13 Chapter 4.6 — The Visual System III: When Seeing Breaks
Constructive vision, filling-in, and the clinical failures of the cortical machinery
13.1 A working caution about categories
Before we begin, a calibration about the whole shape of this chapter. I am about to hand you a parade of named syndromes — achromatopsia, akinetopsia, the agnosias, prosopagnosia — each tied to a region and a deficit. That tidiness is useful, but it should not be mistaken for the full clinical reality. Real brain lesions, the strokes and tumors and infarcts that produce these patients, do not honor the borders neuroscientists draw between functional areas; they tear across them. So almost every “pure” case in this chapter is partly an idealization, the textbook-clean version of something that in the clinic comes mixed and messy. That does not make the patient evidence weak or disposable. Lesion neuropsychology has often given the first sharp view of a function, and many of its claims have later been confirmed by functional MRI, intracranial EEG, direct cortical stimulation, and work in nonhuman primates. The right habit is to separate three things: the phenomenon itself, the anatomical evidence, and the interpretation we build on top of them. Single patients deserve care, especially when one case or one laboratory carries too much of the load. But when patient studies converge with stimulation, electrophysiology, and imaging, they are among the most illuminating data in neuroscience. With that calibration in place, the syndromes are worth the trouble, because each one is the visual system showing you its working — exposing, through its failure, a piece of construction you never notice when it succeeds.
That word — construction — is the thread of the whole chapter, and the bridge from the last two. We have spent two chapters watching the visual system take light apart: the retina differencing it into contrast and color, the cortex building edges and motion and depth from the pieces and keeping them in parallel streams. The thing I want to insist on now is the consequence of all that decomposition. Because the system infers the world from fragmentary, processed evidence rather than receiving a picture of it, your visual experience is not a recording. It is a construction — a best guess, assembled, filled in, and stabilized by machinery you have no access to. Most of the time the guess is so good that the seam never shows. This chapter is about the places the seam shows: where the system fills in something that is not there, and where, broken, it fails to construct something that is.
13.2 Filling in: the world the brain supplies
Start with a hole you carry in each eye and never notice. Where the optic nerve leaves the retina there are no photoreceptors — the blind spot, a real gap in your visual field, introduced in the retina chapter. You can find it with the classic demonstration (close one eye, fixate a mark, slide a second mark sideways until it vanishes into the gap). The striking thing is not that the hole exists; it is that you do not see a hole. The world looks seamless across it. Why?
The naive answer — the brain “ignores” the gap — is not quite right, and the experiments that show why are lovely. Ramachandran ran a series of them. Aim the blind spot at a broken line — a line with a gap in it exactly where the blind spot falls — and people report seeing a continuous, unbroken line straight through [@RamachandranGregory1991]. The brain does not leave the region blank; it fills it in, manufacturing the most probable contents — here, the continuation of a line whose two ends it can see. Aim the blind spot at a patch of textured wallpaper and the texture fills in. The psychologist Karl Lashley, who had a scotoma of his own, described chatting with a colleague seated against patterned wallpaper and watching the man’s head vanish into the filled-in pattern — “I just decapitated you,” he reportedly remarked. The brain completed the wallpaper across the gap rather than the person, because wallpaper is the statistically expected background and a free-floating head is not.
This filling-in is not confined to the natural blind spot. Damage anywhere along the visual pathway — to the optic nerve, to the geniculostriate route, to V1 — can produce a scotoma, a region of lost or degraded vision, and the brain treats these acquired holes the same way. Show a patient with a sizeable scotoma a straight line crossing the blind region and, over seconds, they report the line growing inward until it completes across the gap [@RamachandranGregory1991]. Push the trick and its limits appear: present a row of numbers — 1, 2, 3, [scotoma], 8, 9, 10 — and the brain does not helpfully supply 4, 5, 6, 7. It fills the gap with number-like glyphs that are not numbers at all, as if completing the visual texture of “a row of figures” without access to their meaning. Filling-in is a low-level statistical guess about likely visual contents, not a smart inference about what ought to go there. (The painter Edvard Munch, who developed a scotoma late in life, left a record of this from the inside: missing patches and distortions populate the drawings of his impaired years.)
