Fruit colour conceals endocarp dimorphism from avian seed dispersers in a tropical beach plant, Scaevola taccada (Goodeniaceae), found in Okinawa [accepted version]

INTRODUCTION

Biological communication signals can be mutualistic or parasitic, depending on whether the receivers can gain benefits by responding to the signals (honest signalling) or the signals exist at the expense of the receiver (deceptive or dishonest signalling) (Foster & Delay 1998, Galetti 2002, Stuart-Fox 2005). However, theory predicts that both honest and dishonest signalling strategies will not always be evolutionarily stable in interspecific communication (Freckleton & Côté 2003). Thus, mutualists can become parasitic and vice versa, even in the context of plant-animal interactions: an individual plant may reduce the level of reward to conserve resources without indicating the change in rewards to mutualistic agents. Despite the theoretical prediction, it would be difficult to demonstrate such signalling bimodality in wild animal-plant systems because the transition from mutualistic to parasitic state is expected to be subtle and multimodal (Edwards & Yu 2007).

Scaevola taccada is a widely distributing tropical beach plant, whose fruits have an endocarp composed of a cork substrate around the seed, facilitating flotation (Emura et al. 2014). It is likely that S. taccada seeds are dispersed by ocean currents as well as by frugivorous birds (Emura et al. 2012, Howarth et al. 2003). However, Emura et al. (2014) discovered a cryptic dimorphism in the endocarp of this plant: although most plants bear fruit with the cork endocarp (C+), fruits of some individuals consistently lack such a structure (C–), even though both types of fruit have sugar-containing flesh around the seed. It is probable that these morphotypes have adapted alternatively: plants with the C+ fruit disperse seeds broadly through ocean currents (along with avian dispersal), while those with the C– fruit remain relatively localized through avian dispersal.

Despite the difference in assumed advantages of the endocarp types, the fruit colour seems not to differ between them, at least in the human vision (Emura et al. 2014). The lack of colour differentiation indicates that the fruit colour of C+ plants might relatively be a dishonest signal because of the apparently reduced nutritional contents and increased indigestible substrates. However, whether the fruit colour is actually indiscernible to seed dispersers is unclear, since birds have higher sensitivity to colours than mammals including humans (Endler & Mielke 2005, Tanaka in press, Vorobyev & Osorio 1998). Therefore, the fruit colour in bird vision must be examined to determine whether seed-dispersing birds can distinguish between the colours of C+ fruits and C– fruits.

We focused on the dissimilarity in colour between fruits and leaves, as well as fruits of S. taccada morphotypes. The contrast in colour between fruits and leaves should be a key attractant for seed-dispersing birds, because perceived colour of an object is in principle its contrast against the background colour (Koshitaka et al. 2008, Osorio & Vorobyev 1996, Vorobyev & Osorio 1998). To avoid C+ fruits, seed-dispersing birds should memorise the fruit colour against the leaf background (i.e. fruit-leaf contrast) because the endocarp morphotypes are consistent within individual plants, and birds should move to another plant to encounter fruits of the C– morphotype. Therefore, great dissimilarity in fruit-leaf contrast between morphotypes is essential for S. taccada to establish an honest signal. We hypothesised that the fruit colour of S. taccada should be a dishonest signal if the colour contrasts of fruits against the leaf background are indistinguishable between morphotypes in the vision of potential seed-dispersing birds, and tested it through the avian visual model.

MATERIALS AND METHODS

Study species

Scaevola taccada (Gaertn.) Roxb. is a coastal shrub, distributing widely in tropical and subtropical Indo-Pacific coastal lands, from East Africa to Oceania and Japan. The cork substrate of the endocarp facilitates floatation, as fruits with the cork endocarp float for more than 6 mo while fruits without cork sink relatively rapidly, i.e. within 1 wk (Emura et al. 2014). The endocarp dimorphism might serve as an alternative strategy of seed dispersal within the same species, since a considerable rate of gene flow between individuals of the two morphs has been detected (Emura et al., unpubl. data).

