Module 1: Vision Spectral Atlas

All vertebrate photoreception descends from a single ancestral opsin that diversified in the jawless-fish lineage into five canonical cone classes, plus rhodopsin (Yokoyama, 2008; Bowmaker, 2008):

Placental Mammals Lost Half Their Colour Vision

During the nocturnal bottleneck of Mesozoic mammal evolution, SWS2 and Rh2 were deleted, leaving most placentals as dichromats with SWS1 + LWS only (Jacobs, 2009). Old-World primates re-evolved trichromacy ~30 Mya by duplicating the LWS gene on the X-chromosome, splitting it into M (534 nm) and L (564 nm)—handy for finding ripe fruit against green foliage (Mollon, 1989). New-World primates (e.g. marmosets) retain an allelic LWS polymorphism with mixed phenotypes.

The Govardovskii Template

Govardovskii et al. (2000) fitted hundreds of measured vitamin-A1 pigment spectra to a two-parameter template, which with \(x = \lambda_{\max}/\lambda\)gives:

A single parameter \(\lambda_{\max}\), tuned by key amino-acid substitutions (Yokoyama’s five spectral-tuning sites), predicts the entire absorption curve. Our simulation below uses this template to render every cone/rod in the visual-spectrum atlas.

An arthropod compound eye is built from hundreds to tens of thousands of ommatidia, each a miniature photoreceptor bundle capped by a cornea and crystalline cone. The total interommatidial angle\(\Delta\varphi\) sets the spatial acuity:

Apposition vs. Superposition

In an apposition eye (diurnal insects, crustaceans) each ommatidium has its own isolated optical axis; ray bundles from adjacent facets do not mix. Resolution per facet is high but absolute sensitivity is low. In a superposition eye (moths, krill, deep-sea shrimp) many facets share a single deep rhabdom layer and converging lenses superimpose their images, increasing photon catch \(\sim 100\times\) at the expense of resolution (Nilsson, 1989).

Seitz’s Resolution Limit

Seitz (1970) combined the diffraction limit through a single facet with the photon-noise statistics to find a minimum ommatidium size:

This explains why increasing the number of facets beyond a point gives diminishing returns: each facet becomes too narrow to avoid diffraction blur. Dragonflies push this limit: Anax has ~30,000 facets per eye (Land & Nilsson, 2012), yielding panoramic hemispherical coverage with an acuity peak around\(\Delta\varphi \approx 0.24^\circ\) in the dorsal region used for prey-tracking.

Honeybee Trichromacy

Apis mellifera has cones peaking at 344 nm (UV), 436 nm (blue) and 544 nm (green), shifted to shorter wavelengths than humans (Menzel & Blakers, 1976; Chittka, 1994). Many flowers have UV-absorbing “nectar guides” invisible to humans but vivid to bees—e.g., the central dark bulls-eye ofRudbeckia. Bees thus experience a fundamentally different chromatic world, with colour-mixing rules computable in the hexagonal bee colour space of Chittka (1992).

Camera eyes share the same optical blueprint as a camera: a single aperture (pupil), a lens that forms an inverted image, and a tiled detector array (retina). They have evolved at least seven times independently—in vertebrates, cephalopods, cubozoan jellyfish (Nilsson et al., 2005), alciopid annelids, certain pteropod gastropods, chitons, and spiders. Convergence this dramatic speaks to the optical physics doing most of the design work.

Inverted vs. Everted Retina

Vertebrates carry a peculiar legacy: the retina is inverted—photons must traverse ganglion cells, bipolar cells and blood vessels before reaching the opsin-bearing outer segments. Cephalopods (octopus, squid) have an everted retina where photoreceptors point out toward the light, and photoreceptor axons exit directly to the brain without forming a blind spot. The vertebrate architecture is compensated for by Müller cells, which act as optical fibres guiding light through the neural layers (Franze et al., 2007).