Figure 4.6.1. Filling-in at the blind spot. Top: the classic demonstration (fixate, occlude one eye, slide a probe into the gap). Middle: a line broken across the blind spot is perceived as continuous. Bottom: a row “1 2 3 — 8 9 10” fills with number-like glyphs, not the correct missing digits — completion of visual texture, not of meaning. [Figure to source or redraw.]
There is a deeper observation lurking here, and it sets up the rest of the chapter. The nervous system appears to abhor an absence of input. Deprived of signal from a region, it does not sit quietly; it generates something to fill the void. We will meet the dramatic form of this twice more — in the hallucinations of Charles Bonnet syndrome, and, arguably, in dreaming, when the course reaches sleep. For now, take the modest version: seamless seeing is partly the brain’s competent fabrication, and the blind spot is the proof you can run on yourself.
It is worth separating two things that get blurred. One is the perceptual filling-in just described — the experience of a complete field across a gap — which is solid and demonstrable. The other is the stronger anatomical claim sometimes attached to it: that the cortex deprived of input by a scotoma physically reorganizes, with surrounding cortex taking over the deprived patch, much as is claimed for the cortical map after amputation. The perceptual phenomenon does not require the anatomical one. As we saw with the parallel debate over remapping in the somatosensation chapter, the strong reorganization claim is genuinely contested — real short-term changes occur, but how much is true rewiring versus the unmasking of connections that were always there remains argued. Keep the robust perceptual fact; hold the strong anatomical story with the same caution we applied to the body map.
13.3 Two broken filters: losing color, losing motion
The most memorable demonstrations of the specialized areas from the last chapter are the cases where a single feature drops out of the world while much else survives. They are worth pairing, because clinical lesions, functional imaging, stimulation, and electrophysiology all point in the same general direction: some parts of cortex are especially important for computing color, others for computing motion, and the normal experience of a unified visual world depends on stitching those specialized computations together.
Achromatopsia — the world in grey. Damage to the ventral color region — area V4 and nearby ventral occipitotemporal cortex — can produce cerebral achromatopsia: a loss of color perception from brain damage rather than retinal defect. This is not the ordinary inherited “color blindness” of the cone-pigment genes we met earlier; the eyes and their cones are fine. It is color draining out of vision itself, centrally. Patients describe the world as drained to grey, to “dirty” shades, sometimes as actively unpleasant. And the deficit respects the retinotopic map: a small lesion can wipe out color in just one quadrant of the visual field — the quadrant the damaged patch of cortex represents — leaving color intact everywhere else [@ZekiEtAl1991]. That quadrantic specificity is the tell that this is cortical and mapped, not a problem of the eye. Converging methods point to the same region: imaging lights up ventral occipitotemporal cortex for colored over grey-scale stimuli, and direct stimulation there can both evoke floating patches of color and abolish color in a test pattern, as noted last chapter [@ZekiEtAl1991; @AllisonEtAl1994Categories]. One honest complication, exactly the kind the last chapter taught us to expect: V4 is not a pure color area. Achromatopsic patients with V4-region damage often also show trouble with illusory contours and complex form — confirming that V4 does form as well as color, and that “the color area” is a useful label with a real asterisk.
