Follow one photon,
tear film to cortex.

The first lens in the visual system is not the cornea. It is a film of tears a few micrometres thick, and it is the only optical surface in the body that is torn down and rebuilt every few seconds. Where that film breaks up between blinks, the surface it leaves is no longer smooth, and vision blurs until the next blink restores it. This is why dry eye is a visual complaint and not only a comfort one, and why the blur it produces comes and goes instead of sitting still the way a refractive error does.

Beneath it, the cornea does most of the work: roughly 43 of the eye's 60 or so dioptres of refractive power, and not one of them adjustable. It has five layers, no blood vessels at all, and it takes its oxygen directly from the air through the tear film. Transparency is bought with that avascularity and defended by an endothelial pump on the inner surface, which holds the corneal stroma at exactly the degree of dehydration that keeps it clear. Let that pump fail and the cornea clouds, not because anything has grown in it, but because it has taken on water.

Behind the cornea, the aqueous humour is not filler. The ciliary body secretes it continuously and it drains through the trabecular meshwork at the angle, and the balance between those two rates is what holds the globe at pressure and in shape. When drainage falls behind production, the consequence is not felt at the front of the eye at all. It is transmitted to the far end of the pathway, to the optic nerve head at stage 07, where the axons leave. The front of the eye and the back of the eye are one hydraulic system, which is why a pressure measured at the cornea says something about a nerve 24 mm behind it.

Notice the ordering. Light has been bent twice before it is a millimetre into the eye, and neither of those surfaces can be adjusted by any effort of yours. Everything the eye does deliberately to focus happens later, and works on a much smaller share of the total power than most people assume.

The pupil is a hole. The iris is the muscle that sizes it, running from roughly 8 mm across in darkness to about 2 mm in bright light. That is a sixteen-fold change in area, which sounds substantial until you set it against the range of light the eye actually works across: starlight to a snowfield at noon spans around ten orders of magnitude. The pupil covers almost none of that range. The retina covers the rest, chemically, and takes minutes rather than milliseconds to do it.

What the pupil does buy is depth of focus. A small aperture forgives an imprecise focus, in an eye exactly as in a camera. This is why reading glasses feel less urgent under a bright lamp than a dim one, and why the same person is measurably more short-sighted at dusk.

Behind the iris sits the only adjustable optic in the eye, and its mechanism runs opposite to intuition. The lens hangs from zonular fibres attached to the ciliary muscle. The muscle contracts in order to release tension on those fibres, which lets the lens thicken under its own elasticity. Focusing on something close is therefore the active, effortful state, and focusing on the horizon is the resting one. The lens is not pulled into shape. It is let go into it.

Which means presbyopia is not a muscle giving up. The lens itself stiffens continuously from childhood onward, and accommodative range falls along a near-straight line for decades until, somewhere in the mid-forties, what is left no longer spans the distance to a page. The ciliary muscle still contracts as well as it ever did. The lens has simply stopped answering. That is why presbyopia arrives on a schedule, arrives for everyone, and is unmoved by anything done to the muscle.

One more thing about the lens. It has no blood supply and no way to clear old protein. Crystallins laid down before birth are still sitting in the centre of the lens at eighty. It is among the oldest continuously present tissue in the body, and its slow yellowing shifts the spectrum reaching the retina toward long wavelengths across a lifetime, quietly cutting the short-wavelength light that also happens to be the signal your body clock reads at stage 09.

Most of the eye is vitreous: a gel roughly 99 percent water, held in shape by a sparse mesh of collagen fibrils and hyaluronan. Optically it is almost nothing, which is the point. It is also the part of the eye that changes most visibly with age. The gel liquefies in pockets, the mesh coarsens, and eventually the whole vitreous body collapses forward and peels away from the retinal surface. The debris left casting shadows on the retina is what people report as floaters; the tug on the retina during separation is what they report as flashes.

Then comes the strangest fact in the entire pathway. The retina is installed backwards. Light that has survived the cornea, the pupil, the lens and the vitreous does not arrive at the photoreceptors. It arrives at the ganglion cells, and must pass through their axons, through two synaptic layers, through the interneurons, and through the full thickness of the neural retina before it reaches a photoreceptor at all. The sensors sit at the very back, pointing away from the light, with the wiring stacked in front of them.

This is not sloppiness and it is not evidence of anything. The photoreceptors have to sit against the retinal pigment epithelium, because the RPE performs the pigment recycling and the daily digestion of spent outer segments without which photoreceptors do not survive. Given that constraint, the neural layers had to go somewhere, and in front is where they went.