Walls’ Hypothesis of Eye Origin

Gordon Walls (The Vertebrate Eye and its Adaptive Radiation, 1942) argued that the vertebrate retina’s inversion is a consequence of its embryonic origin as an outgrowth of the neural tube, with the photoreceptors on the outersurface that becomes internal upon invagination. Cephalopod eyes develop from ectodermal placodes, giving the opposite polarity. Nilsson (1996, 2009) showed that gradual selection on a flat photosensitive patch can reach a camera eye in\(< 400{,}000\) generations—and has done so repeatedly.

Resolution Limits

Two physical bounds compete: diffraction through the pupil, and retinal sampling by photoreceptor density:

Human central fovea: \(D \approx 5\) mm, \(f \approx 17\)mm, \(d_{\text{cone}} \approx 2.5\) μm, giving \(\alpha \approx 0.5\) arcmin (60 cycles/deg peak).

Along the edge of a Pecten scallop mantle sit up to 200 blue eyes the size of pinheads. Each contains a concave multilayer reflector behind the retina, made of stacked guanine crystal plates. Light passes through the pupil, transits the retina, reflects off the mirror, and re-excites the retina from behind (Land, 1965, 1968).

Palmer et al. (2017; Science 358, 1172) imaged the mirror at nanometer resolution and showed it is tiled like a telescope’s segmented mirror, with the facets optimised for off-axis imaging of dim points. The scallop has TWO retinas per eye: a proximal retina receiving the focused image from the mirror, and a distal retina sampling the out-of-focus light—a differential comparison that detects motion (a predator’s shadow) with high contrast. This is the only known biological imaging system using reflection rather than refraction as its primary focusing mechanism.

Raptors, especially the wedge-tailed eagle Aquila audax and the peregrine falcon Falco peregrinus, have the sharpest vision on earth. Reymond (1985) measured wedge-tailed eagle acuity at 140 cycles/degree—more than twice human peak—achieved via:

• Two foveae per eye: a deep central fovea for monocular sideways scanning, and a shallow temporal fovea for binocular forward vision during the strike.
• Cone density ~1,000,000/mm² in the central fovea (humans: 200,000/mm²).
• Elongated eyeball: focal length ~35 mm vs human 17 mm, doubling image magnification on the retina.
• Coloured oil droplets in cone inner segments acting as in-line long-pass filters, narrowing each cone’s spectral response and sharpening colour discrimination (Hart, 2001).

Oil-Droplet Filtering

Birds place a pigmented oil droplet at the base of each outer segment. The droplet absorbs shorter wavelengths before they reach the opsin, effectively subtracting part of the Govardovskii curve and producing a narrower, redshifted spectral sensitivity. Four oil-droplet types (T=transparent, C=clear, Y=yellow/carotenoid-rich, R=red) combine with four cone opsins (SWS1, SWS2, Rh2, LWS) to produce near-orthogonal tetrachromatic channels.

Why Two Foveae?

An eagle hunting at 100 m must monitor a wide area (requiring panoramic vision) and, upon spotting prey, transition to precise distance judgement for the kill (binocular stereopsis). The central fovea has a deep concave pit that magnifies the image like a telephoto lens (Snyder & Miller, 1978), while the temporal fovea is flatter and is used binocularly when the head is pointed forward.

The stomatopod Odontodactylus scyllarus has the most elaborate visual system ever discovered. Marshall (1988, 2007) mapped 16 photoreceptor types per eye, organised in a midband of six rows:

• 12 spectral channels spanning 310–700 nm, with lambda_max at 313, 336, 380, 400, 425, 455, 488, 515, 543, 570, 615, and 640 nm.
• 4 linear polarisation channels in the dorsal and ventral hemispheres.
• Circular-polarisation sensitivity uniquely in stomatopods (Chiou et al., 2008) — the only known biological circular-polarisation vision.
• Each eye moves independently and can fixate targets with a single specialised midband region.

The Thoen Paradox

Despite 12 colour channels, behavioural experiments by Thoen et al. (2014; Science 343, 411) showed that Haptosquilla trispinosa discriminates wavelengths only coarsely—about 25 nm apart, compared to 1–4 nm in humans. The interpretation: mantis shrimp use a labeled-line code rather than interpolating between opposing channels. Their 12 channels act like a spectrograph with discrete bins, enabling rapid (low-latency) classification without the neural circuitry needed for opponent computations. This trades spectral resolution for speed—critical for an animal that delivers a 23-m/s strike with its raptorial appendage.

