to extrastriate visual areas including V5/MT, V4 and parietal cortex. None of that traffic passes through the primary visual cortex, which is precisely why destroying the primary visual cortex does not silence it.

A second bypass sits at the thalamus itself. The lateral geniculate nucleus, the relay that normally feeds V1, can project directly to extrastriate visual areas without routing through V1 at all. So there are at least two anatomical ways for visual information to reach surviving cortex in a brain whose main visual entry point is gone. Nothing about blindsight requires the signal to arrive by magic; it requires only that the wiring diagram have more than one entrance, and it does.

One further line of evidence belongs here, and it belongs in its own paragraph because it is a different class of evidence entirely. Blindsight-like phenomena have been documented in monkeys with surgical V1 lesions: ablated animals show residual detection, localization, and some form discrimination that parallels the human cases (Cowey and Stoerig, 1995). That matters exactly because it is not a case study. In the animal work the location and extent of the lesion are experimentally controlled and known precisely, whereas the human cases arise from accident, stroke, or surgery for unrelated reasons and are studied after the fact. This is not the same finding as DB's or GY's and should never be quoted as if it were. It is an independent finding that happens to agree with them, which is worth more than a louder version of the same one.

Type 2 blindsight is the case in which the patient does report something. Not normal seeing, but a feeling that something is there (Weiskrantz, 1998). GY describes sensations in his blind field as a feeling of something happening, distinct both from ordinary vision and from a complete absence of experience. This has been described as a form of degraded phenomenal experience rather than true unconscious processing. The distinction is not a technicality. Type 1, no awareness at all, is the clearest case of visual processing running with nothing conscious attached to it, which makes Type 1 the only version that does the job theories of consciousness want blindsight to do.

The classic and most-cited skeptical case is Campion, Latto and Smith, Is Blindsight an Effect of Scattered Light, Spared Cortex, and Near-Threshold Vision?, in Behavioral and Brain Sciences, volume 6, 1983, pages 423 to 448. Their argument is that the reported blindsight performance of the early studies could be explained without positing any unconscious vision at all, by three ordinary mechanisms: light scattering from a bright stimulus into the intact visual field; incompletely destroyed striate cortex leaving small functioning islands behind; and near-threshold conscious vision that patients simply under-reported. Their own conclusion was that an adequate case for blindsight had not been made by the evidence then available. This is a peer-reviewed, formally target-articled position, not a fringe dismissal, and it deserves to be met rather than waved at.

There is a specific empirical version of the spared-cortex limb. Fendrich, Wessinger and Gazzaniga, writing in Science in 1992, used careful eye-tracking to confirm retinal stability and located a small, discrete island of residual visual sensitivity inside a blindsight patient's scotoma, separated from the region of macular sparing. Their argument was that spared tissue of this kind, rather than a genuine unconscious pathway, could account for at least some reported blindsight in some patients. Note the scope of that claim, because it is often overstated in both directions: it is a finding in one patient, and it is a claim about some cases, not a refutation of all of them.

The substantial reply exists and it is worth naming properly. Alan Cowey's 30th Sir Frederick Bartlett Lecture, published in the Quarterly Journal of Experimental Psychology in 2004 under a title that announces its intent, Fact, Artefact, and Myth About Blindsight, addresses the scattered-light and spared-cortex objections directly. It is standardly summarized as concluding that the core phenomenon survives them once stimulus placement is controlled, eye position is tracked, and lesion boundaries are verified with modern neuroimaging rather than assumed from a surgeon's notes. The title alone signals that Cowey treated the skeptical case as something to answer rather than something to dismiss, which is the correct posture and not the one our own file adopted.

The strongest single piece of patient-specific evidence against the spared-islands limb comes from Zeki and Ffytche, in Brain in 1998. Using functional imaging on GY, they found activation in extrastriate visual areas, notably V5/MT, without detectable activation in the spared portions of his primary visual cortex for the motion stimuli tested. That is evidence for a genuine extrastriate route rather than reliance on a hidden intact patch of V1. Secondary syntheses of this and related work report that the majority of blindsight subjects studied to date do not appear to have spared V1 islands large enough to explain their performance, and this article carries that statement as the well-corroborated secondary claim it is rather than as a first-hand finding.

The objection has not been dissolved. It has been answered, and those are different outcomes, so treat any account of blindsight that presents the phenomenon as uncontested, including our own file's, as having skipped the part that matters. Our file answers the objection in a single sentence, saying that early criticisms about light scattering have been controlled for in numerous studies and that blindsight is a genuine phenomenon. That sentence is directionally correct and it is not an argument: it names no study, no critic and no control. It sits beside a counter-arguments section claiming that no active scholarly dispute exists here at all, which is simply not true of this subject. What fails here is not the phenomenon. It is the framing that closes the question.

The standard philosophical reading is that blindsight shows visual information can be accessed, used to guide behavior, without being phenomenally experienced. That maps onto Ned Block's 1995 distinction between phenomenal consciousness, the felt character of an experience, and access consciousness, the availability of information to the rest of the system. The reading is standard, and the exact mapping between the phenomenon and the philosophical distinction is itself debated.

Recurrent processing theory, in Victor Lamme's 2006 statement of it, holds that conscious visual perception requires recurrent, feedback processing involving V1, and that feedforward processing alone, which is what survives V1 damage, produces only unconscious processing. That is consistent with blindsight. Consistent with is not the same as demonstrated by, and the distance between those two phrases is where most of this section lives.

The global workspace reading, as our own file glosses it, is that V1 damage may prevent visual information from entering the global workspace: the information is processed locally but is never broadcast widely enough across cortical networks for conscious access. That is a gloss on blindsight rather than a developed argument, and the theory's own dossier belongs to its own file, not this one.

And here is the complication that keeps the verdict honest. It is tempting to read blindsight as proof that V1 is irrelevant to consciousness. Our companion file on the neural correlates of consciousness does not license that reading. Crick and Koch originally argued that V1 is not a neural correlate of consciousness, on the grounds that its activity does not track conscious perception in binocular rivalry. But that same file notes that some evidence, super-resolution perception and blindsight recovery among it, suggests V1 may play a role after all, and that which cortical areas are in or out of the neural correlates of consciousness remains actively debated. Blindsight is a data point inside that argument. It is not the end of it.