Behavioural elements reflect phenotypic colour divergence in a poison frog

Abstract

The coexistence of both aposematic and cryptic morphs as different anti-predator strategies within a species seems to be an unusual phenomenon in nature. The strawberry poison frog, Oophaga pumilio, shows an astonishing colour diversity among populations in western Panama. In this study we selected a red and a green colour morph from two Panamanian islands (Isla Solarte and Isla Colón) for behavioural observations and measurements of conspicuousness. We found that red frogs were more visible to both conspecific frogs and potential predators than green frogs. Interestingly the difference in conspicuousness was most pronounced at the substrate that males used as principal calling places. Red males were more active and spent more time foraging than green males, which spent more time hidden. The association between conspicuousness of colouration and behaviour results in a more aposematic and a more cryptic anti-predator strategy. This is the first study which links differences in conspicuousness between animals on their natural backgrounds to differences in foraging as well as anti-predator behaviour and discusses the results in light of previous findings of toxicity analyses and potential costs and benefits of aposematism. To this end, our study adds a novel perspective for explaining extreme colour diversity between populations within an initially aposematic species.

Appendices

Appendix 1: Visual modelling

An object’s contrast against the background results from the ratio of reflected light across all wavelengths from object and background (q) that reaches the visual system of the viewer. The contrast also depends on the visual system of the viewer, e.g. the sensitivity of the present cone types (S _i ). We calculated the colour contrast for an avian model predator with tetrachromatic vision (blue tit, Hart et al. 2000; Hart 2001a) as well as for strawberry poison frogs with trichromatic vision (Siddiqi et al. 2004). Sensitivity data for all cone types were available for blue tits and some other passerine birds (Hart et al. 1998; Hart 2001a, b). Since spectral sensitivities do not vary much between 26 bird species investigated (Hart 2001b) and the only bird species known to predate on poison frogs is the rufous motmots Baryphthengus martii (Master 1999) which closely related to passerine birds (Livezey and Zusi 2007) we considered blue tit spectral curves as appropriate for a first approximation for calculating conspicuousness of the frogs for an avian model predator. The maxima of spectral sensitivity of blue tits are 372 nm for the ultraviolet/violet sensitive receptor (UV/VS), 449 nm for the short wave length sensitive receptor (SWS), 502 nm for the mid wave length sensitive receptor (MWS) and 563 nm for the long wave length sensitive receptor (LWS). In contrast to birds strawberry poison frogs do not possess UV/VS receptors. Their sensitivity maxima are SWS: 466 nm, MWS: 489 nm and LWS: 561 nm (Siddiqi et al. 2004).

In the first step of the visual model, the entire quantity of light (total quantum catch = Q) that reaches the visual system was calculated for each cone type (i) of the avian model and the frog model as the integrated product of reflectance spectrum of the frog, the receptor sensitivity spectrum and irradiance spectrum:

$$ Q_{i} = \int R (\lambda )S_{i} (\lambda )I(\lambda )d\lambda $$

(1)

where R(λ) is spectral reflection of the frog at wave length λ, S _i (λ) is the sensitivity of cone type i of the viewer for wave length λ, I(λ) is the irradiance for wave length λ (= spectrum of light that enters the eye) and d(λ) is the interval of 5 nm usually used for integrations over the visual spectrum (here 350–800 nm) (Vorobyev et al. 1998).

Analogous calculations were performed for the background. As a result of these calculations the total quantum catch (Q) that reached the visual system was $Q_{i}^{F}$ for the frogs and $Q_{i}^{B}$ for the background. In the second step of the visual model the ratio of the frog to all quantum catches of the background was computed to get the value:

$$ q_{i} = {\frac{{Q_{i}^{F} }}{{Q_{i}^{B} }}} $$

(2)

for each cone type i.

For the next step it was assumed that the photoreceptor response follows the Weber-Fechner laws (Vorobyev et al. 1998; Chiao et al. 2000) which states that the intensity of the receptor signal (f) of cone type i is given by the natural logarithm of the quantum catch q _i :

$$ f_{i} = \ln q_{i} $$

(3)

