A colour to birds and to humans: why is it so different? [accepted version]

Visual Modelling: the New Standard

It has started with a paper by Vorobyev and Osorio (1998), which formulates a mathematical framework to approximate a colour that an animal can perceive, synthesizing the knowledge of eye anatomy and physiology with that of neural processing of colours, based on the multidimensional chromaticity model by Goldsmith (1990), the so-called visual model. Since then, studies based on this framework have been increasing year by year (Fig. 1), helped by Vorobyev et al. (1998) and Endler and Mielke (2005), in which more practical applications of the visual model are proposed. Nowadays, the visual model has become more or less a standard for quantifying colours to birds. Here, I review the biological bases of the importance of visual modelling to most ornithologists, focusing on the causes of the difference in colours to birds and to humans.

What is Colour?

Colour is a sense that we have in our brain, which is formed as a ratio of multiple neural stimuli from retinal photoreceptor cells, processed by and inputted through retinal neurons. What animals perceive in this process is also called hue, a chromatic aspect of colours (Endler and Mielke 2005; Vorobyev et al. 1998). The sense of colour is widespread in animals with eyes, but the type of colour vision differs between them (Bowmaker 2008; Collin et al. 2009). I first define the type of colour vision, and then illustrate what determines the difference in perceived colours between birds and humans in terms of both proximate and ultimate causes.

The photoreceptor cells responsible for colour processing are called cones or single-cones, besides the rod cell that is responsible for twilight vision. The number of single-cone types on the retina differs between animals, which should render types of colour vision different between them. The known range of this number is from 1 to 4, each referred to as monochromacy (1 cone cell type), dichromacy (2 types), trichromacy (3 types), and tetrachromacy (4 types) (Vorobyev et al. 1998). This definition, however, should be treated carefully because all types of cone cells in the eye are not always used for colour processing (Koshitaka et al. 2009), and because few studies demonstrate that birds are in fact tetrachromats (Osorio et al. 1999). Nevertheless, at least in birds, despite few verification by behavioural experiments (Maier 1992; Remy and Emmerton 1989), a number of microspectroscopic studies demonstrate that birds have 4 types of single-cones (e.g., Wright and Bowmaker 2001; Hart 2002, 2004), which are probably used for colour processing (Osorio et al. 1999; Vorobyev and Osorio 1998).

Causes of the Difference between Birds and Humans

Proximate Causes

The bird-human colour difference is partly attributed to the difference in photoreceptor sensitivity (Vorobyev and Osorio 1998; Endler and Mielke 2005). On the outer segment of photoreceptor cells is a visual pigment, which is composed of a kind of vitamin A, retinal, and a protein, so-called photopsin (a kind of vertebrate opsins) (Collin et al. 2009; Hart and Hunt 2007). When a photon of a certain wavelength hits the visual pigment, the retinal is released to trigger a chemical response to activate neurons attached to the photoreceptor cell with synapses. A number of variants of photopsins are known, each of which differs in the wavelength of photons that respective visual pigments can capture (Collin et al. 2009; Shi and Yokoyama 2003). Fig. 2 shows the difference in the sensitivity of cone photoreceptors between humans and birds of several kinds. Although UVS (ultraviolet-sensitive; Hart et al. 2000) and VS (violet-sensitive; Hart 2002, 2004; Hart and Vorobyev 2005) cones of birds as well as the SWS (short-wavelength-sensitive) cone of humans are variants of the same photopsin, SWS1, the sensitivity is different between each other (Collin et al. 2009; Shi and Yokoyama 2003).

