Avian Visual Cognition

Husband & Shimizu Home Page

Next Section:

Taking Flight: Post Retinal Processing
___________

 

 

IV. Evolution of Retinal Structures

Despite the sophisticated visual abilities that birds and primates share, the visual systems of non-mammalian amniotes (reptiles and birds) contain some interesting differences compared to primates.  The answer to the question of whether the visual systems of birds are qualitatively or quantitatively different may lie in the particulars of their retinal structures and the organization of their central visual pathways.  The following data concentrates on that derived primarily from studies in the pigeon (Columba livia). 

Retinal Morphology and Processing  

The basic retinal structure of vertebrates is very similar, usually characterized by the presence of five major layers and five major cell types.  The five layers are the outer nuclear and plexiform layers, the inner nuclear and plexiform layers, and the ganglion cell layer.  Five major classes of retinal neurons are also recognized:  photoreceptors, bipolar cells, horizontal cells, amacrine cells, and ganglion cells.  The photoreceptors (rods and cones), bipolar cells, and horizontal cells make synaptic contacts with each other in the outer plexiform layers. The bipolar, amacrine, and ganglion cells make contact in the inner plexiform layers.  

Despite basic similarities with primates, the avian retina contains important differences in the number, ratio, distribution, and morphological subclasses of retinal cells, and in their physiological responses.  For example, the avian inner nuclear layer is especially rich in horizontal and amacrine cells compared to that of primates.  The anatomist S. Ramón y Cajal noted the complexity of inter-connections in the avian retina; he thought this probably represented a great deal of intra-retina visual processing (Rodieck, 1973).  This may reflect an important difference between primate and avian visual systems in that complex processing which occurs in higher forebrain areas in primates may be achieved at a lower level in birds.

Photoreceptor & photopigment evolution. Which of the two primary classes, rods or cones, is the ancestral photoreceptor?  Given the tremendous variation seen photoreceptors across vertebrate and invertebrate species, this in not an easy question to answer based on simple phylogenetic assumptions.  In addition, it is often difficult to clearly distinguish certain rod and cone types from each other, or classify them into one or the other category.  Rods appear to be relatively more conserved in vertebrates in terms of pigments and structure than cones, and therefore could be considered the more ancestral form.  However, rods in some respects are more morphologically complex than cones, having developed extreme sensitivity (capable of detecting as little as one photon of light).  The cyclostomes (e.g., lampreys and hagfishes) descended from a now extinct line of jawless, fish-like vertebrates of the Devonian period.  These organisms are thought to have a similar morphology to some of the earliest vertrbrates (Bowmaker, 1991).  Lampreys have two classes of photoreceptors, consisting of a short outer-segment type and a long outer-segment type.  It is still debated whether these are rods or cones.  Nevertheless, lampreys appear to have duplex retinae with receptors demonstrating wavelength sensitivity maxima at 517nm for the shorter, more "rod-like" type and 555nm for the longer, more "cone-like" type. 

Based on several lines of evidence the ancestral vertebrate visual system probably had a relatively unspecialized photoreceptor that was not strictly classifiable as either a rod or cone.  This receptor probably exhibited two spectral classes of photopigment at a very early stage in evolution (Bowmaker, 1991).  The radiation of amphibians, reptiles, birds, and mammals led to immense changes in photoreceptors and pigment types due to the differing demands of their operating environments.  

The functional spectral range of photopigments is constrained by several factors.  Vertebrates have photopigments with ranges of maximum spectral sensitivity from the ultraviolet (350nm) to the far red (630nm).  The earth's atmosphere filters much of the spectral radiation, and the 80% which reaches the earth's surface is in the 300-1100nm range.  The energy distribution has a maximum of 480nm and a quantum distribution with a maximum at 555nm (Bowmaker, 1991).  Photons above 850nm are not energetic enough to photoisomerize organic molecules, and photons below 300nm can be destructive to organic molecules (Bowmaker, 1991).  Thus, maximum biologically relevant range is about 300-850nm.  This range is itself often constrained by ecological factors, for example the photopigments of deep sea fish.  Light traveling through the water is reduced at both short and long wavelengths; at increasing depths the light is not only attenuated but is reduced first at longer wavelengths, and then shorter ones as the depth increases.  Most deep sea fish have all rod retinas, with maximum sensitivities at 470-480nm, which matches what would be predicted from the wavelength transmission properties of clear water (Bowmaker, 1991).  