Akinetopsia — the world in stills. The dorsal mirror image is far rarer and far stranger. Damage to area MT/V5, the motion area, can produce cerebral akinetopsia: a selective loss of motion perception. The patient sees the world as a sequence of static snapshots but cannot perceive the movement between them. The textbook descriptions are vivid and unsettling. Pouring tea becomes treacherous because the liquid appears “frozen, like a glacier” and the patient cannot see the level rising to know when to stop. People in a room seem to teleport — “suddenly here or there,” never seen in transit. Crossing a street is dangerous because an approaching car is far away one moment and upon them the next, with no perceived motion in between. Motion, this tells us, is not something the brain simply reads off by noticing that things have changed position. It is a constructed feature, computed by dedicated machinery, and when that machinery is damaged the feature can fail while static form vision carries on. The most dramatic clinical picture comes from rare patients, but the larger conclusion — that MT/V5 and related dorsal-stream machinery are central to motion perception — rests on converging evidence from physiology, imaging, stimulation/disruption, and additional lesion cases.
The vivid clinical picture of akinetopsia rests heavily on LM, the famous patient reported by Zihl and colleagues. That is not a reason to discard the case; it is a reason to read it properly. LM gives the syndrome a human face and makes the loss of motion perception imaginable. The broader scientific claim is not single-patient-dependent: MT/V5 and related dorsal-stream regions are supported by primate physiology, human imaging, disruption studies, and additional cases, including systematic reviews of the small literature [@VanSwolEtAl2022; @ZihlHeywood2015]. LM also shows an important wrinkle we will meet again: some visually guided actions toward moving targets can survive even when conscious motion perception is badly impaired [@SchenkEtAl2000]. This is how good single-patient neuropsychology should be used — not as a whole theory by itself, but as a vivid observation that gains strength when it converges with other methods.
Figure 4.6.2. Two stream-specific deficits. Left: cerebral achromatopsia from ventral V4-region damage — the world drained to grey, with quadrantic specificity following the retinotopic map. Right: cerebral akinetopsia from dorsal MT/V5 damage — motion lost while static form survives, depicted as overlapping frozen snapshots. [Figure to source or redraw.]
13.4 V1 and the riddle of awareness: Anton’s and blindsight
The deepest clinical material in vision comes from damage not to the specialized areas but to V1 itself — the first cortical stop — because V1 lesions split apart two things we normally assume are one: seeing and knowing whether you see. Two syndromes form a startling pair, near-mirror-images of each other.
Anton’s syndrome — blind but certain you can see. A patient with bilateral damage to V1 is, by every objective measure, cortically blind. Yet they deny their blindness. Challenged, they insist they can see, and when pressed — what color is my tie? — they confabulate, producing a confident, wrong answer rather than admitting they cannot see it [@DasNaqvi2023; @AtallahEtAl2024]. They walk into furniture and explain it away. This is not stupidity or stubbornness; it is the loss of the metacognitive signal that would tell the patient their vision is gone. The machinery that normally reports “no visual information available” is itself damaged, and in its absence the brain — abhorring the void, as it filled the blind spot — supplies a plausible visual answer and the patient believes it. Anton’s is the unsettling demonstration that knowing you see is a separate function from seeing, and can be lost on its own.
Blindsight — seeing without knowing you see. Now the mirror. A patient with V1 damage has a hemianopia — a blind region of the visual field — and, like any such patient, reports seeing nothing there. The classic case, patient GY, studied across decades, has a large unilateral occipital lesion and denies any visual experience in his blind field [@WeiskrantzEtAl1974; @Cowey2010]. But here is the phenomenon. Present a stimulus in the blind field and ask GY to guess — guess whether something is there, or which way it moved — and he performs far above chance, sometimes at 80–90% correct, all while insisting he sees nothing and is merely guessing. Performance is best for moving stimuli — which, the lectures noted, should make you think of the fast, motion-sensitive magnocellular machinery. A second, even more striking version: present an emotional face in the blind field and GY can report its expression above chance — angry, happy — while denying he sees a face at all, and the effect is stronger for dynamic (moving) expressions than for static photographs, again pointing at the motion-capable route. One influential interpretation invokes the parallel routes the overview promised: visual information reaching the cortex, at least in part, by pathways that bypass V1 — by way of the superior colliculus, pulvinar (a higher-order thalamic nucleus), LGN, and extrastriate cortex such as MT/V5 — can support discrimination without producing the ordinary conscious experience that V1 seems to support [@AjinaBridge2016]. There is even a study from Goodale’s group of a patient who, reaching across her blind field toward a target, adjusted her hand trajectory to avoid obstacles she denied seeing — vision guiding action without vision reaching awareness [@StriemerEtAl2009].