Polarisation Vision

Microvilli in adjacent rhabdomeres are orthogonally oriented, making the photoreceptors intrinsically sensitive to linear polarisation. Mantis shrimp further possess a quarter-wave retarder built from unusual R8 distal cells, converting circular to linear polarisation. This is used in mate signalling: O. scyllarus telson reflects circularly polarised light, invisible to predators without the retarder.

Most diurnal birds are tetrachromats: SWS1 (violet or UV), SWS2 (blue), Rh2 (green), LWS (red), each paired with coloured oil droplets. SWS1 lambda_max ranges from 355 nm in European starlings Sturnus vulgaris(Wilkie et al., 2000) to 405 nm in pigeons. The shift from violet-sensitive (VS) to ultraviolet-sensitive (UVS) is controlled by a handful of amino-acid substitutions in helix III, most notably F86C (Yokoyama & Shi, 2000).

UV vision is used for:

• Sexual selection: bluetits Cyanistes caeruleus have UV crown reflectance invisible to humans but visible to potential mates (Andersson et al., 1998).
• Prey detection: kestrels Falco tinnunculus follow vole urine trails that fluoresce in UV (Viitala et al., 1995).
• Foraging: many berries and flowers have UV-reflective patterns invisible to mammalian seed-dispersers.
• Navigation: UV polarisation patterns aid migration.

Chicken “Vision-Perfected” Retina

Kram et al. (2010) showed that chicken cone mosaics are organised as a hyperuniform random tiling—statistically irregular yet with long-range uniform density—optimal for sampling natural scenes without aliasing. This pattern has since been engineered into camera sensors.

Crotalinae (pit vipers: rattlesnakes, copperheads, bushmasters) and pythons have loreal or labial pit organs that form a pinhole thermal camera, imaging thermal infrared at 7–15 μm. The organ is a bilobed cavity about 1 mm across, innervated by ~7,000 trigeminal fibres densely packed under a thin (~15 μm) tympanic membrane.

TRPA1 is the Thermal Detector

Gracheva et al. (2010; Nature 464, 1006) showed that the molecular sensor is the TRPA1 ion channel—not an opsin. TRPA1 normally responds to noxious chemicals (wasabi, acrolein) in other animals, but in pit vipers it has been reshaped by gene duplication and positive selection into a heat-threshold channel firing above\(\sim 28^\circ\text{C}\). Blackbody infrared at 7–15 μm heats the pit membrane by tens of millikelvin, and TRPA1-expressing C-fibres encode the thermal image with \(\sim 0.003 ^\circ\text{C}\)sensitivity (Bullock & Cowles, 1952; Newman & Hartline, 1981).

Spatial Resolution

The pinhole aperture and detector density set a spatial resolution of\(\sim 4^\circ\)—coarse, but sufficient to localise warm-blooded prey at 1 m in total darkness. The optic tectum fuses the thermal and visual images so the snake experiences a single multimodal scene (Newman & Hartline, 1981).

Boid vs. Pythonid Convergence

Boas and pythons have independently evolved labial pit organs with the same convergent solution of TRPA1 re-tuning. This is a clean example of convergent molecular evolution: the same gene was re-purposed for the same function in two separate lineages (Gracheva et al., 2010).

In the bathypelagic zone (\(>1000\) m), most bioluminescence is blue (~475 nm) because seawater transmits blue most efficiently. Nearly all deep-sea fish have a single visual pigment peaked near 480 nm. But three genera of stomiid dragonfish—Malacosteus niger, Aristostomias,Pachystomias—have evolved red photophores emitting at 705–710 nm, together with long-wavelength visual pigments peaking at \(\lambda_{\max} \approx 540\) and\(\sim 590\) nm (Douglas et al., 1998, 2000).