Eye organs that transmit light, i.e., ocular media, also play substantial roles in differentiating the perceived colours by birds and by humans. First, there are some mechanical differences between birds and mammals found in the cornea, one of ocular media. Cornea is composed of collagen fibres. By the collagen nanostructure, the avian cornea let ultraviolet light pass through, while the mammalian cornea blocks the ultraviolet light (Tsukahara et al. 2010). This difference probably causes the difference in the ultraviolet sensitivity between birds and humans: although the SWS photoreceptor of humans has sensitivity similar to that of peafowls’ VS photoreceptor (the dotted line in Fig. 2a) (Bowmaker and Dartall 1980; Hart 2002), humans cannot see ultraviolet light, i.e., the range below ca. 400 nm of wavelengths (dashed lines in Fig. 2) (Surrridge et al. 2003). This is supported by the fact that people who don’t have cornea, i.e., aphakia, are known to have slight UV sensitivity (Griswold and Stark 1992).

A difference also exists in the cornea at the physiological level: a protein called ALDH3A1 absorbs ultraviolet light in mammalian cornea (Estey et al. 2007), while a protein called apolipoprotein is found in avian cornea instead of the ALDH3A1, which removes lipids that capture free radicals caused by ultraviolet radiation (Tsukahara et al. 2011). These findings suggest that birds have coping mechanisms against ultraviolet light and its influence inside the eye, while mammals’ eye keeps UV out with double lines of defence. Nevertheless, visual sensitivity to UV and its usage in mammals remains largely unknown: the SWS1 photopsin of mice and rats is highly sensitive to UV, with a λ_max, wavelength for the peak sensitivity, of about 359 nm (Shi and Yokoyam 2003). Also, reindeer are thought to have UV sensitivity (Hogg et al. 2011). Moreover, recent work demonstrates that cornea of a number of species of mammals transmit UV (Douglas and Jeffery 2014). More comprehensive work is necessary to know how mammals sense and use ultraviolet light as a whole.

Birds have two other structures that all eutherian mammals lack: oil droplets and double-cones. Oil droplets are located on the basement of the cone outer segment, and function as photon filters to reduce noises (Hart and Vorobyev 2005). Double-cones are a kind of cone cells, but differ form the single-cones in the structure, which are composed of two fused cone cells as the name suggests (Goldsmith 1990; Collin et al. 2009; Osorio et al. 1999). They outnumber all the single-cone types in bird retina (Hart 2001), and are thought to be responsible for various functions in achromatic, i.e., colour-independent, visual perception, e.g., texture perception, motion detection, and perception of lightness (Osorio et al. 1999). The last function is different from that of humans, as humans perceive lightness with stimuli of combined inputs from both M and L cones (Surridge et al. 2003).

The greatest difference between birds and humans is the number of types of the cone photoreceptor cell on the retina: according to the definition above, birds have tetrachromacy, i.e., having 4 types of cones, whilst humans basically have trichromacy, with 3 cone types, namely, lacking VS/UVS cones. We have so-called three primary colours owing to the trichromacy, unlike most mammals that have dichromacy. Such a difference between respective vision types is a dimensional difference, strictly in a mathematical sense (Fig. 3), as every colour has a set of components of the number of the photoreceptor types that are used for colour processing in respective animals. For example, considering the state of a colour in number, grey for humans is composed of three elements, which are 0.333:0.333:0.333, while grey for birds is composed of four elements, 0.25:0.25:0.25:0.25, even though both sum up to 1.

Ultimate Causes

The difference in perceived colours between birds and humans should have been affected strongly by the evolution of colour vision types in vertebrates. The colour vision type of extant vertebrates is diverse across taxa: tetrachromacy is found in lampreys, teleost fish, reptiles and birds; trichromacy in amphibians (Collin et al. 2009; Siddiqi et al. 2004), marsupials (Arrese et al. 2002), and Simiiformes (New, Old World monkeys, and apes) (Surridge et al. 2003); dichromacy in most eutherian mammals and monotremes (Jacobs 2009); and monochromacy in sharks (Hart et al. 2011; but see Bowmaker 2008; Collin et al. 2009). Therefore, a complex evolution of the vision type likely occurred in the entire Vertebrata.