The size of the indigestible part of C+ fruits is 6.61 mm in diameter, whereas the one for C– is 3.93 mm due to the lack of cork endocarp (Emura et al. 2014). This size difference results not only in a 10% reduction in the volume of digestive pulp in C+ fruits, but also in an almost five times greater volume of the indigestible part of C+ than that of C–, 0.15 ml for C+ and 0.031 ml for C–, assuming spherical shapes (i.e.

where r =

). Moreover, the pulp of C+ fruits contains more water than that of C– (Emura et al., unpubl. data), suggesting a much lower nutritional reward to seed dispersers than the one expected by the structural difference between morphotypes. Therefore, the total amount of reward a C+ fruit provides to birds should be less than that of the C– type, and simultaneously, the amount of indigestible substrate should be greater despite the lack of difference in external sizes (Emura et al. 2014).

.

The single- and double-cone sensitivity of the blue tit Cyanistes caeruleus (Hart et al. 2000) with its cone abundance (Hart 2001) were adopted because observed seed dispersers of the plant are the blue rock-thrush Monticola solitarius (Emura et al. 2012) and the brown-eared bulbul Hypsipetes amaurotis stejnegeri (Kawakami et al. 2009), both of which are classified as Passerida, and thus, likely to have the UVS (ultraviolet sensitive) vision (Ödeen et al. 2011).

We then calculated the JND (just noticeable difference), a discrimination threshold between a given pair of colours, from their photon captures based on the Vorobyev-Osorio model (Vorobyev & Osorio 1998): JNDs were obtained from

Distributions of first principal coordinates of hue (a-c), luminance (d-f), and integrated JNDs (first principal components; g-i) for the colours of fruits and leaves of Scaevola taccada in relation to the endocarp morphotype with (C+) and without cork (C–) compared within and between organs. Boxes are composed of medians with first and third quartiles with whiskers indicating ranges. Numbers represent eigenvalues for respective first principal coordinates. Colour indication corresponds to those in Figure 1 and 2.

Figure 4.

for luminance (Siddiqi et al. 2004), where ∆f represents the logarithmic ratio of photon captures from the focal pair of colours by a given type of photoreceptors (i.e.

), and ω represents the relative abundance of each cone type in the posterior dorsal area of the retina, incorporating the Weber fraction of 0.05 (an error rate in the Weber-Fechner law by convention) (Vorobyev et al. 1998). We set a JND of 2 as the standard for the minimum discriminability, which was stricter than those in most visual modelling studies, as we were testing the lack of differences between the morphotypes. The reflectance spectra were drawn with the ‘pavo’ package (Maia et al. 2013), and the tetrahedral plots were drawn with the ‘lattice’ package (Sarker 2008), both in the R software, version 3.1.1 (R Foundation for Statistical Computing, Vienna; http://www.r-project.org/).

Statistics

Representative values (logarithmic means) for photon captures were estimated using log-linear mixed-effect models with the ‘lme4’ package (http://cran.r-projectorg/package=lme4) in R. We assigned individual plant as a random effect, but not assigned collection site because its explanatory power was too small in these models. Deterministic JNDs between morphotypes were calculated from exponential-transformed parameter estimates for photon captures in these models.

To evaluate the variance of discriminability between measured colours, we first calculated JNDs between all possible combinations of measured colours. The number of JNDs amounted to 97,020 (the cumulative sum of n – 1, i.e.

MCMC-sampled Bayesian posterior probabilities for the differences in first principal coordinates of hue (a-c), luminance (d-f), and integrated JNDs (first principal components; g-i) for the colours of fruits and leaves of Scaevola taccada in relation to the endocarp morphotypes with (C+) and without cork (C–) compared within and between organs. Values represent the effect sizes (posterior means) with lower and upper ends of the 95% credible intervals in the parentheses, with probabilities that the posterior means are statistically different from 0. Solid lines indicate the mean, and grey lines 0. Distributions with negative means were negative-transformed to aid interpretation. The positions of panels correspond to those in Figure 3.