The result is a private communication channel: the dragonfish can see its own red bioluminescence but its prey (blue-sensitive shrimp, other fish) cannot. Even more remarkable, Malacosteus uses a bacteriochlorophyll-derived photosensitiser, absorbed via dietary copepods, to extend its own retinal sensitivity into the deep red (Douglas et al., 1998)—a rare case of heterologous photopigment in an animal.

Mouse, Cat, Bushbaby: Tapetum Lucidum

Most mammals are dichromats (SWS1 + LWS) reflecting their nocturnal ancestry. Carnivores and primates with nocturnal habits (cats, lorises, bushbabies) carry a tapetum lucidum—a reflective layer behind the retina that doubles photon catch at the cost of retinal blurring. Cats have rod:cone ratios of \(\sim 25{:}1\) compared to humans’\(\sim 20{:}1\) but with a far larger tapetum, explaining their glowing eyes at night.

Mouse cones are special: they carry an opsin gradient across the retina, with UV (SWS1 at ~360 nm) in the ventral retina and green (M-opsin at ~508 nm) dorsally (Baden et al., 2013). The animal looks up at the sky with UV-sensitive cones (to detect aerial predators silhouetted against UV-rich sky) and down at the ground with green cones.

Polarisation in Cephalopods

Octopus and cuttlefish are colour-blind (monochromat, one opsin), but exploit polarisation vision for communication, camouflage detection, and seeing through reflective water surfaces. Temple et al. (2015; Curr. Biol. 25, R209) showed cuttlefish can discriminate polarisation angles of \(<1^\circ\), the sharpest known. Dynamic body patterns include polarised displays invisible to most predators but conspicuous to conspecifics.

Polarisation in Bees

The bee dorsal-rim area has specialised ommatidia with orthogonal microvilli encoding e-vector orientation. Von Frisch (1949) showed bees use the polarisation pattern of the sky as a celestial compass, allowing navigation even when the sun is obscured by clouds. The insect central complex compiles these compass readings into a ring-attractor head-direction signal (Seelig &amp. Jayaraman, 2015).

Jumping Spider Anterior Median Eyes (Salticidae)

Jumping spiders—famously Phidippus and Portia—possess four pairs of eyes. The two anterior median (AM) eyes are tiny telescopes with a four-layer tiered retina behind a movable muscle-actuated ocular tube. Each retinal layer has different spectral tuning, creating a chromatic aberration depth-from-defocus system: the spider can estimate range to a prey by comparing focus across layers without needing stereopsis (Nagata et al., 2012; Science 335, 469). This is the only known use of chromatic aberration as a depth cue in animal vision.

Chameleon Independent Eye Control

Chameleons rotate their eyes independently through up to 180° of arc, monitoring two different visual fields simultaneously. When a prey is detected, the eyes converge for fused binocular vision before the tongue strike. Harkness (1977) and Ott & Schaeffel (1995) showed that chameleons use a negative lens (diverging, unlike most vertebrates) combined with a concave cornea to achieve telephoto magnification, giving high acuity through a narrow field of view—perfect for a sit-and-wait predator.

Four-Eyed Fish Anableps

The Amazonian Anableps anableps swims at the water surface and has eyes split horizontally by a pigmented band. The upper half of each eye sees in air, the lower half underwater. Because the refractive index of water (1.33) and air (1.00) create different effective focal lengths, the lens is oval with different radii in the two hemispheres (Schwab et al., 2001). The retina has two separate cone-density peaks, one per half-eye. This is the cleanest example of asymmetric optical design driven by environmental niche.

Box Jellyfish Camera Eyes

The cubozoan jellyfish Tripedalia cystophora has 24 eyes arranged in four clusters (rhopalia). Six of the 24 are sophisticated camera eyes with a lens, iris and retina (Nilsson et al., 2005;Nature 435, 201), remarkable in an animal without a true brain. The eyes are upper- and lower-directed, used to navigate the mangrove roots where the jellyfish hunts copepods. All six rhopalia send crude positional signals to a decentralised ring nerve; decision-making emerges from this distributed network.