Opponent-processing functions in color vision probably evolved relatively early in the history of vertebrate visual systems.  Opponent processing has been found in the retina of the bowfin fish Amia calva, a species considered one of the most unchanged or conserved from its evolutionary ancestry (Burkhardt et al., 1983).  The presence of an opponent processing system in this "hold-over" from the dinosaurs, and its existence in a variety of fish, reptiles, birds, and mammals, indicates that such a system evolved very early as a mechanism for spectral differentiation.  It most likely originated in some common vertebrate ancestor many millions of years ago rather than through convergent evolutionary processes.

Avian Retina

Photoreceptors - double cones. Although the avian retina has a duplex nature (i.e., containing both rods and cones) similar to that of primates, it also possesses (along with reptiles, many fish, and amphibians) photoreceptors known as double cones.  In fact, all vertebrate classes, with the notable exception of placental mammals, possess such double cones (Sillman, 1973).  The double cone consists of a principle cone (similar in structure to a normal single cone) and an accessory cone which curves around the inner segment of the principle cone (see Figure).  In the scheme of photoreceptor evolution proposed by Walls (1942), the double cones were probably present in the Captorhinda (ancestral amniotes) and possibly in the dinosaurs as well.

The principle and accessory segments of the double cone may contain different visual pigments. The interaction of signals from cones with different pigment types is crucial to color vision. The electrical signals generated by the double cones in response to different wavelengths of light indicate that each member of the pair has strong interactions with the other.  This may be functionally comparable to the type of interaction which occurs in primates, but in the context of single cones in a trichromatic system influencing the retinal ganglion cells.  Hence, whereas initial color processing occurs at the ganglion cell level in primates, the avian visual system may perform color processing steps within the photoreceptors themselves.  Despite the differences inherent in these systems, they appear to achieve comparable color abilities.  

Avian photoreceptors - oil droplets. The single cones, as well as one or both segments of double cones, may also contain an oil droplet.  Oil droplets are a common feature of the cones in many vertebrate retinas, but are rarely found in rods.  These drops consist of lipids in which carotenoid pigments are dissolved; they can appear transparent or clear, pale yellow, green, orange, or red.  They are positioned at the distal end of the inner cone segments, covering the entire width of the receptor.  Light passes through the droplet before entering the photosensitive outer segment.  Colored oil droplets are presumed to have evolved from the colorless cone forms found in other vertebrates like the fish coelacanth (Bowmaker, 1991).  Evidence suggests that oil droplets are more rapidly influenced by natural selection that opsins (photopigment elements); (Varela et al., 1993).  Oil droplets were probably present in the retina of Captorhinda and most likely in the dinosaurs. 