The optical cost is real, and the retina answers it twice. First, the Müller cells, which span the entire retinal thickness, behave as living optical fibres: they carry a higher refractive index than the tissue around them and guide light through the inner layers to the receptors with far less scatter than the same thickness of undifferentiated tissue would produce. Second, at the one place where scatter would be least acceptable, the retina simply moves the obstruction out of the way. The inner layers are swept aside, and the depression that results is the foveal pit at stage 04. The fovea is a pit because the retina got out of its own light.

The retina lines most of the inside of the eye, but the part that does the work you would call seeing is small. The macula is about 5.5 mm across. Inside it the fovea is a pit 1.5 mm wide, and at the floor of that pit the foveola is 0.35 mm: an area smaller than a pinhead, containing only cones, where cone density reaches its peak and acuity with it.

In angular terms, which is the honest way to state it, the foveola covers a little over one degree of visual field. Hold your thumb up at arm's length; the nail is about that size. Everything you believe you are seeing in sharp, saturated detail at this moment is being reconstructed from a patch of retina that small, thrown around the scene several times a second, and stitched into an apparently uniform whole by the machinery at stage 10. The sense of a wide, evenly detailed visual world is a construction, and it is the most convincing illusion the brain produces.

The macula's yellow comes from lutein and zeaxanthin concentrated in the layers in front of the receptors, where they absorb short-wavelength light before it can reach them and quench the reactive species that light produces. That is the mechanism the whole macular-pigment literature is built on, and it is why the supplement question attaches to this stage of the pathway and to no other.

At the very centre, the retina gives up its own blood supply. The foveal avascular zone, roughly half a millimetre across, has no capillaries at all, because capillaries sitting in front of the most acute part of the retina would be intolerable optically. The cones there are fed only by diffusion from the choroid behind. It is the most metabolically demanding tissue in the eye, supplied by the least direct route available, and that combination is why so much of what goes wrong with central vision goes wrong precisely here.

One consequence of the pit is worth following. Because the inner layers are swept aside to make it, foveal cones cannot connect straight down to their bipolar cells. Their axons run obliquely outward instead, forming the Henle fibre layer, and reach their partners sideways. The centre of your vision is wired at an angle in order to get around its own architecture.

Everything up to here has been optics. At the photoreceptor outer segment, light stops being light. A photon is absorbed by a molecule of 11-cis retinal buried inside an opsin protein, and the only thing it does is straighten one kink in that molecule into the all-trans form. That isomerisation is the entire light-dependent step in vision. Nothing after it requires a photon. All of it is chemistry, and all of it is amplification.

The bent retinal changes the shape of its opsin, the activated opsin switches on hundreds of copies of the G protein transducin, each of which activates a phosphodiesterase that hydrolyses thousands of molecules of cyclic GMP per second. A single absorbed photon becomes a macroscopic electrical event, which is why a dark-adapted rod can produce a measurable response to one quantum of light. There is no more sensitive detector in biology, and remarkably few in physics.

Now the part that surprises people who assume the eye works like a camera sensor. The signal runs backwards. In darkness, cyclic GMP holds cation channels open and the photoreceptor sits depolarised, pouring current across its membrane and releasing neurotransmitter continuously. That is the dark current. Light causes the cGMP to be destroyed, the channels to close, and the cell to hyperpolarise and go quiet. Photoreceptors do not fire when they see light. They stop.

That inversion has a metabolic consequence that runs through everything else on this site. The expensive state is darkness, because darkness is when the current is running and the pumps are working to sustain it. Photoreceptors are among the most energy-hungry cells in the body, and their peak demand is at night. Layer onto that the fact that the outer segments are continuously rebuilt from the base and shed from the tip, with the retinal pigment epithelium phagocytosing the discarded discs every morning, and you have a tissue that consumes enormous energy simply to stay where it is.

Finally, the loop has to close. All-trans retinal cannot absorb another photon; it has to be transported out to the retinal pigment epithelium, re-isomerised back to 11-cis, and returned. That is the visual cycle, and its slowness is something you have felt. Walking into a dark room, the first minutes of recovery are cones. The rod-cone break arrives several minutes in, and full dark adaptation takes something closer to half an hour. You are not waiting for your pupils. You are waiting for chemistry.

Between the photoreceptors and the output neurons sit two synaptic layers and three classes of interneuron, and what happens there is not relaying. It is computation, and it discards most of what the receptors captured.

Bipolar cells split the signal immediately into two channels: ON cells that depolarise when light increases, OFF cells that depolarise when it decreases. From this point onward the visual system carries brightening and darkening as separate, parallel streams, all the way to the cortex. Horizontal cells then feed inhibition sideways, so that the response of any bipolar cell is suppressed in proportion to the illumination of the ring of retina around it. Amacrine cells, of which there are more than thirty types, do the equivalent job in time rather than space, sharpening transients and building the first direction-selective responses.