Even in Mammalia, two vision types are found: trichromacy in marsupials (Arrese et al. 2002; Jacobs 2009) and Simiiformes except in males of New World monkeys (Platyrrhini) (Surridge et al. 2003); and dichromacy in monotremes and other eutherians (Jacobs 2009). Considering the type of photopsins (Fig. 4), each colour vision type probably has an independent evolutionary origin. In particular, the trichromacy of humans and primates certainly emerged in the ancestral Simiiform primate (New, Old World monkeys, and apes), reacquiring the MWS photoreceptor as a result of mutation in the LWS gene on the X-chromosome. This gene was duplicated only in Catarrhini (Old World monkeys and apes), and because of the lack of duplication, Platyrrhini have an intraspecific sex-linked polymorphism (Jacobs 1996): only females are diploids in terms of that locus, and thus females are either dichromats (i.e., homozygotes) or trichromats (heterozygotes), while males are always dichromats because of the haploidy of the X-chromosome. Such a recent origin of the MWS photoreceptor is strongly related to the great overlap in MWS and LWS sensitivity in the human (and primate) colour vision (Fig. 2a), and probably to achromatopsia in humans (Mollon 1989).

Here, I illustrate 2 mutually exclusive scenarios to explain the evolution of colour vision in vertebrates, which are simplest and at the extreme opposite: one assumes monochromacy at the origin, while the other tetrachromacy (Fig. 4). Putting the differentiation of the LWS photopsins in Simiiformes aside, it is more probable, parsimoniously speaking, that the ancestor of the entire Vertebrata was a tetrachromat than a monochromat. When assuming the monochromacy as the very ancestral character, photoreceptor types should have been acquired totally 14-16 times independently. On the contrary, assuming the tetrachromacy as the origin, 6-8 loss-of-function events are only required to explain the extant distribution of vision types in Vertebrata. The increase of a photoreceptor type, hence rising of the dimensionality of colour vision, should be much more difficult to occur than the decrease, as it should involve changes not only in the spectral sensitivity but also in the neural networks. Indeed, this might not be the most plausible scenario, and I would admit that the hypotheses are too simple. Nevertheless, it is probable that the very ancestor of the entire Vertebrata had several types of single-cones rather than just a single type (Bowmaker 2008; Collin et al. 2009).

The reduced number of photoreceptor types in mammals could be attributed to the nocturnal way of life probably in the ancestor, and in the vast majority of extant species (Jacobs 2009). By contrast, the acquisition and the prevalence of the trichromacy in catarrhine monkeys could have coincided with a change from nocturnal to diurnal living (Surridge et al. 2003). Overall, the trichromacy of primates has been reacquired recently in vertebrate evolution, and the colour vision of humans is still moderate or even weak among vertebrates. On the contrary, the tetrachromacy of birds, which is of a higher dimension than trichromacy, could have been retained throughout the long history of vertebrate evolution. Therefore, ignoring these differences in the analyses should falsify the results, and I should call it an artefact.

Application in Visual Modelling

The above-mentioned proximate characters of eyes are all incorporated in the visual model (Endler and Mielke 2005; Vorobyev and Osorio 1998). The visual model is designed to infer discriminability between 2 given colours, namely, to estimate how different the 2 colours are in a unit discrimination threshold, i.e., just noticeable difference, or JND, a psychophysical measure of difference based on the Weber-Fechner law. The actual set of data required is composed of reflectance spectra of light from the objects in interest on wavelengths from 300 nm to 700 nm, which are usually collected with a spectrophotometer or extracted from digital camera image files (Stevens et al. 2007). The first step is to know how much is the portion of light that each photoreceptor type receives, the so-called photon capture (also referred to as cone catch, quantum catch, etc.). Photon capture is the integrated product of a measured reflectance spectrum and the sensitivity of a cone type, which is modified with reflectance, transmittance, and absorbance of ocular media above mentioned (Hart and Vorobyev 2005). The error rate, i.e., noise-to-signal ratio or Weber fraction, is associated with the abundance of respective cone types on the retina (Vorobyev et al. 1998), on the assumption that the input signal from a cone type should be more accurate if the cone type is denser on the retina, and vice versa. The difference in vision types is indeed treated as a dimensional difference.

