Functional Ecology Appendix S1 Description of bird, snake and crab vision models. Bird vision model Calculations were done based on a set of functions in Matlab R2009a (The MathWorks Inc, USA). The model starts by calculating the cone quantum catches (qi) for each photoreceptor class for froglets and ambient radiance spectra as 750 ππππππ = ∫ π ( λ)πΌ(λ)π΄α΅’(λ)πλ, λ=300 here π΄π (π) represents the absorptance spectrum for each of the four photoreceptor cone classes of the bird integrated over 1 nm intervals from 300 to 750 nm. π (π) and πΌ(π) represent the reflectance spectrum of the froglet´s skin and the irradiance spectra measured in the field respectively. Resulting photon quantum catches were standardized in order to account for variation in light conditions, using the von Kries transformation adaptation coefficient. This method assumes that photoreceptors adjust their sensitivity in proportion to the background light environment πΎπππππ = 1 , 750 ∫π=300 π΄α΅’(π)πΌ(π)ππ This procedure ensures that colour perception relies on colour constancy, whereby the visual system removes variation in ambient light so that colours look similar under variable light conditions (Cuthill 2006). Therefore, the adjusted quantum catch data for each photoreceptor class was calculated as ππ = πΎπππππ π₯ ππππππ Flores, E. E., Stevens, M, Moore, A. J., and Blount, J. D. Diet, development and the optimisation of warning signals in post-metamorphic green and black poison frogs Functional Ecology Values of relative single cone quantum catches were then included in a Principal Component Analysis (see Data analyses). Following calculations of quantum catches, the model assumes that the signal of each cone channel is proportional to the logarithm of the adjusted quantum catches; as such the contrast between a pair of stimuli was calculated as the quotient of adjusted photon quantum catches βππ = πππ [ππ1 (ππππ π ππππ‘ππ)] [ππ2 (ππππππππ’ππ π ππππ‘ππ)] The Vorobyev-Osorio colour discrimination model is based on evidence that colour discrimination is determined by noise arising in the photoreceptors and is independent of light intensity. Noise in each photoreceptor channel (ππ ) was calculated as ππ = ππ √ππ Where ππ was taken as 0.05 and represents the Weber fraction of the most abundant cone type (Siddiqi et al. 2004) and ππ is the relative number of receptor types in the retina of the blue tit (Hart et al. 2000) (ππΏ = 1.00, ππ = 0.99, ππ = 0.71, πππ = 0.37). Colour (chromatic) discrimination in the tetrachromatic visual model was calculated as JND values using the following equation π½ππ·ππππ πππππ (πππ ππ )2 (βππΏ − βππ )2 + (πππ ππ )2 (βππΏ − βππ )2 + (πππ ππΏ )2 (βππ − βππ )2 √+(ππ ππ )2 (βππΏ − βπππ )2 + (ππ ππΏ )2 (βππ − βπππ )2 + (ππ ππΏ )2 (βππ − βπππ )2 = (πππ ππ ππ )2 + (πππ ππ ππΏ )2 + (πππ ππ ππΏ )2 + (ππ ππ ππΏ )2 2 Flores, E. E., Stevens, M, Moore, A. J., and Blount, J. D. Diet, development and the optimisation of warning signals in post-metamorphic green and black poison frogs Functional Ecology Since in our model, overall perceived luminance is considered to arise from stimulation of double cone photoreceptors, luminance discrimination was evaluated as βππ· π½ππ·ππππ ππ’πππππππ = ( ) ππ· Overall variation in colour was based on values from single cone photon catch scores (and a Principal Component Analysis of them) and variation in overall luminance was based on double cone photon catch scores (ΔqD). Colour and luminance discrimination (conspicuousness) was based on JND values. Snake vision model Calculations were done based on a set of functions in Matlab R2009a (The MathWorks Inc, USA). The model starts by calculating the cone quantum catches (qi) for each photoreceptor class for froglets and ambient radiance spectra as 750 πππ ππππ = ∫ π ( λ)πΌ(λ)π΄α΅’(λ)πλ, λ=300 here π΄π (π) represents the absorptance spectrum for