, where n = 441) for within-organ comparisons (i.e. fruit-fruit or leaf-leaf), and 194,922 (n²) for the between-organ comparison (i.e. fruit-leaf). Next, we formulated distance matrices from all the calculated JNDs, and then converted these matrices into given sets of coordinates of a size n (i.e. 441 for fruit and leaf, respectively) through the principal coordinate analysis (PCoA; Legendre & Legendre 1998). We used the ‘cmdscale’ command default in R, setting the dimensional parameter ‘k’ as 3 since the original colour data scattered in a tridimensional space. Each measured colour thus had three unique coordinates, which represent its relative positions on respective axes among all the analysed colours. We chose given sets of coordinates with eigenvalues greater than 1 to analyse. First principal coordinates for both hue and luminance JNDs were integrated in a principal component analysis using a default command ‘prcomp’ in R to estimate the discriminability of colours of the objects to birds as a whole, since birds might not perceive the two aspects of colour separately (Osorio et al. 1999).

We then compared Bayesian posterior probabilities for these first principal coordinates/components between morphotypes inferred with the ‘MCMCglmm’ package (Hadfield 2010), in which plant identity and collection site were assigned as random effects. The number of iterations was set as 510,000, the thinning interval 1000, and the burn-in (pilot run) 10,000. The total Markov-chain Monte-Carlo (MCMC) samples were thus 500 for each parameter. From the inferred posterior probabilities, we estimated the difference in the discriminability between C– and C+, and calculated the probability that the differences of posterior means were greater than 0 (i.e. proportion of cases in which the difference of each MCMC sampled C– from that of C+ was below 0). Note that a statistically significant difference is not always related to the difference in perceived colours, as colours with JNDs less than 2 would be indiscernible to the viewer (Siddiqi et al. 2004), irrespective of whether such a difference is biologically substantial. Distributions with negative means were negative-transformed to aid interpretation.

RESULTS

Representative photon capture and deterministic JND

Although photon captures by some single-cone types differed significantly between C+ and C- (fruit UVS: x²₁ = 6.49, P = 0.011; leaf UVS: x²₁ = 25.1, P < 0.0001; leaf SWS: x²₁ = 5.74, P = 0.016; P > 0.05 for all else; Figure 1c-d), no JND greater than 2 was detected in the within-organ comparisons: 0.54 and 0.13 for fruit hue and luminance respectively, and 0.59 and 1.36 for leaf hue and luminance, respectively. Photon capture from leaves by double-cones was significantly different between these two morphoypes (x²₁ = 7.17, P = 0.007), while those from the fruit did not (x²₁ = 0.096, P = 0.76; Figure 1e). Again, no JND greater than 2 was detected between leaves of the two morphs. JNDs between fruits and leaves of any combinations of morphotypes were much greater than 2 both in hue and luminance (Table 1).

Discriminability with variance

The distribution of measured colours was different between fruits and leaves, but apparently not between the endocarp morphotypes in the avian tetrachromatic colour space (Figure 2). The distributions of first principal coordinates/components for JNDs appeared not to differ between the morphotypes (Figure 3). Eigenvalues of all the first principal coordinates/components were exclusively greater than 1, and thus their unit could be an approximation of the original parameter (i.e. hue and luminance JNDs, though its precision varied with eigenvalues) (Figure 3).

All the posterior probabilities for the differences in the first principal coordinates between the morphotypes were not greater than 0 at the 5% level (Figure 4a-f) except in the leaf hue JND (P = 0.048; Figure 4b). Nevertheless, none of the differences between the posterior means (i.e. effect sizes for the approximate JNDs for both C– and C+) was greater than 2.