Despite the establishment of the framework, there are still assumptions in the visual model. First, although differences greater than 1 JND should by definition be discriminable to subject animals in the Weber-Fechner law, this measure is not always true, particularly in conditions with less ambient light (Olsson et al. 2015; Siddiqi et al. 2004). By convention, 3 JND is treated as a sufficient threshold for discrimination in practice by subject animals (Siddiqi et al. 2004; Stevens et al. 2013). Moreover, it is even possible that the effect of noises might vary between photoreceptor types (Angueyra and Rieke 2013; Olsson et al. 2015). Nevertheless, such information is usually unavailable for wild birds, and thus all we could do is pay attention on them in application.

Another notice on practical application is that our knowledge about bird vision is still very limited. For example, visual cortices of the brain are known to play a considerable role in colour discrimination in primates (Shapley and Hawken 2011), but such information lacks in birds. To know more properly the colour that birds see, we still have to wait for future advancements in psychophysical and neurophysiological studies about colour processing mechanisms in birds.

Statistical Problems and Solutions

So far, it is demonstrated that the visual model is a strong tool to quantify a colour perceived by non-human animals with least artefacts. However, there remains an issue to mention: statistics. We need to be careful to analyze the output data of the visual model because their structure is different from what we usually face in most ecological studies. In this final section, I illustrate the characteristics of visual model outputs, and propose a simple solution, in which several traditional methods are used in combination.

Similarity and Congruence Principle

Perception of a colour is strongly affected by the colour of its background in the sight of viewers, and thus the contrast against the background is key to perception of colours. This is related to a mechanism called colour constancy to keep the colour of the same object constant under different ambient light, which is widespread in colour sensing animals (e.g., Koshitaka et al. 2008). The colour constancy would be achieved because the difference in irradiated light would be ignored by the neural system based on the reflection from the background, as the object and the background are mostly in the same light environment. This also means that it would be difficult for an animal to perceive the colour of objects alone without referring to another colour. Understandably, any analytical methods must conform to this manner of colour perception, as done by the visual model.

The problem is, however, the nature of the data structure. First, as long as being quantified in the colour space, respective colours are in trigonometric relationships with each other, and thus geometrically mutually dependent (Endler and Mielke 2005). Moreover, the output of the visual model is the measure of approximated similarity between a given pair of colours (Vorobyev and Osorio 1998), not simply the position of a colour in the colour space (Endler and Mielke 2005; Vorobyev et al. 1998). Data with such dependence could not be dealt with in the paradigm of linear models, or even that of non-parametric tests because such dependence is not postulated in these procedures (Biondini et al. 1991; Endler and Mielke 2005). This is what we call the congruence principle.

Few methods can cope with this issue, but each has drawbacks. First, there are the Mantel and the permutation tests: both estimate correlation between 2 datasets, in each of which data scatter multidimensionally. However, due to the nature of correlational analyses, it is difficult to reach at the causal relationship with these methods (reviewed in Tanaka et al. in press). Next, Endler and Mielke (2005) propose a very fine method for analysis, called LSED-MRPP (LSED: least sum of Euclidean distance; MRPP: multi-response permutation procedure) based on the paradigm of the permutation test. However, it appears not to have prevailed very much, probably because of the complexity and the difficulty in practical uses. There might be some alternatives, but probably with little accessibility. To sum, some studies seemingly violate statistical principles (reviewed in Tanaka et al. in press), while statistical procedures are not described precisely in many studies.

Multidimensional Scaling

I would suggest here to adopt the multidimensional scaling (MDS), or the principal coordinate analysis (PCoA; Legendre and Legendre 1998), which is formulated to transform a set of geometric (Euclidean) distances or measures of similarity to a set of coordinates (principal coordinate). Transforming distance/similarity matrices to coordinates is particularly important in visual modelling because it liberates the geometric dependence between respective measured colours. Hereafter, one can avoid violating the congruence principle, and thus adopt linear models or non-parametric tests for analysis, which should be more common in ecology. Because of this, the MDS is quite suitable for the visual model, and in fact, it is adopted in visual modelling of colours perceived by bees (Backhaus et al. 1987).