each of the three photoreceptor cone classes (i.e. UV, SW and LW) integrated over 1 nm intervals from 300 to 750 nm. π (π) and πΌ(π) represent the reflectance spectrum of the froglet´s skin and the irradiance spectra measured in the field respectively. Resulting photon quantum catches were standardized in order to account for variation in light conditions, using the von Kries transformation adaptation coefficient. This method assumes that photoreceptors adjust their sensitivity in proportion to the background light environment Flores, E. E., Stevens, M, Moore, A. J., and Blount, J. D. Diet, development and the optimisation of warning signals in post-metamorphic green and black poison frogs Functional Ecology πΎππ ππππ = 1 , 750 ∫π=300 π΄α΅’(π)πΌ(π)ππ This procedure ensures that colour perception relies on colour constancy, whereby the visual system removes variation in ambient light so that colours look similar under variable light conditions (Cuthill 2006). Therefore, the adjusted photon quantum catch data for each photoreceptor class was calculated as ππ = πΎππ ππππ π₯ πππ ππππ Values of the relative single cone quantum catches were then included in a Principal Component Analysis (see Data analyses). Following calculations of quantum catches, the model assumes that the signal of each cone channel is proportional to the logarithm of the adjusted quantum catches; as such the contrast between a pair of stimuli was calculated as the quotient of adjusted quantum catches βππ = πππ [ππ1 (ππππ π ππππ‘ππ)] [ππ2 (ππππππππ’ππ π ππππ‘ππ)] The Vorobyev-Osorio visual discrimination model is based on evidence that color discrimination is determined by noise arising in the photoreceptors and is independent of light intensity. Noise in each photoreceptor channel (ππ ) was calculated as ππ = ππ √ππ Where ππ was taken as 0.05 according to Siddiqi et al. (2004) and represents the Weber fraction of the most abundant cone type and ππ is the relative number of receptor types in the retina of the snake (data for Thamnophis sirtalis, Sillman et al. 1997) (ππΏ = 0.85, Flores, E. E., Stevens, M, Moore, A. J., and Blount, J. D. Diet, development and the optimisation of warning signals in post-metamorphic green and black poison frogs Functional Ecology ππ = 0.10, πππ = 0.05). Colour (chromatic) discrimination in the trichromatic visual model was calculated as JND values using the following equation π½ππ·π ππππ πππππ’π 2 =√ (πππ )2 (βππΏ − βππ )2 + (ππ )2 (βππΏ − βπππ )2 + (ππΏ )2 (βπππ − βππ )2 (πππ ππ )2 + (πππ ππΏ )2 + (ππ ππΏ )2 We considered overall perceived luminance in snakes as in other vertebrates to arise from stimulation of LW cone photoreceptors with sensitivity in the long wavelength part of the spectra; therefore luminance discrimination was evaluated as βππΏ π½ππ·π ππππ ππ’πππππππ = ( ) ππΏ Overall variation in colour was based on values from single cone photon catch scores (and a Principal Component Analysis of them) and variation in overall luminance was based on LW cone photon catch scores (ΔqL). Colour and luminance discrimination (conspicuousness) was based on JND values. Crab vision model Similar calculations were done based on a set of functions in Matlab R2009a (The MathWorks Inc, USA). The model starts by calculating the cone quantum catches (qi) for each photoreceptor class for froglets and ambient radiance spectra as 750 ππππππ = ∫ π ( λ)πΌ(λ)π΄α΅’(λ)πλ, λ=300 Flores, E. E., Stevens, M, Moore, A. J., and Blount, J. D. Diet, development and the optimisation of warning signals in post-metamorphic green and black poison frogs Functional Ecology here π΄π (π) represents the absorptance spectrum for each of the two photoreceptor cone classes (i.e. SW and LW) integrated over 1 nm intervals from 300 to 750 nm. Resulting photon quantum catches were standardized in order to account for variation in light conditions, using the