Oil droplets, they are widely thought to act as cut-off filters, absorbing light below their characteristic wavelengths of transmission and conveying longer wavelengths to their associated photopigments.  This would have a net effect of shifting maximal sensitivity towards longer light wavelengths.  Other theories of oil droplet function include the reduction of chromatic aberration and enhancement of contrast.  For instance, Walls (1942) argued that oil droplets probably do not serve a color function because the avian fovea is devoid of any red oil droplets.  He concluded that it was not logical for the area thought to have the highest acuity to lack color discrimination ability at this end of the spectrum.  Yellow oil droplets could function to reduce the chromatic aberration and scattering attributable to shorter wavelengths.  This would be similar to the reduction of short wavelength scattering performed by the yellowed lenses in animals like diurnal squirrels and humans (Walls, 1942).  Additionally, Walls suggested the red and yellow droplets could both serve to enhance contrast under different conditions.  For example, yellow droplets would help enhance contrast between an object as seen against a blue sky, while red droplets would reduce the effect of a green background (e.g., viewing an object on the ground).  A high density of red and orange oil droplets in the pigeon retina are found in the superior dorsal quadrant, an area that has been called the "red field."  Its shape is approximately circular and occupies almost all of the superior dorsal quadrant of the retina, extending nearly to the fovea.  The remainder of the retina in the pigeon is often termed the "yellow field."  It has been suggested the pigeon retina's ventro-nasal yellow field is in an ideal position for viewing the blue sky, and concurrently the retina's dorso-temporal red field is well suited for viewing the green ground (Sillman, 1973).  For more details on the implications of these areas of the pigeon retina, see the chapter by Blough, 2001 on visual search.  

Colorless oil droplets found in many birds may relate to the ability to perceive very short spectral wavelengths (e.g., ultraviolet or near ultraviolet).  The yellow tint of the primate lens filters out such wavelengths, and cones containing photopigments sensitive to ultraviolet (UV) or near ultraviolet wavelengths (and presumably the ability to perceive such have not been found in primates.  There is evidence, however, that such mechanisms exist in a wide a variety of other organisms like honeybees, fish, and birds.  The performance of visual discriminations based on UV wavelengths indicates the pigeon has the ability to detect such light; however it has been very difficult to chemically isolate the photopigment which performs this function.  UV wavelengths could serve a signaling function, since bird plumage tends to have shorter wavelength reflectance than many other natural objects (Varela et al., 1993).  Ultraviolet reflectance could also be used as a cue in discriminating foods (e.g., plants, seeds, berries) or other natural objects.  Sensitivity to UV may also play a role in aerial navigation as an adaptation to the coloration of an unclouded sky (Valera et al., 1993).  Short-wavelength gradients would vary depending on the sun's angle in the sky.  These gradients would increase in saturation as the sun moves from directly overhead to an angle of 90 degrees.  Color gradients in the sky due to UV or polarized light may also assist navigation when the sun is hidden by clouds.  It is difficult to appreciate the possibly rich realm of a UV "color" space, since we have no direct experience with it.

Birds appear to have excellent color vision, which may be based on as many as four or five different photopigments. This contrasts with the three cone pigments in primates. (See Figure 1)Evidence for tetra- or pentachromatic vision in birds versus the trichromacy present in primates comes from analysis of the chemical properties Click here to view Figure 1 of visual pigments and behavioral discriminations.  Pigeons possess a rod photopigment (rhodopsin) with a maxima of 500nm, and iodopsin and related cone pigments with maxima at about 413, 467, 514, and 567nm (Varela et al., 1993).  Behaviorally, the best color discrimination in pigeons occurs at 460, 530, and 595nm (Emmerton & Delius, 1980).  Pigeons can also discriminate wavelengths toward the ultraviolet end of the spectrum, in the range of 360-380nm.

The inner nuclear layer - bipolar cells. Common to all vertebrate retinae are bipolar cells which show center-surround opponency and separate ON- and OFF- center types.   The defining characteristic of bipolar cells is their dendritic contact with the photoreceptors.  The bipolar cells of birds, reptiles, and amphibians are very similar, consisting of two types recognized by the anatomist Ramon Cajal:  outer (or large) and inner (or small) bipolar cells (Rodieck, 1973).  Cajal also noted that the inner plexiform layer of birds is very thick compared to the other amniotes (Rodieck, 1973).  In mammals, bipolar cells tend to share characteristics with teleost fish rather than reptiles or birds, and consist of three recognized types: rod, cone, and giant bipolars (Rodieck, 1973).  An important difference across species is the degree to which the bipolar cells arborize into more than one sublamina of the inner plexiform layer.  The echidna (one of the monotremes) shares this trait with several non-mammalian species.  It has been suggested that this organization indicates more mixing of the ON- and OFF- pathways in non-mammalian retinas (Rodieck, 1973).  This could have effects on the processing of weak contrast targets and backgrounds, or the ability to ascertain rapid intensity shifts across the retina.  If the degree of inter-sublaminar spread is functionally significant, it may indicate a higher sensitivity to low spatial frequencies or motion detection in non-mammalian amniotes compared to primates.  Conversely, such functions in primates could be compensated for by more central mechanisms, perhaps in the tectum or higher areas.