In short, the procedure I propose is first to calculate pair-wise JNDs between all possible combinations of measured colours in a form of distance matrix (Fig. 5a), and next, to analyze the produced principal coordinates by the MDS with linear models. The output principal coordinate of the MDS would represent the (multidimensional) position of each measured colour in the unit of approximate JND. As the example, I reanalyzed the dataset of Tanaka et al. (2011), in which colours to birds between the gape, the wing-patch of Horsfield’s hawk-cuckoo chicks (see Tanaka et al. 2011; Tanaka and Ueda 2005 for detail), and the gape of host chicks were compared without the MDS.

The results are as follows: cuckoo gape was treated as the intercept (linear coefficient = 2.82 ± 1.48 SE) with the rest two as dummy variables (Agresti 2002), and from cuckoo gape, host gape was significantly different (partial coefficient = -6.05 ± 1.83, x²₁ = 10.9, P = 0.001), while cuckoo patch was not (partial coefficient = -0.38 ± 1.44, x²₁ = 0.068, P = 0.79). These partial coefficients represent the estimated differences from the intercept, i.e., effect size, in the unit approximate JND. Since the sum of each principal coordinate (i.e., centroid) is always 0, the absolute value of the partial coefficient roughly represents the JND of the focal colour against the colour that is set as the intercept, i.e., cuckoo gape in this case. The estimated JND for host gape against cuckoo gape was thus 6.05 ± 1.83, while the one for cuckoo patch was 0.38 ± 1.44. Therefore, the colour of the wing-patch of hawk-cuckoo chicks would not be discriminable to host parents from that of the gape, while the colour of the gape of host chicks would be enough discriminable from the colour of the gape of hawk-cuckoo chicks. The conclusion does not diverge from that in Tanaka et al. (2011), but one can assess the effect more directly and robustly with this procedure.

The versatility of the MDS in visual modelling is not limited to the availability of variance for JNDs. First, it can easily be applied to other measures of colour difference, such as colour contrast, i.e., Euclidean distance between a given pair of colours in the colour space (∆_T in Endler and Mielke 2005). Next, the dimensionality of the dataset is controllable: the number of output coordinates, i.e., dimension, can be changed in the MDS by setting the dimension parameter, e.g., k in the cmdscale command in R (R Core Team 2014). As a result, variance can be divided into those in respective dimensions; e.g., first and second principal coordinates of the hawk-cuckoo dataset have respective variances (Fig. 5a). This also helps examine how good is the approximation of JNDs by the first principal coordinate. In this instance, the first principal coordinate explained 76% of the total variance with the eigenvalue of 2.28, and thus it represented JNDs well. Finally, the estimated JND for hue can be integrated with that for luminance (perceived lightness), approximated from double-cone photon captures, through the principal component analysis (PCA). This integration would be similar to what is done with the LSED-MRPP by Endler and Mielke (2005), and might better represent the colour that birds actually see than without integration, on the assumption that these chromatic and achromatic aspects of colours are integrated in their colour processing system (Osorio et al. 1999). Practical applications of this method and details are described and discussed in Tanaka et al. (in press).

Conclusion

There must be a huge biological difference between colours that birds see and that humans see, which is not only in the sensitivity to a particular type of light but also in the dimension of the colour vision system, which stem from the long evolutionary history of vertebrates. The visual model could circumvent the effect of such differences, but one must be careful of not only the application but also the way to analyze the output of the visual model.