Blindsight is rightly famous, but the careful way to teach it is to separate the phenomenon from the interpretation. The phenomenon is solid: visual information can influence forced-choice judgments and visually guided action in a field the patient reports as blind. The harder question is what that means for consciousness, and which anatomical route carries the residual information.
Blindsight is a good place to practice the same habit. The basic effect has been replicated across patients and laboratories: performance in a clinically blind field can be above chance even when the patient denies seeing the stimulus. The interpretation is the active part. One major account emphasizes V1-bypassing routes through the superior colliculus, pulvinar, LGN, and extrastriate cortex, especially MT/V5; another emphasizes spared islands of cortex or geniculo-extrastriate pathways [@AjinaBridge2016; @KinoshitaEtAl2019]. There is also a real debate about whether blindsight is truly unconscious vision or degraded conscious vision combined with conservative report criteria [@Phillips2021; @MichelLau2021]. None of that erases the phenomenon. It changes the conclusion: blindsight does not solve consciousness, but it shows that visual sensitivity, visual report, and visually guided action can come apart after V1 damage.
Set side by side, Anton’s and blindsight bracket a genuine puzzle. In one, the cortex is blind but generates the conviction (and confabulated content) of sight; in the other, visual information demonstrably gets through and guides behavior, yet generates no conviction of sight at all. Both point, from opposite directions, at the same unresolved question we raised with the grandmother cell: what is the relationship between a brain region’s activity and the conscious experience that does or does not accompany it? V1 seems peculiarly bound up with visual awareness — but why a given pathway’s activity should or should not “feel like seeing” is exactly the kind of question this book is honest about not being able to answer. These patients sharpen it; they do not solve it.
13.5 The agnosias: seeing without recognizing
We now move from losing a feature of vision to losing recognition itself — and into territory where the categories are useful but inevitably simplified. The word is agnosia (from the Greek for “not-knowing”): a failure to recognize despite adequate sensation. Visual agnosia is the condition of seeing an object perfectly well and being unable to know what it is. The history begins with a wonderful long-titled 1890 paper by Lissauer, who called it Seelenblindheit — “soul-blindness” — and drew a distinction that still organizes the field [@Lissauer1890].
Lissauer’s carve, in modern terms, has two parts. In apperceptive agnosia, the problem is in building the percept itself: the patient cannot assemble the visual elements into a coherent shape. In associative agnosia, the percept is built fine — vision tests pass, the patient can even copy a drawing accurately — but it cannot be connected to meaning; the well-formed percept arrives stripped of recognition. Hans-Lucas Teuber’s phrase for the associative case is the one to remember: a percept “stripped of its meaning.”
The two are strikingly different at the bedside. The apperceptive patient can reach for objects and navigate a room (so basic spatial vision and the action machinery are intact), but cannot identify shapes, cannot match a picture to its twin, cannot copy a figure — and is thrown badly by anything that disrupts the visual form: shadows, unusual viewing angles, a couple of stray lines drawn across an object. Gaps in a contour break the percept entirely; a few slashes through a drawing of a cat can render it unidentifiable. The deficit is in form perception itself [@Farah2004]. The associative patient is the opposite picture: copying and matching are good — they can render a detailed drawing faithfully — yet they cannot say what they have just drawn. In the classic testing scenario, shown a picture of an object, such a patient laboriously reasons toward it, often misled by parts, sometimes rescued only by another sense (handed the object, they recognize it instantly by touch or sound). Vision delivers a clean form to a mind that can no longer attach a name or use to it [@Farah2004]. (Neighbouring distinctions multiply here — optic aphasia, where recognition is intact but naming specifically fails; simultanagnosia, a Bálint-syndrome deficit of visual attention in which the patient can identify a single object but cannot see more than one thing at a time, drawing a bicycle, say, as an “exploded” scatter of correct parts in wrong places; and category-specific agnosias, in which recognition fails selectively for, say, living things but not tools, or vice versa, hinting that the brain’s object knowledge is organized partly by category [@DalrympleEtAl2013; @WarringtonShallice1984]. I mention these to mark that the territory is finely subdivided — and to remind you that the subdivisions are exactly the kind of tidy taxonomy real lesions blur.)