The inner nuclear layer - horizontal cells. The horizontal and amacrine cells serve to mediate the lateral spread of visual activation in the retina.  Horizontal cells perform this modulation in the outer plexiform layer.  Cajal distinguished two types of horizontal cells common to birds and reptiles, brush-shaped and stellate (Rodieck, 1973).  In mammals and amphibians, Cajal distinguished two somewhat different classes, the inner and outer horizontal cells.  The number of horizontal cells is mostly consistent across fish, reptiles, birds, and mammals.  However, Cajal noted a correlation of rod structure and density to the size of the horizontal cells (Rodieck, 1973). In mammals and teleost fishes, the rods are somewhat thin and associated with large horizontal cells, while in birds and reptiles the rods tend to be associated with smaller horizontal cells. 

Physiologically, horizontal cells can be divided into two general types (Rodieck, 1973): 
1) those that hyperpolarize or depolarize depending on the stimulating wavelength   (chromaticity or C-type) 
2) those that hyperpolarize to light regardless of wavelength (luminosity or L-type)
The C-type is usually found in species of fish with good color vision, but interestingly, avian and mammalian horizontal cells are of the L-type.  The significance of this difference in fish versus amniotes, who have somewhat comparable color capabilities, is unknown.  One obvious possibility may be that the C-type horizontal cell is specialized to respond to the spectral conditions unique to the underwater environment.

The inner nuclear layer - amacrine cells. Amacrine cells function similarly to horizontal cells in transferring information laterally across the retina. There are over 20 morphological types of amacrine cells, which use at least eight different neurotransmitters (Kandel et al., 1991).  Amacrines may also shape the complex receptive fields seen in some retinal ganglion cells.  They are in a position to modulate antagonistic inputs from bipolar cells to the receptive field surrounds of ganglion cells.  The complexity of ganglion cell receptive fields usually correlates with the amount of amacrine cell input.  The ON-OFF retinal ganglion cells with directional selectivity receive input from amacrine cells (Dowling, 1990).  Amacrine cells are very diverse between amniote forms and even within members of the same species.  In addition, there are regional variations in amacrine cell density.  For example, in the pigeon there are high numbers of amacrine cells found in the red field (dorsotemporal quadrant) of the retina (Nalbach et al., 1993).  The predominance of complex responses and directional selectivity in avian retinal ganglion cells compared to those in mammals may be attributable to amacrine circuitry.

Retinal ganglion cells. There are wide differences in the number, types, and distribution of retinal ganglion cells (RGCs) between birds and primates.  There are 1 million RGCs in rhesus monkey and man compared to 2 million or more in chick, pigeon, and quail (Thompson, 1991).  In general, nocturnal animals and deep-sea fish tend to have lower quantities of ganglion cells, and in other organisms the number of ganglion cells correlates well with the number of cones present in their retina.  The large variation seen in RGCs make them difficult to classify.  Both morphological and physiological criteria for recognized classes have shifted as different variations are discovered and described.  Early classifications by dendritic fields or response types (ON-, OFF-, and ON-OFF) have become uncertain given new data, which has extended the known diversity in their morphology and response properties.  RGCs in birds possess very complex receptive fields; along with the simple luminosity-responsive cells, there are some which exhibit their most vigorous responses to horizontal or vertical stimuli, and stationary or moving edges.  These responses in the bird are similar to those found in the RGCs of reptiles and amphibians (Sillman, 1973).  