Integrated discriminability: chromatic and achromatic thresholds

The posterior probabilities of differences in the first principal components between the morphotypes, which were integrated from the first principal coordinates of hue and luminance JNDs, were not greater than 0 at the 5% level (Figure 4g-i), and none of the differences between posterior means was greater than 2.

DISCUSSION

Fruit colour difference between morphotypes

Though the fruit colour of S. taccada was highly discriminable from its leaf colour in the deterministic approach, it is certain that such a measure of differences is insufficient to infer how seed-dispersing birds can distinguish between fruits of the two morphotypes, i.e. C+ and C–, because the variability of perceived differences was not considered. Actually, the deterministic JNDs of luminance between fruit and leaf colours varied roughly with the morphotype (Table 1), though such a difference is untestable using conventional methods. By processing data so that such dissimilarity was quantifiable, we showed that not only the perceived dissimilarity within fruits or leaves, but that between fruits and leaves were not different between morphotypes (Figure 4). These results suggest that the fruit colours of the different morphotypes would be indistinguishable from the point of view of avian seed dispersers. Thus, to avoid consuming fruits with cork, birds would have to rely on other discrimination cues, such as the size-weight relationship of fruits, or spatial locations of individual plants. In fact, frugivorous birds seem not to distinguish the morphotypes as they consumed C+ fruits (Emura et al. 2012, Kawakami et al. 2009) in areas where both morphotypes were found (Emura et al. 2014), and both types of seed were found in faeces and pellets of birds in our collecting sites (Emura, pers. obs.). To sum, the endocarp dimorphism is cryptic sensu stricto to frugivorous birds, and thus the fruit colour of S. taccada is a relatively dishonest signal about its reward.

Although we do not know the gene expression mechanism of the endocarp dimorphism of S. taccada, theories might help disentangle the evolutionary relationship of the dimorphism and the dishonesty (Gigord et al. 2001, Gray & McKinnon 2007). For example, we would assume that Mendel’s law of dominance would apply both for C+ and a putative mutant allele for a novel fruit colour, and both are genetically uncorrelated traits. Although the novel colour gene seems to be adaptive, this new colour morph still could not provide an honest signal to seed dispersers because fruit colours are expected to segregate independently of the endocarp types according to Mendel’s law. Therefore, a required condition for an honest signal is a strong genetic correlation between dominant alleles of the endocarp and fruit colour genes. However, such genetic hitchhiking would be stochastically unlikely (Barton 2000), so is the signal honesty. The low probability of genetic linkage implies that divergent selection to promote speciation is required to make the fruit colour an honest signal under endocarp dimorphism (Gray & McKinnon 2007). Therefore, any kind of habitat segregation between respective morphotypes is necessary to prevent avian seed dispersal as well as gene flows through pollination (Forsman et al. 2008, Gillespie et al. 2012, Gray & McKinnon 2007). In fact, some close relatives of S. taccada that have different fruit colours and lack the cork endocarp, such as S. montana in New Caledonia, which inhabit inland areas that are far from coastal areas where S. taccada is found (Howarth et al. 2003). Considering such genetic constraints as improbable genetic linkage or introgression, the dishonesty in the reward signal of S. taccada might have been formed as a byproduct, rather than for the sake of the plant. Further genetic analyses could reveal such evolutionary dynamics of S. taccada.