The result is the receptive field that Kuffler described in 1953 and that has organised the field ever since: centre-surround. A retinal neuron does not report how bright its patch of the world is. It reports how much brighter or darker that patch is than its immediate neighbourhood. Illuminate the centre and the surround equally and the cell says almost nothing at all.

This is why an evenly lit wall generates so little traffic, and why the visual system is built around edges. You can see the mechanism working directly: the Mach bands that appear at the boundaries between adjacent grey strips are not in the light, they are in the lateral inhibition. The same circuit is why a grey square looks lighter on a dark background than on a pale one. These are not failures of perception. They are the arithmetic of the retina, visible.

Then there is the compression, which is the number that reframes everything else. Something like 95 million photoreceptors feed roughly 1.2 million optic nerve fibres. Around eighty to one, before the signal has left the eye. The retina is not a sensor that hands off an image. It is a piece of brain, developmentally and functionally, that throws away most of what it collects and forwards a heavily processed summary. Whatever it discarded here is gone. No later stage can recover it.

Retinal ganglion cells are the only neurons in the eye that talk to the brain. Everything upstream is preparation; these cells carry the result. They are central nervous system neurons with unusually long axons and a punishing energy budget, and they are the cells that die in glaucoma and in optic neuropathy. When they are gone they do not come back.

They are not one population but several, running in parallel and never fully merging. Midget cells, around eighty percent of the total, have tiny receptive fields and carry fine spatial detail and red-green colour. Parasol cells have large fields and fast responses, and carry motion and luminance contrast. Bistratified cells carry the blue-yellow signal. A fourth group, the intrinsically photosensitive ganglion cells, contain their own photopigment and respond to light directly without needing rods or cones at all; they are not building a picture, and their destination is stage 09.

Their axons run across the inner surface of the retina, unmyelinated, because myelin would scatter light in front of the receptors. Transparency is bought at the cost of conduction speed and a higher energy demand per millimetre, which is one reason this layer is fragile. Its thickness is measurable on OCT and thinning of it is one of the most useful structural biomarkers in ophthalmology.

All of them converge on one point and leave together through a sieve of collagen called the lamina cribrosa. At that point there is no room for photoreceptors, and there are none. Every eye has a blind spot roughly fifteen degrees from fixation, on the temporal side of the visual field, and about five degrees across: several full moons wide, sitting permanently in the visual field of each eye.

You have never noticed it. Partly because the two eyes cover each other's holes, but that is not the interesting part, because the blind spot is invisible with one eye closed too. What happens is that the visual system fills the region in with whatever surrounds it. Put the blind spot over a striped background and you see unbroken stripes across it. The brain is not reporting an absence of data. It is producing a confident answer where it has none, and it does not flag the difference. That is worth sitting with, because it is a demonstration, in your own head, that seeing is inference rather than recording.

The lamina cribrosa deserves one more line. It is a mechanical weak point: the place where a pressurised eye meets an unpressurised skull, and where the axons must pass through holes in a collagen mesh. Raise the pressure inside the eye and this is where the damage concentrates. It is the anatomical reason that a plumbing problem at stage 01 becomes a neurological one at stage 07.

Behind the eye the axons acquire myelin, produced by oligodendrocytes, and the character of the structure changes. The optic nerve is not a peripheral nerve at all. Developmentally and histologically it is a tract of the central nervous system that happens to run outside the skull, which is why it does not regenerate the way a nerve in your arm would, and why demyelinating disease of the brain shows up in it.

The two nerves then meet beneath the hypothalamus and do something that looks pointless until you work out what it accomplishes. Fibres from the nasal half of each retina cross to the other side. Fibres from the temporal half stay put. Half the traffic changes lanes.

Because the eye's optics invert the image, the nasal retina views the temporal field. So the effect of a partial crossing is that everything to the left of where you are looking, seen by either eye, ends up together in the right hemisphere, and everything to the right ends up in the left. The brain is not organised by eye after the chiasm. It is organised by half of the world.

The clinical payoff is unmatched anywhere else in neurology. Because the fibres are sorted geographically, the shape of a person's field loss reports the location of the lesion, without imaging. Loss confined to one eye is in front of the chiasm. Loss of both outer halves, bitemporal, means the crossing fibres themselves are being compressed at the midline, and the classic cause is a pituitary tumour growing upward. Loss of the same side in both eyes, homonymous, is behind the chiasm; the further back the lesion, the more precisely the two eyes' defects match each other. And an occipital lesion often spares the very centre, because the foveal representation at stage 10 is large and supplied from two directions.

This is the reason anatomy is worth learning rather than looking up. The map is not decoration. It is diagnostic.

Most of the optic tract terminates in the lateral geniculate nucleus of the thalamus, a structure of six layers, each of which receives input from one eye only. Two ventral layers take the fast magnocellular input from parasol cells, four dorsal layers take the parvocellular input from midget cells, and thin koniocellular sheets between them take the blue-yellow signal. The streams that separated in the retina are still separated here.