In general, there is an inverse correlation between the presumed complexity of visual analysis occurring in the RGC layer and the importance of visual cortex.  It has been suggested that the complexity of RGC receptive fields is related to the extent of binocular processing in the animal.  This relationship is one of reduced RGC response-complexity in those organisms exhibiting larger binocular fields or extensive binocular processing.  For example, owls, with their frontally-placed eyes and highly developed retino-thalamic connections to visual Wulst, have ganglion cells which tend toward simpler ON- or OF-center units (Pettigrew, 1978).  These RGC properties are also found in primate retina.  The relative simplicity of RGC receptive field responses in owls and primates versus that of more lateral-eyed birds and mammals may related to the "correspondence problem."  For stereovision to occur at the retinal level there would have to be a large quantity of ganglion cells to code the enormous number of possible orientation, direction, and size parameters to match before binocular processes continue to telencephalic levels.  The general idea that "complex processing" occurs at the level of the ganglion cells in birds but has been shifted to diencephalic or telencephalic levels in primates may be too gross a generalization.  There are other interpretations of the data relating to binocular vision.  To quote Pettigrew (1991): "Differences in the proportion of the specialized retinal ganglion cells turn out, then, to be quantitative rather than qualitative, and provide the following useful generalization:  the proportion of concentrically organized, non-specialized retinal ganglion cells, as a function of the highly specialized retinal ganglion cells, tends to increase with the importance of binocular vision for the animal." 

It may be that simple-response ganglion cells developed as derivations after the retino-tectal system was in place.  Most of the complex RGC receptive fields that are found in mammals are of retino-collicular cells.  These cells have physiological properties similar to those seen in nonmammalian vertebrates.  In mammals and possibly birds, visual evolution may progress with the addition of specific retino-geniculate ganglion cell types and connections to the existing retino-collicular groups.  

The distribution of RGCs differs widely across organisms.  High packing densities of RGCs are typically correlated with high densities of their associated cones.  These densities, in turn, are assumed to equate with areas of high visual acuity for the organism.  Areas are sites on the retina in which the cones become more densely packed at the expense of rods.  Areas can either be circular (an area centralis) or elongated (a visual streak).  Due to the typical 1:1 ratio of cones to ganglion cells, the density of cones and associated cells in highly developed areas like those in diurnal birds, lizards, and primates, would cause a large bulging of the retinal layer.  This bulging does not occur, however, due to the development of a pit within the area, called a fovea, which is formed by the lateral displacement of the proximal neurons and fibers.  Foveal pits tend to be more pronounced and deeper in diurnal birds in comparison to primates (including humans).  Even in the absence of a distinct pit, many organisms have areas of higher photoreceptor concentrations in certain regions of the retina.   

Foveae in a diurnal sea bird (Arctic Tern; top) & primate (human, bottom).

 

The distribution of RGCs tends to correlate better with behavior and habitat than phylogeny (Thompson, 1991).  Therefore, while RGC distribution may be unreliable in terms of reconstructing phylogenies, there is good evidence that comparisons of RGC densities across species can yield important information regarding the visual worlds of these organisms.  

Regional differences in ganglion cell distributions have typically been associated with where in the visual field important events requiring high detail vision would occur for the organism (Thompson, 1991).  In birds, there can be a single or multiple areas, a horizontal streak, or both.  Owls possess a single fovea, but instead of being located centrally it is more temporally placed.  Birds whose feeding strategies necessitate good speed and distance estimation in daylight (e.g., hawks eagles, and terns feeding "on the wing") have both central and temporal foveae, with the temporal foveae presumably assisting in good stereoscopic depth perception.  Visual streaks are usually associated with the horizontal meridian.  An example of a bird with such a visual streak is the white-capped albatross; this streak may serve to enhance sensitivity to objects on the horizon when flying or feeding over the surface of the water (Sillman, 1973). A more lateral-eyed bird like a pigeon has two areas: 
xxx1) a shallow central fovea, located slightly ventral to the horizontal median and posterior to the vertical median and
xxx2) an area located in the posterior dorsal quadrant which is almost as rich in cells as the fovea itself. 
The central area may be used for monocular viewing of objects in the lateral visual field, while the superior dorsal area may assist in viewing objects binocularly in front of the birds (e.g., in both predatory owls or ground-feeding species like the pigeon).  Most birds (e.g., crows and sparrows) fall into the category of central monofoveal, with a single fovea near the center, slightly above and nasal to the optic disc.  Some birds, such as the domestic chicken and California quail, do not appear to have a fovea (Sillman, 1973).  