Application of the PCoA to visual modelling

The nature of colour perception by colour-sensing animals prevents application of common statistical procedures (Tanaka in press). We review statistical constraints on the visual model and discuss how to circumvent statistical problems encountered in previous studies. Few techniques exist to test the lack of difference in colour contrasts between samples. In this study, we employed principal coordinate analysis (PCoA, or multi-dimensional scaling), which a set of coordinates in a given dimension is formulated from a distance or similarity matrix (Legendre & Legendre 1998). The PCoA is particularly compatible with visual modelling because a JND is an approximated perceived dissimilarity, i.e. the psychophysical distance between given two colours (Backhaus et al. 1987, Vorobyev & Osorio 1998). Four major methods have been adopted in the studies of avian visual modelling to quantify colour differences: (1) representative (averaged) reflectance/photon captures and deterministic JNDs (Avilés 2008, Feeney et al. 2014, Tanaka et al. 2011); (2) estimating volumes in the colour space occupied by given colours (Langmore et al. 2011, Maia et al. 2013, Stoddard & Prum 2008, 2011; Stoddard & Stevens 2011); (3) averaging JNDs (Antonov et al. 2011, Avilés 2008, Avilés et al. 2011, Cassey et al. 2009); and (4) the Mantel test (Avilés et al. 2012, Fadzly et al. 2013). However, each of these methods has drawbacks.

First, the representative approach begins with calculation of averages for reflectance spectra (Avilés 2008) or photon captures (Tanaka et al. 2011; this study), or centroids of focal colours in the colour space (Feeney et al. 2014). Then, JNDs for focal colours against a given set of average photon captures (Avilés 2008, Avilés et al. 2011), JNDs of all colours from the centroid of the focal colours (Feeney et al. 2014), or JNDs between averages are calculated (Tanaka et al. 2011; Table 1 of the present study). Although such results are intuitive, the representative approach partially or even wholly discounts the variance of data (i.e. JNDs are deterministic), which could cause type I errors. Second, the volume estimation calculates a volume that encompasses outlying data points (i.e. minimum convex polygon) of measured colours in the tetrahedron (Maia et al. 2013, Stoddard & Prum 2008). While the volume estimation efficiently reduces the dimension of data, it disregards the density of distribution, which could cause type II errors by over-weighting outliers. The third method, averaging JNDs, treats each JND value as a repeated measurement, and then analyses the effects of covariates on the set of the JND values (i.e. as a response variable) with a linear model (Cassey et al. 2009). However, the averaging of JNDs severely violates two fundamental principles of statistics: (1) the congruence principle, and (2) the independence of data. In reference to the congruence principle, treating a set of (Euclidean) distances as a linear variable is to ignore the mutual geometric dependence of each datum, which is non-existent in linear variables (Biondini et al. 1991, Endler & Mielke 2005, Endler et al. 2005). With regard to the independence principle, each colour has n – 1 JNDs, so the number of analyzed JNDs in this procedure must be n(n – 1). However, a single JND always stems from a pair of measured colours, so the total number of JNDs cannot be the sum of the number of JNDs that are related to each measured colour. The original number of JNDs is the cumulative sum of n – 1, i.e.

, and thus the number of analyzed JNDs will be literally duplicated.

Finally, the Mantel test should be statistically more robust than the other methods as it satisfies the congruence principle, and thus is capable of directly analysing distance matrices. However, the Mantel test is designed to estimate correlation between two distance matrices, and thus causal relationships could not easily be detected. Moreover, the Mantel test is not applicable to longitudinal data (i.e. repeated measurements), and thus pseudoreplication is unavoidable. Repeated measurement is the strongest way to reduce measurement errors (Bolker et al. 2009), and is especially important in the visual modelling framework since such errors could occur at a scale beyond human perception. The partial Mantel test might do better at avoiding pseudoreplication, although its validity is controversial (Raufaste & Rousset 2001).

The PCoA, or MDS, clearly has superiority over the procedures described above, because of its statistical validity. The concept of the PCoA is similar to that of the Mantel test, in that both can analyse distance matrices, but different in that the PCoA yields linear coordinates from distance matrices (Legendre & Legendre 1998). The resultant set of coordinates corresponds exactly to a well-known normally distributed variable, the principal component, and is thus compatible with linear models, currently the commonest framework in biostatistics (Bolker et al. 2009). Even though the PCoA does have some drawbacks, such as the reduction of dimensions inevitably entailing missing information, it could reduce the risk of false discoveries or false rejections of null hypotheses, while avoiding statistical violations such as incongruence or pseudoreplication. To sum, applying the PCoA to visual modelling analyses could expand the application of hypothesis testing statistically more robustly, particularly when the data structure is such highly multi-dimensional as avian tetrachromatic colours (Backhaus et al. 1987).