Acknowledgements

I thank the speakers in the symposium at IOC26, W Kitamura, G Morimoto, T Yamasaki, and particularly M Stevens, the keynote speaker, for their contribution. I thank members of the Scientific Programme Committee for their editorial support, and two anonymous referees for their very helpful comments. I thank N Hart for providing photoreceptor sensitivity data. I received financial supports from Grants-in-Aid by the Japan Society for the Promotion of Science grant no. 24770028 (Young Scientists B) and 23255004 (Basic A).

References

• Agresti A (2002) An Introduction to Categorical Data Analysis. Wiley Inter-Science, Hoboken, NJ, USA

• Angueyra JM, Rieke F (2013) Origin and effect of phototransduction noise in primate cone photoreceptors. Nat Neurosci 16: 1692-1700

• Arrese CA, Hart NS, Thomas N, Beazley LD, Shand J (2002) Trichromacy in Australian marsupials. Curr Biol 12: 657-660

• Backhaus W, Menzel R, Kreißl S (1987) Multidimensional scaling of color similarity in bees. Biol Cybern 56:293-304

• Biondini ME, Mielke PW Jr, Redente EF (1991) Permutation techniques based on Euclidean analysis spaces: a new and powerful statistical method for ecological researches. In: Feoli F, Orlóci L (eds) Computer Assisted Vegetation Analysis. Kluwer Academic Publisher, Dordrecht, pp 221-240

• Bowmaker JK, Dartnall HJA (1980) Visual pigments of rods and cones in a human retina. J Physiol 298: 501-511

• Bowmaker JK (2008) Evolution of vertebrate visual pigments. Vision Res 48: 2022-2041

• Collin SP, Davies WL, Hart NS, Hunt DM (2009) The evolution of early vertebrate photoreceptors. Phil Trans R Soc B 364: 2925-2940

• Douglas RH, Jeffery G. 2014. The spectral transmission of ocular media suggests ultraviolet sensitivity is widespread among mammals. Proc R Soc Lond B 281: 20132995

• Endler JA, Mielke PW Jr (2005) Comparing entire colour patterns as birds see them. Biol J Linn Soc 86: 405-431

• Estey T, Piatigorsky J, Lassen N, Vasiliou V (2007) ALDH3A1: a cornel crystalin with diverse functions. Exp Eye Res 84: 3-12

• Goldsmith TH (1990) Optimization, constraint, and history in the evolution of eyes. Q Rev Biol 65: 281-322

• Griswold MS, Stark WS (1992) Scotopic spectral sensitivity of phakic and aphakic observers extending into the near ultraviolet. Vision Res 32: 1739-1743

• Hart NS (2001) Variations in cone photoreceptor abundance and the visual ecology of birds. J Comp Physiol A 187: 685-698

• Hart NS (2002) Vision in the peafowl (Aves: Pavo cristatus). J Exp Biol 205: 3925-3935

• Hart NS (2004) Microspectrophotometry of visual pigments and oil droplets in a marine bird, the wedge-tailed shearwater Puffinus pacificus: topographic variations in photoreceptor spectral characteristics. J Exp Biol 207: 1229-1240

• Hart NS, Hunt DM (2007) Avian visual pigments: characteristics, spectral tuning and evolution. Am Nat 169: S7-S26Hart NS, Partridge JC, Cuthill IC, Bennett ATD (2000) Visual pigments, oil droplets, ocular media and cone photoreceptor distribution in two species of passerine bird: the blue tit (Parus caeruleus L.) and the blackbird (Turdus merula L.). J Comp Physiol A 186: 375-387

• Hart NS, Theiss SM, Harahush BK, Collin SP (2011) Microspectorphotometric evidence for cone monochromacy in sharks. Naturwissenschaften 98: 193-201

• Hart NS, Vorobyev M (2005) Modelling oil droplet absorption spectra and spectral sensitivities of bird cone photoreceptors. J Comp Physiol A 191: 381-392

• Hogg C, Neveu M, Stokkan K-A, Folkow L, Cottrill P, Douglas R, Hunt DM, Jeffrey G (2011) Arctic reindeer extend their visual range into the ultraviolet. J Exp Biol 214: 2014-2016