Comparing Avian and Primate Retinae  

Is the avian retina "special?"  The avian eye possess structures not found in the eye of primates. When studying the avian visual system, it is natural to make comparisons between it and the visual system in humans and non-human primates.  The presence of only two basic photoreceptor types (single rods and cones) and lack of oil droplets in placental mammals may lead to the impression that birds have an unusally complex retinal structure compared to other organisms, including mammals. This is not the case; the avian retina (with a few exceptions) is very similar to that of modern reptiles. In terms of the comparative anatomy of the retina, it is the placental mammals, and especially the primates, who are the "odd ones" among the vertebrates.  

Mammalian retina. The differences in the eye and retina among the three mammalian classifications are instructive in determining the overall course of visual system evolution.  Given the separate lineages and evolutionary history of monotremes, marsupials, and placentals, one cannot consider modern monotremes or marsupials as an early stage of the placentals.  However, given the similarities of the monotreme eye with the reptilian, it could be argued that in at least their visual system they are sufficiently conserved over their evolution from their reptilian forebears to make useful comparisons against the marsupials and placental mammals. 

The monotreme eye, with the exception of the oculo-rotary muscles and the ciliary body, is very similar to that of a reptiles and birds.  So similar, in fact, that Walls (1942) suggested that it might easily be mistaken for the eye of a nocturnal reptile.  The monotreme retina extends farther temporally than nasally, perhaps suggesting the importance of the binocular field (Walls, 1942).  The echidna has a pure rod retina, with three outer layers of nuclei and two inner layers.  There is a single row of scattered ganglion cells.  According to Walls (1942) the photoreceptors are probably derivatives of those in sauropsida.  In the platypus the single cones and double cones still exhibit oil-droplets, but these are mostly colorless.  The rod and cone nuclei are not well differentiated, both being cone like.  The absence of cones in the echidna and their simplification in the platypus are likely the consequence of adaptation to a nocturnal (or at least dim-light) lifestyle.  

The marsupial eye seems to possess both mammalian and reptilian characteristics.  Walls (1942) characterized the marsupials (e.g., opposums, kangaroos) as having mammalian eyes (i.e., eye shape, lens, etc.) but a more reptilian retina.  This reptile-like retina is similar to that found in monotremes.  The photoreceptors are similar to those in the platypus, with both single cones and double cones with oil droplets, and slender rods.  The single and double cones are very similar to their corresponding types in reptiles. 

The placental mammals seem to have "lost" the double cones and oil droplets characteristic of their reptilian forebears.  Most placental mammals have duplex retinae, and generally the placental retina is simpler than that of other amniotes. Many placental mammals appear to have no cones (e.g., armadillo, bats, hedgehogs); those that do possess only single cones without oil-droplets.  No placental is known to have double cones or oil droplets in its single cones.  The ganglion cells usually form one layer, except in an area centralis or in the fovea, such as that found in primates.  It would appear that the stem ancestral placental mammal must have lost the double cones and oil droplets, undergoing a kind of simplification, at least of the retinal structures.  Walls (1942) presumes, based on the transition from striated to unstriated intra-ocular muscles in the reptile-monotreme transition, that accommodation became unimportant for early mammals. This may be because it was never necessary for them to close the pupil quickly, implying that they operated primarily under scotopic and/or nocturnal conditions.  It is commonly held that at some point in their evolution, the mammals went through a period of nocturnal activity, probably coinciding with the ascent of the dinosaurs.