Conclusion

Our study revealed that the fruit colours of S. taccada are not discernibly different to avian seed dispersers between morphotypes, based on the outputs from our robust statistical modelling. Thus, the seed-dispersal strategy of C+ individuals of the plant is likely to result in a dishonest signal, which can impose greater costs on the avian seed dispersers. However, it is not certain that the signalling dishonesty exists solely for the benefit of the plant, rather than as a byproduct of other factors.

Acknowledgements

We thank W. Kitamura for comments on statistics and W. Feeney for his helpful comments and suggestions. This study was financially aided by Grants-in-aid from the Japan Society for the Promotion of Science: Young Scientists B: 24770028 (K. D. T.); Basic A: 23255004 (K. U.); by the Japan Science Society: Sasakawa Scientific Research Grant: 26-445 (N. E.).

Literature Cited

• ANTONOV, A., AVILÉS, J. M., STOKKE, B. G., SPASOVA, V., VIKAN, J. R., MOKSNES, A., YANG, C., LIANG, W. & RØSKAFT, E. 2011. Egg discrimination in an open nesting passerine under dim light conditions. Ethology 117:1128-1137.

• AVILÉS, J. M. 2008. Egg colour mimicry in the common cuckoo Cuculus canorus as revealed by modelling host retinal function. Proceedings of the Royal Society of London series B – Biological Sciences 275:2345-2352.

• AVILÉS, J. M., SOLER, J. J. & HART, N. S. 2011. Sexual selection based on egg colour: physiological models and egg discrimination experiments in a cavity-nesting bird. Behavioral Ecology and Sociobiology 65:1721-1730.

• AVILÉS, J. M., VIKAN, J. R., FOSSØY, F., ANTONOV, A., MOKSNES, A., RØSKAFT, E., SHYKOFF, J. A., MØLLER, A. P. & STOKKE, B. G. 2012. Egg phenotype matching by cuckoos in relation to discrimination by hosts and climatic conditions. Proceedings of the Royal Society of London series B – Biological Sciences 297:1967-1976.

• BACKHAUS, W., MENZEL, R. & KREISSL, S. 1987. Multidimensional scaling of color similarity in bees. Biological Cybernetics 56:293-304.

• BARTON, N. H. 2000. Genetic hitchhiking. Philosophical Transactions of the Royal Society of London Series B – Biological Sciences 355:1553-1562.

• BIONDINI, M. E., MIELKE, P. W. & REDENTE, E. F. 1991. Permutation techniques based on Euclidean analysis spaces: a new and powerful statistical method for ecological researches. Pp. 221-240 in Feoli, F. & Orlóci, L. (eds.). Computer assisted vegetation analysis. Kluwer Academic Publisher, Dordrecht.

• BOLKER, B. M., BROOKS, M. E., CLARK, C. J., GEANGE, S. W., POULSEN, J. R., STEVENS, H. H. & WHITE, J.-S. S. 2009. Generalized linear mixed models: a practical guide for ecology and evolution. Trends in Ecology and Evolution 24:127-135.

• CASSEY, P., EWEN, J. G., MARSHALL, N. J., VOROBYEV, M., BLACKBURN, T. M. & HAUBER, M. E. 2009. Are avian eggshell colours effective intraspecific communication signals in the Muscicapoidea? A perceptual modelling approach. Ibis 151:689-698.

• EDWARDS, D. P. & YU, D. W. 2007. The role of sensory traps in the origin, maintenance, and breakdown of mutualism. Behavioral Ecology and Sociobiology 61:1321-1327.