The word relay undersells it. The LGN receives more synaptic input descending from the cortex than ascending from the retina. Whatever it is doing, it is not passing messages along unaltered; the cortex is talking back to it constantly, and the traffic is mostly in the direction you would not expect.

From there the optic radiations sweep back to the occipital lobe, and their route matters clinically. The fibres carrying the upper visual field loop forward through the temporal lobe before turning back, a detour called Meyer's loop, which is why temporal lobe surgery can produce a defect in the upper quadrant of the opposite field.

But a substantial minority of retinal axons never go to the LGN at all, and what they do is older than image formation. One group reaches the superior colliculus, a layered midbrain structure that maps visual space directly onto motor commands and turns the eyes and head toward a stimulus. That is the circuit that has already moved your eyes toward movement in the periphery before you could say what moved. In people whose primary visual cortex is destroyed, this route appears to survive: they report seeing nothing, and yet perform above chance at locating stimuli they deny perceiving, which is the phenomenon called blindsight.

Another group ends in the pretectum, which drives the pupillary light reflex through the Edinger-Westphal nucleus and the third cranial nerve. This is why a clinician shines a light in your eye: they are testing a circuit that runs retina, midbrain, brainstem, nerve, iris, and never once passes through the part of the brain that sees.

And a third group, from the melanopsin ganglion cells, ends in the suprachiasmatic nucleus of the hypothalamus, the body's master clock. Melanopsin is most sensitive around 480 nanometres, in the blue. That single fact is the whole biological basis of the argument about light at night. Light striking your retina in the evening is a timing signal before it is anything visual, and the receptor that reads it does not care whether you are looking at the screen or merely near it.

The primary visual cortex sits at the back of the occipital lobe and is laid out retinotopically: neighbouring points in the visual field land on neighbouring points of cortex. But the map is wildly distorted. The central ten degrees of the visual field, a tiny fraction of what you can see, occupies at least half of V1. Cortical magnification is the formal name; the practical meaning is that the brain's allocation of tissue matches the retina's allocation of receptors, and neither matches your subjective impression of seeing a wide, uniform scene.

Within that map, Hubel and Wiesel found the cortex sorted into columns responding to specific edge orientations, and into alternating bands dominated by one eye or the other. The first cortical operation on the visual signal is to ask what direction the edges run. Not what is here, but which way does it go.

Beyond V1 the pathway fans out. V2 and V4 build form and colour constancy, the ability to see a shirt as the same white indoors and outdoors even though the light reaching your eye differs enormously. Area MT is dominated by motion. Broadly, the fibres running upward into the parietal lobe carry spatial and action-guiding information, and those running down into the temporal lobe carry identity, an organisation usually described as the dorsal and ventral streams, or where and what.

And then there is the problem of the input itself. Your eyes make several large saccades a second, and during each one vision is actively suppressed, so you are functionally blind for a meaningful fraction of every waking hour. You have a permanent hole in each field from stage 07. Your retina reports contrast, not brightness. The whole thing arrives inverted, compressed eighty to one, from a sensor array that is denser in the middle by orders of magnitude.

Between saccades the eye still never holds still, and that turns out to be necessary rather than sloppy. Stabilise an image perfectly on the retina and it fades from perception within seconds, because the system responds to change and an unchanging signal stops being reported. The tiny microsaccades and drifts of fixation exist to keep refreshing it. You are able to keep seeing a stationary object only because your eye refuses to stop moving.

What you experience is a stable, sharp, continuous, gap-free world. None of those four properties are in the data. All of them are constructed. That is not a caveat at the end of the anatomy; it is the conclusion the anatomy was leading to. Vision is not a recording of what is in front of you. It is the brain's best current explanation of a sparse, interrupted, heavily edited signal, and it presents that explanation with no indication anywhere that it is one.

Read the ten stages back in order and a pattern emerges. The tear film destabilises, the lens stiffens and yellows, the pupil shrinks, the vitreous liquefies, the RPE fills with the residue of its own housekeeping, Bruch's membrane thickens, ganglion cells are lost and the nerve fibre layer thins. Different tissues, different timescales, and underneath them a short list of processes that are not specific to the eye at all.

What singles the visual pathway out is not that it ages faster. It is that so much of it is post-mitotic and metabolically extreme at the same time. Photoreceptors, RPE cells and ganglion cells largely cannot be replaced, and all three run energy budgets near the top of anything in the body. Tissue that cannot divide and cannot rest accumulates damage in a way that tissue doing only one of those things does not.

The terms below are the systemic processes involved, and they are where this laboratory's research sits. They are stated as mechanisms under study. Nothing here is a claim that any catalogued material treats, prevents or reverses any condition, in the eye or anywhere else.