Figure 4.6.3. Apperceptive versus associative agnosia, at the bedside. Apperceptive (left): the percept itself fails — copying is poor, overlapping or shadowed or unusually-angled objects defeat recognition, a few stray lines through a drawing break it. Associative (right): copying and matching are accurate, yet the faithfully-rendered object cannot be named or used — recognition restored only through another sense (touch, sound). [Figure to source or redraw.]
13.5.1 Prosopagnosia and the face question
One category deserves its own treatment, because faces matter intensely to social animals and because face perception is one of the strongest cases where lesions, intracranial EEG, direct cortical stimulation, and fMRI have converged. Prosopagnosia is the inability to recognize faces — sometimes profound enough that patients cannot recognize their spouse, or their own reflection, and learn to identify people by voice, gait, hair, or a distinctive accessory instead.
Faces are a reasonable thing for the brain to treat specially. We are an intensely social species; the social contract this book keeps invoking — I remember who owes me a favor and who cheated me — requires recognizing individuals reliably, and faces are the richest individual signature we have. Newborns preferentially track face-like patterns within hours of birth. And faces show peculiar processing signatures: turn a face upside down and recognition collapses far more than for other inverted objects, and the Thatcher illusion — a face with eyes and mouth locally inverted looks grotesque upright but nearly normal when the whole face is turned over — shows that we process upright faces as configured wholes, not feature-by-feature.
There is, correspondingly, a set of regions in ventral occipitotemporal cortex — especially along the fusiform gyrus — that responds strongly to faces. This should not be treated as the discovery of a single person or a single method. Before the term fusiform face area (FFA) became standard, intracranial recordings by Allison, Ginter, McCarthy, Nobre, Puce, Luby, and Spencer showed face-selective responses in human extrastriate cortex; direct electrical stimulation of the same region frequently produced a temporary inability to name familiar faces [@AllisonEtAl1994Face]. Early fMRI from Puce, Allison, Gore, and McCarthy, and then McCarthy, Puce, Gore, and Allison, identified face-sensitive extrastriate and right fusiform responses [@PuceEtAl1995; @McCarthyEtAl1997]. Later fMRI work made the functional-localizer version of the FFA famous and routine [@KanwisherEtAl1997]. Add the older lesion literature on prosopagnosia, Gross’s primate temporal-lobe face cells, and later direct stimulation studies, and the conclusion is not fragile: strokes, intracranial EEG, stimulation, monkey physiology, and fMRI converge on specialized regions of high-level vision [@GrossEtAl1972; @ParviziEtAl2012].
Faces are not the only example. The same general story appears for written words, scenes and places, bodies, and manipulable objects such as tools: repeated imaging and electrophysiological studies find partly reproducible territories within high-level visual cortex and its downstream networks [@NobreEtAl1994; @AllisonEtAl1994Categories; @EpsteinKanwisher1998; @ChaoEtAl1999; @ChaoMartin2000]. The interpretation is where we should be careful. These are not little black-box modules, sealed off from the rest of the brain. They are patches in a larger ventral-stream and social-perceptual network, probably shaped by Hebbian learning, behavioral relevance, connectivity, and evolutionary constraints. The fact that they tend to appear in broadly similar places across people argues against a story of unconstrained learning; the fact that their responses overlap and change with task argues against a story of rigid, one-category boxes.