• Jacobs GH (1996) Primate photopigments and primate color vision. Proc Natl Acad Sci USA 93: 577-581

• Jacobs GH (2009) Evolution of colour vision in mammals. Phil Trans R Soc B 364: 2957-2967

• Koshitaka H, Kinoshita M, Vorobyev M, Arikawa K (2008) Tetrachromacy in a butterfly that has eight varieties of spectral receptors. Proc R Soc Lon B 275: 947-954

• Legendre P, Legendre L (1998) Numerical ecology, second English ed Elsevier, Amsterdam

• Maier EJ (1992) Spectral sensitivities including the ultraviolet of the passeirform bird Leiothrix lutea. J Comp Physiol A 170: 709-714

• Mollon JD (1989) “Tho’ she kneel’d in that place where they grew...” The uses and origins of primate colour vision. J Exp Biol 146: 21-38

• Osorio D, Vorobyev M, Jones CD (1999) Colour vision of domestic chicks. J Exp Biol 202: 2951-2959

• Osorio D, Miklósi A, Gonda Zs (1999) Visual ecology and perception of coloration patterns by domestic chicks. Evol Ecol 13: 673-689

• R Core Team (2014) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. <http://www.R-project.org/>

• Shapley R, Hawken MJ (2011) Color in the cortex: single- and double-opponent cells. Vision Res 51: 701-717

• Shi Y, Yokoyama S (2003) Molecular analysis of the evolutionary significance of ultraviolet vision in vertebrates. Proc Natl Acad Sci USA 100: 8308-8313

• Siddiqi A, Cronin TW, Loew ER, Vorobyev M, Summers K (2004) Interspecific and intraspecific views of color signals in the strawberry poison frog Dendrobates pumilio. J Exp Biol 207: 2471-2485

• Stevens M, Párraga CA, Cuthill IC, Partridge JC, Troscianko TS (2007) Using digital photography to study animal coloration. Biol J Linn Soc 90: 211-237

• Stevens M, Troscianko J, Spottiswoode CN (2013) Repeated targeting of the same hosts by a brood parasite compromises host egg rejection. Nat Comm 4: 2475

• Surridge AK, Osorio D, Mundy NI (2003) Evolution and selection of trichromatic vision in primates. Trends Ecol Evol 18: 198-205

• Tanaka KD, Denda T, Ueda K, Emura N (in press) Fruit colours conceals endocarp dimorphism from avian seed dispersers in a tropical beach plant, Scaevola taccada (Goodeniaceae), found in Okinawa. J Trop Ecol [DOI: 10.1017/S0266467415000218]

• Tanaka KD, Morimoto G, Stevens M, Ueda K (2011) Rethinking visual supernormal stimuli in cuckoos: visual modeling of host and parasite signals. Behav Ecol 22: 1012-1019

• Tanaka KD, Ueda K (2005) Horsfield’s hawk-cuckoo nestlings simulate multiple gapes for begging. Science 308: 653

• Tsukahara N, Tani Y, Lee E, Kikuchi H, Endoh K, Ichikawa M, Sugita S (2010) Microstructure characteristics of the cornea in birds and mammals. J Vet Med Sci 72: 1137-1143

• Tsukahara N, Tani Y, Nihei K, Kabuyama Y, Sugita S (2011) High levels of apolipoproteins found in the soluble fraction of avian cornea. Exp Eye Res 92: 432-435

• Vorobyev M, Osorio D (1998) Receptor noise as a determinant of colour thresholds. Proc R Soc Lond B 265: 351-358

• Vorobyev M, Osorio D, Bennett ATD, Marshall NJ, Cuthill IC (1998) Tetrachromacy, oil droplets and bird plumage colours. J Comp Physiol A 183: 621-633

• Wright MW, Bowmaker JK (2001) Retinal photoreceptors of paleognathous birds: the ostrich (Struthio camelus) and rhea (Rhea americana). Vision Res 41: 1-12