von Kries transformation adaptation coefficient. This method assumes that photoreceptors adjust their sensitivity in proportion to the background light environment πΎπππππ = 1 , 750 ∫π=300 π΄α΅’(π)πΌ(π)ππ This procedure ensures that colour perception relies on colour constancy, whereby the visual system removes variation in ambient light so that colours look similar under variable light conditions (Cuthill 2006). Therefore, the adjusted photon quantum catch data for each photoreceptor class was calculated as ππ = πΎπππππ π₯ ππππππ Values of the relative single cone quantum catches were then included in a Principal Component Analysis (see Data analyses). Following calculations of quantum catches, the model assumes that the signal of each cone channel is proportional to the logarithm of the adjusted quantum catches; as such the contrast between a pair of stimuli was calculated as the quotient of adjusted quantum catches βππ = πππ [ππ1 (ππππ π ππππ‘ππ)] [ππ2 (ππππππππ’ππ π ππππ‘ππ)] Flores, E. E., Stevens, M, Moore, A. J., and Blount, J. D. Diet, development and the optimisation of warning signals in post-metamorphic green and black poison frogs Functional Ecology The Vorobyev-Osorio visual discrimination model is based on evidence that colour discrimination is determined by noise arising in the photoreceptors and is independent of light intensity. Noise in each photoreceptor channel (ππ ) was calculated as ππ = ππ √ππ Where ππ was taken as 0.12 based on physiological receptor noise data according to Hempel De Ibarra et al. (2000) for honeybees Apis mellifera and ππ is the relative number of receptor types that for our purposes were set as (ππΏ = 0.5, ππ = 0.5) following (Cummings et al. 2008). Colour (chromatic) discrimination in the dichromatic visual model was calculated as JND values using the following equation 2 (βπ − βπ )2 π πΏ π½ππ·ππππ πππππ’π = √ (ππ )2 + (ππΏ )2 We considered overall perceived luminance in crab to arise from stimulation of LW cone photoreceptors with sensitivity in the long wavelength part of the spectra, luminance discrimination was evaluated as βππΏ π½ππ·ππππ ππ’πππππππ = ( ) ππΏ Overall variation in colour was based on values from single cone photon catch scores (and a Principal Component Analysis of them) and variation in overall luminance was based on LW cone photon catch scores (ΔqL). Colour and luminance discrimination (conspicuousness) was based on JND values. Flores, E. E., Stevens, M, Moore, A. J., and Blount, J. D. Diet, development and the optimisation of warning signals in post-metamorphic green and black poison frogs Functional Ecology References Cummings, M.E., Jordao, J.M., Cronin, T.W. & Oliveira, R.F. (2008) Visual ecology of the fiddler crab, Uca tangeri: effects of sex, viewer and background on conspicuousness. Animal Behaviour, 75, 175–188. Cuthill, I.C. (2006) Color perception. Bird coloration, vol. I: Mechanisms and measurements (eds G.E. Hill & K.J. McGraw), pp. 3–40. Harvard University Press, Cambridge. 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: Sensory, Neural, and Behavioral Physiology, 186, 375– 387. Hempel De Ibarra, N., Vorobyev, M., Brandt, R. & Giurfa, M. (2000) Detection of bright and dim colours by honeybees. The Journal of Experimental Biology, 203, 3289–3298. 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 Biology, 207, 2471–2485. Sillman, A.J., Govardovskii, V.I., Rohlich, P., Southard, J.A. & Loew, E.R. (1997) The photoreceptors and visual pigments of the garter snake (Thamnophis sirtalis): a microspectrophotometric, scanning electron microscopic and immunocytochemical study. Journal of Comparative Physiology A: Sensory, Neural, and Behavioral Physiology, 181, 89–101. . Flores, E. E., Stevens, M, Moore, A. J., and Blount, J. D. Diet, development and the optimisation of warning signals in post-metamorphic green and black poison frogs