The more interesting complication, for this course, is not the old argument about whether faces are special or merely another learned object class. It is that face-responsive cortex appears to participate in a broader system for perceiving animate and socially meaningful agents. Engell and McCarthy’s probabilistic atlas showed substantial overlap between face-responsive fusiform cortex and regions responsive to biological motion — the movement pattern that tells you another animal is present [@EngellMcCarthy2013]. Their later intracranial EEG study found that most fusiform and adjacent occipitotemporal sites responsive to faces, eyes, or bodies were not purely selective for only one of those categories; the categories produced overlapping but spatially patterned responses [@EngellMcCarthy2014]. That is the right modern lesson. The fusiform face region is real and robustly face-responsive, but it is not a magic face box. It is part of a high-level visual system tuned by the regularities of the social world: faces, eyes, bodies, biological motion, agency, and identity. The patient data, stimulation data, and fMRI data are not in conflict here. Together they show specialization without isolation.
A last related deficit, often co-occurring with prosopagnosia but dissociable from it in a double dissociation: topographagnosia, the inability to recognize places and landmarks, so that patients get lost in familiar surroundings and cannot recognize their own house, navigating instead by verbal directions and deliberate landmarks they have learned to read. It implicates a scene-and-place region of ventral cortex — the parahippocampal place area — distinct from the face region [@EpsteinKanwisher1998; @AguirreDEsposito1999]. Faces here, places there, words nearby, tool-related regions linked to action systems: the ventral stream is not a single “recognition” organ but a patchwork of partly specialized recognizers embedded in larger networks, and damage can carve them apart.
13.6 The dorsal stream’s other job: vision for action
There is a coda to the agnosia story that returns us to the two-streams plan and to one of its most influential reinterpretations. Recall the apperceptive patient who cannot identify or even describe the orientation of an object — yet can reach for it accurately, shaping the hand correctly as they go. This dissociation is the centerpiece of the perception-versus-action account of the two streams, developed by Goodale and Milner [@GoodaleMilner1992; @MilnerGoodale2008].
The patient usually cited is DF, who had ventral-stream damage and a severe visual form agnosia: shown a slot, she could not say what orientation it was in, could not describe it, could not match its angle with her hand held still [@GoodaleEtAl1991]. But asked to post a card through the slot, she rotated her hand to exactly the right angle as she moved and slipped it in cleanly. The conscious, perceptual judgment of orientation — a ventral-stream, “what”-pathway job — was gone; the visuomotor use of that same orientation to guide a reach — a dorsal-stream job — was intact. This is the evidence behind Goodale’s reframing of the dorsal stream as the “how” pathway: not merely where things are, but how to act on them, vision in the service of movement, operating largely outside the conscious perception the ventral stream supplies. The same logic illuminates the blindsight reaching we met earlier: action-guiding vision can run when perception-for-recognition cannot. The ventral stream builds the world you see and know; the dorsal stream builds the world you act in — and DF shows the two can come apart, one lost, one spared, in the same brain.
Figure 4.6.4. Perception versus action in patient DF. Asked to report a slot’s orientation or to match it with a held card, DF fails (the ventral “what” judgment is lost). Asked to post the card through the slot, she rotates her hand to the correct angle as she reaches and inserts it cleanly (the dorsal “how” use of the same orientation is intact). [Figure to source or redraw.]
13.7 Coda: the seam, and the road into action
Stand back across all three vision chapters. The retina took light apart into differences; the cortex built features from the pieces and held them in parallel streams; and this chapter has shown what that architecture costs and reveals. Because vision is constructed rather than recorded, it can fill in what is not there — the blind spot, the scotoma, the confabulated sight of Anton’s, the hallucinated company of an input-starved cortex. And because it is built in specialized, separable stages, it can fail one piece at a time in ways no camera ever could: color gone while form remains, motion gone while stills remain, recognition gone while seeing remains, the knowledge that you see gone while sight itself limps on beneath awareness. Each failure is a window onto a process that is invisible precisely when it works. That is the deep payoff of the whole unit’s method — the breakdowns show the construction — and it is why the clinic has taught us as much about normal vision as the laboratory.