• EMURA, N., KAWAKAMI, K., DEGUCHI, T. & SONE, K. 2012. Potential role of frugivorous birds in the recovery process of forest vegetation after feral goat eradication in Mukojima Island, the Bonin Islands. Journal of Forest Research 17:352-359.

• EMURA, N., DENDA, T., SAKAI, M. & UEDA, K. 2014. Dimorphism of the seed-dispersal organ in a pantropical coastal plant, Scaevola taccada: heterogeneous population structures between islands and microhabitat types. Ecological Research 29:733-740.

• ENDLER, J. A. & MIELKE, P. W. 2005. Comparing entire colour patterns as birds see them. Biological Journal of the Linnean Society 86:405-431.

• ENDLER, J. A., WESTCOTT, D. A., MADDEN, J. R. & ROBSON, T. 2005. Animal visual systems and the evolution of color patterns: sensory processing illuminates signal evolution. Evolution 59:1795-1818.FADZLY, N., BURNS, K. C. & ZUHARAH, W. F. 2013. Evaluating frugivore-fruit interactions using avian eye modelling. Tropical Life Sciences Research 24:31-50.

• FEENEY, W. E., STODDARD, M. C., KILNER, R. M. & LANGMORE, N. E. 2014. “Jack-of-all-trades” egg mimicry in the brood parasitic Horsfield's bronze-cuckoo? Behavioral Ecology 25: 1365-1373.

• FORSMAN, A., SANESJÖ, J., CAESAR, S. M. & KARLSON, M. 2008. A model of ecological and evolutionary consequences of color polymorphism. Ecology 89:34-40.

• FOSTER, M. S. & DELAY, L. S. 1998. Dispersal of mimetic seeds of three species of Ormosia (Leguminosae). Journal of Tropical Ecology 14:389-411.

• FRECKLETON, R. P. & CÔTÉ, I. M. 2003. Honesty and cheating in cleaning symbioses: evolutionarily stable strategies defined by variable pay-offs. Proceedings of the Royal Society of London series B: Biological Sciences 270:299-305.

• GALETTI, M. 2002. Seed dispersal of mimetic fruits: parasitism, mutualism, aposematism or exaptation? Pp 177-191 in Levey, D. J., Silva, W. R. & Galettti, M. (eds.). Seed dispersal and frugivory: ecology, evolution and conservation. CABI Publishing, New York.

• GIGORD, L. D. B., MACNAIR, M. R. & SMITHSON, A. 2001. Negative frequency-dependent selection maintains a dramatic flower color polymorphism in the rewardless orchid Dactylorhiza sambucina (L.) Soò. Proceedings of the National Academy of Sciences, U.S.A. 98:6253-6255.

• GILLESPIE, R. G., BALDWIN, B. G., WATERS, J. M., FRASER, C. I., NUKULA, R. & RODERICK, G. K. 2012. Long-distance dispersal: a framework for hypothesis testing. Trends in Ecology and Evolution 27:47-56.

• GRAY, S. M. & MCKINNON, J. S. 2007. Linking color polymorphism maintenance and speciation. Trends in Ecology and Evolution 22:71-79.

• HADFIELD, J. D. 2010. MCMC methods for multi-response generalized linear mixed models: the MCMCglmm R package. Journal of Statistical Software 33:1-22.

• HART, N. S. 2001. Variations in cone photoreceptor abundance and the visual ecology of birds. Journal of Comparative Physiology A 187:685-698.

• HART, N. S., PARTRIDGE, J. C., CUTHILL, I. C. & BENNETT, A. T. D. 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). Journal of Comparative Physiology A 186:375-387.

• HOWARTH, D. G., GUSTAFSSON, M. H. G., BAUM, D. A. & MOTLEY, T. J. 2003. Phylogenetics of the genus Scaevola (Goodeniaceae): implication for dispersal patterns across the Pacific Basin and colonization of the Hawaiian Islands. American Journal of Botany 90:915-923.