Bottleneck Theory of Mammalian Evolution
 

"The eye of man, with its pretty good accommodation, its fovea, its miscellaneous yellow filters, and its capacity for color vision, possesses in substantial degree the physiological capabilities of  the standard sauropsidan eye as we see it in the lizard or the bird.  But it has gained these powers  through a lengthy process of re-differentiation, which was carried out largely within the confines of the primate order itself." 

Walls (1942); reprinted in Cronly-Dillon & Gregory (1991), pg. 478

The "Bottleneck Theory" proposed by Walls (1942) asserts that during the rule of the dinosaurs, the ancestral mammals were forced out of a diurnal lifestyle to a nocturnal one.  The ancestral mammal probably was a small-eyed, nocturnal insectivore (and an accompanying rod-dominated retina is assume.  These mammals presumably possessed the double cones and oil droplets inherited from their reptilian ancestors.  This includes a well-developed collothalamic visual system.  During the "Age of Reptiles", with most of the land areas inhabited by numerous carnivorous dinosaurs.  The mammals were forced by predation pressures to adapt a more nocturnal lifestyle.   During this period, the cones underwent simplification, resulting in a predominantly rod-containing retina.  The oil droplets disappeared, or became colorless, since scotopic conditions prevailed for these forms.  The early placentals had large pupils, simple large lenses with unstriated intra-ocular muscles which did not support accommodation.  Placentilian retina has single cones with no oil droplets or other special features found in other amniote cones.  In essence, the placental mammals have cones reduced to their simplest form (Walls, 1942).  Only in the primates do the cones exhibit such a sophisticated capacity for color vision.  If the duplex retina persisted throughout the placental line early on, then one would think that all mammals, and not just the simians, would have retained a complete color vision system like that found in birds.  It is more likely that at one time all living placentals had rod-only retinae and that subsequent placentals evolved duplex retina from pure rod ones.  

After mammals escaped the evolutionary bottleneck, they reacquired cones and the capacity for color vision.  After the mass extinction of the dinosaurs at the Cretaceous-Tertiary boundary (the KT boundary, about 65 MYA), mammals "passed through" the evolutionary bottleneck.   Released from the predation pressures of the dinosaurs, they were relatively free to rapidly develop and expand into various new ecological niches that were opened up for them.  The return to diurnal niches of some forms led to the reemergence of color vision.  Color vision appears to have evolved twice separately in the mammalian line with the diurnal squirrels and diurnal primates.  

Support for the bottleneck theory comes from analysis of the visual system in snakes.  Perhaps surprisingly, support for this theory comes from comparing the visual systems of placental mammals and snakes.  Evidence suggests that snakes arose from diurnal lizards which adapted a burrowing lifestyle and subsequently underwent the atrophy and eventual loss of the limbs.  Both the snakes and placental mammals went from being nocturnal to diurnal (snakes lost their legs, went burrowing, then came back up).  The eye of placental mammals and snakes is more spherical than birds and other reptiles.  Their spherical eye possesses a lens which acquired a yellow tint, substituting for the yellow oil drops lost from a reptile ancestor.  During their burrowing period, the snake eye lost their "reptilian" oil droplets and double cones.  Upon re-emergence above ground, the early snakes (Boidae, including modern pythons and boas) gained rods containing rhodopsin and single cones (Walls, 1942).   In a sense, the snake visual system "devolved", and it was reinvented, as in the placental mammals.  

To summarize, the retina of reptiles and birds might have evolved differently from the primates primarily because of the period of nocturnal existence that early mammals underwent.  One can look to the time when dinosaurs ruled as one of the pivotal points in the evolution of the mammalian eye and retina.  During this period, the birds maintained a predominantly diurnal lifestyle, relatively safe from predation pressures from the dinosaurs because of their ability to fly.  From the early placental mammals, the primates eventually evolved superior color and object recognition abilities.  

Next Section: Taking Flight: Post-Retinal Processing