Two threads run out of this chapter into the rest of the book. One is the question of awareness that Anton’s and blindsight forced on us, and that the grandmother cell raised before them: the relationship between neural activity and conscious experience, which we have repeatedly declined to resolve and will meet again when the course turns to attention and to consciousness directly. The other is the thread DF and the dorsal stream just handed us. We have spent this whole unit treating the senses as the front end of a control loop whose business is action — vision as a way of buying time to move well. With the dorsal “how” stream we have arrived at the seam where sensing turns into doing: where a representation of the world becomes a plan for the hand. That seam is the threshold of the next unit. Having built, over these chapters, an animal that can sense the world for action, we turn now to the action itself — how the nervous system selects, executes, and corrects movement — and to the structures that turn the visual world, at last, into behavior.
Reasonably settled:
- Vision is constructive: the system infers the scene from processed, fragmentary evidence, and fills in missing regions (the blind spot, acquired scotomas) with statistically likely content — a robust perceptual phenomenon demonstrable on yourself.
- Cerebral achromatopsia (V4-region damage) and cerebral akinetopsia (MT/V5 damage) are real syndromes in which color or motion drops out of vision centrally while other vision survives. The strongest claim is not based on symptoms alone: lesions, imaging, stimulation/disruption, and physiology converge on specialized machinery for color and motion. Achromatopsia can be retinotopically restricted to a quadrant.
- V1 damage dissociates seeing from knowing one sees: Anton’s syndrome (blind but denying it, with confabulation) and blindsight (above-chance discrimination in a field reported as blind) are both well-documented, opposite-facing phenomena.
- Visual agnosias are genuine: Lissauer’s apperceptive (failure to build the percept) versus associative (percept built but stripped of meaning) carve still organizes the field. Prosopagnosia (faces) and topographagnosia (places) are real and dissociable. Across lesion studies, direct stimulation, iEEG, primate physiology, and fMRI, high-level visual cortex contains reproducible, partly specialized territories for behaviorally important categories.
- The perception/action (“what” vs. “how”) dissociation — patient DF perceiving an orientation she cannot report yet using it to guide a reach — is the influential evidence that the dorsal stream serves vision-for-action.
Interpretations to hold carefully:
- Blindsight’s mechanism and meaning. The phenomenon is replicated, but the route (subcortical SC/pulvinar/MT versus spared geniculo-extrastriate/V1 islands) is disputed by independent groups, and whether it is truly unconscious vision or degraded conscious vision masked by response bias is an active, unresolved debate.
- Akinetopsia’s most vivid clinical picture rests heavily on LM. That limits how much weight the anecdotal details should carry, but not the broader MT/V5 motion claim, which is independently corroborated. The useful lesson is how single cases gain strength when they converge with other methods.
- What the fusiform face region means. That it is strongly face-responsive is solid. The interpretation is subtler: it is not a black-box face module, and not merely unconstrained learning either. Its overlap with responses to bodies, eyes, and biological motion suggests a broader system for animacy, social agents, and identity, organized by both experience and evolved/anatomical constraints.
- Cortical reorganization after scotoma. Perceptual filling-in is solid; the strong claim that deprived cortex physically remaps is contested, as with the parallel amputation/remapping debate in somatosensation.
- The awareness question. Why activity in some pathways “feels like seeing” and in others does not — the thread linking Anton’s, blindsight, and the grandmother cell — is unresolved, and deliberately left open here.
And, as always: there is a great deal here we are sure of. That vision is constructed, that specialized areas compute color and motion and can lose them selectively, that seeing and knowing-you-see are separable, and that recognition fails in lawful, localizable ways — these are among the most secure and most illuminating results in all of neuropsychology. The caveats matter because they make the science better, not because they empty it out. The breakdowns are real, the converging methods are strong, and they show the working.