• KAWAKAMI, K., MIZUSAWA, L. & HIGUCHI, H. 2009. Re-established mutualism in a seed-dispersal system consisting of native and introduced birds and plants on the Bonin Islands, Japan. Ecological Research 24:741-748.

• KOSHITAKA, H., KINOSHITA, M., VOROBYEV, M. & ARIKAWA, K. 2008. Tetrachromacy in a butterfly that has eight varieties of spectral receptors. Proceedings of the Royal Society of London series B – Biological Sciences 275:947-954.

• LANGMORE, N. E., STEVENS, M., MAURER, G., HEINSOHN, R., HALL, M. L., PETERS, A. & KILNER, R. M. 2011. Visual mimicry of host nestlings by cuckoos. Proceedings of the Royal Society of London series B – Biological Sciences 278:2455-2463.

• LEGENDRE, P. & LEGENDRE, L. 1998. Numerical ecology. (Second edition). Elsevier, Amsterdam. 853 pp.

• MAIA, R., ELIASON, C. M., BITTON, P.-P., DOUCET, S. M. & SHAWKEY, M. D. 2013. pavo: an R package for the analysis, visualization and organization of spectral data. Methods in Ecology and Evolution 4:906-913.

• ÖDEEN, A., HÅSTAD, O. & ALSTRÖM, P. 2011. Evolution of ultraviolet vision in the largest avian radiation - the passerines. BMC Evolutionary Biolology 11:313.

• OSORIO, D., MIKLÓSI, A. & GONDA, Z. S. 1999. Visual ecology and perception of coloration patterns by domestic chicks. Evolutionary Ecology 13:673-689.

• OSORIO, D. & VOROBYEV, M. 1996. Colour vision as an adaptation to frugivory in primates. Proceedings of the Royal Society of London series B – Biological Sciences 263:593-599.

• RAUFASTE, N. & ROUSSETT, F. 2001. Are partial Mantel tests adequate? Evolution 55:1703-1705.

• SARKER, D. 2008. Lattice: multivariate data visualization with R. Springer, New York. 268 pp.

• SIDDIQI, A., CRONIN, T. W., LOEW, E. R., VOROBYEV, M. & SUMMERS, K. 2004. Interspecific and intraspecific views of color signals in the strawberry poison frog Dendrobates pumilio. Journal of Experimental Biolology 207:2471-2485.

• STODDARD, M. C. & PRUM, R. O. 2008. Evolution of Avian plumage color in tetrahedral color space: a phylogenetic analysis of New World buntings. The American Naturalist 171:755–776.

• STODDARD, M. C. & PRUM, R. O. 2011. How colorful are birds? Evolution of the avian plumage color gamut. Behavioral Ecology 22:1042-1052.

• STODDARD, M. C. & STEVENS, M. 2011. Avian vision and the evolution of egg color mimicry in the common cuckoo. Evolution 65:2004-2013.

• STUART-FOX, D. 2005. Deception and the origin of honest signals. Trends in Ecology and Evolution 20:521-523.

• TANAKA, K. D. A colour to birds and to humans: why is it so different? Journal of Ornithology (in press)

• TANAKA, K. D., MORIMOTO, G., STEVENS, M. & UEDA, K. 2011. Rethinking visual supernormal stimuli in cuckoos: visual modeling of host and parasite signals. Behavioral Ecology 22:1012-1019.

• VOROBYEV, M. & OSORIO, D. 1998. Receptor noise as a determinant of colour thresholds. Proceedings of the Royal Society of London series B – Biological Sciences 265:351-358.

• VOROBYEV, M., OSORIO, D., BENNETT, A. T. D., MARSHALL, N. J. & CUTHILL, I. C. 1998. Tetrachromacy, oil droplets and bird plumage colours. Journal of Comparative Physiology A 183:621-633