To calculate the colour contrast ΔS (the perceived difference between two colours: here between frog and background) the formulas (4) and (5) were used (Vorobyev et al. 1998; Stuart-Fox et al. 2003; Siddiqi et al. 2004). Formula (4) was developed for the trichromatic visual system (frog) and formula (5) was developed for the tetrachromatic visual system (bird):

$$ \begin{aligned} \Updelta S^{2} = & \omega_{S}^{2} (f_{L} - f_{M} )^{2} + \omega_{M}^{2} (f_{L} - f_{S} )^{2} + \omega_{L}^{2} (f_{S} - f_{M} )^{2} \\ /(\omega_{S} \omega_{M} )^{2} + (\omega_{S} \omega_{L} )^{2} + (\omega_{M} \omega_{L} )^{2} \\ \end{aligned} $$

(4)

$$ \begin{aligned} \Updelta S^{2} = & (\omega_{U} \omega_{S} )^{2} \left( {f_{L}^{t} - f_{M}^{t} } \right)^{2} + (\omega_{U} \omega_{M} )^{2} \left( {f_{L}^{t} - f_{S}^{t} } \right)^{2} + (\omega_{U} \omega_{L} )^{2} \left( {f_{M}^{t} - f_{S}^{t} } \right)^{2} \\ + (\omega_{S} \omega_{M} )^{2} \left( {f_{L}^{t} - f_{U}^{t} } \right)^{2} + (\omega_{S} \omega_{L} )^{2} \left( {f_{M}^{t} - f_{U}^{t} } \right)^{2} + (\omega_{M} \omega_{L} )^{2} \left( {f_{S}^{t} - f_{U}^{t} } \right)^{2} \\ /(\omega_{U} \omega_{S} \omega_{M} )^{2} + (\omega_{U} \omega_{S} \omega_{L} )^{2} + (\omega_{U} \omega_{M} \omega_{L} )^{2} + (\omega_{S} \omega_{M} \omega_{L} )^{2} \\ \end{aligned} $$

(5)

The subscripts indicate the different photoreceptors (cone types i): U for UV/VS, S for SWS, M for MWS, and L for LWS.

In these formulas the Weber fraction ω _i describes the noise to signal ratio of the different cone types which varies for different wave lengths. The Weber fraction ω _i is inversely proportional to the relative number of receptor cells (η _i ) of each cone type ($\omega_{i} = \upsilon /\sqrt {\eta_{{{}^{i}}} }$, whereby υ is the noise to signal ratio of a single cone and a constant for all cone types). The number of receptor cells (η _i ) differs between species (Hart 2001b). For most studies no values for η _i of potential predators are available, and as a surrogate values of bird species with well-known visual systems like blue tit (Parus caeruleus), European blackbird (Turdus merula) or European starling (Sturnus vulgaris) were often used for visual modelling (e.g. Stuart-Fox et al. 2003; Cummings et al. 2008). For strawberry poison frogs η _i values were estimated in Siddiqi et al. (2004). Since calculations of Hart (2001a) and Stuart-Fox et al. (2003) indicate that for the birds differences in η _i do not greatly affect the qualitative results, the η _i data for European blackbirds presented by Hart et al. (2000), Hart (2001a) were also used in this study (Table 1). To verify whether differences in relative numbers of η _i would greatly influence model calculations in our system we also performed calculations of ΔS with the η _i of blue tits. The results for ΔS for η _i of blackbirds and blue tits were very similar _. As proposed by Stuart-Fox et al. (2003) and Siddiqi et al. (2004) we applied the estimate of ω _L  = 0.05 for both the avian and the frog model and subsequently calculated ω _U, ω _S and ω _M (Table 1; Vorobyev and Osorio 1998).

Apart from the colour contrast the brightness (or achromatic) contrast is important for prey detection. The brightness contrast seems to be essential for long range tasks or detection of small targets (Osorio et al. 1999; Théry et al. 2004). For birds it is assumed that double cones which contain the LWS pigment are responsible for achromatic tasks (Hart et al. 1998; Vorobyev et al. 1998). Also in frogs the LWS class is assumed to be involved in perceiving brightness (Siddiqi et al. 2004). Thus, we estimated brightness contrasts with regard to the spectral sensitivities of the LWS cone only:

$$ \Updelta B = {\frac{{f_{L} }}{{\omega_{L} }}} $$

(see also Stuart-Fox et al. 2003; Siddiqi et al. 2004).

See Table 2.

Table 2 Visual modelling: relative number of receptor types η _i and Weber fractions ω _i used in this study

Appendix 2

See Fig. 5.

Fig. 5

Irradiance measurements: mean irradiance for Isla Solarte and Isla Colón

Appendix 3

See Fig. 6.

Fig. 6

Spectral reflectance of ventral and dorsal body parts of strawberry poison frogs from (a) Isla Solarte and (b) Isla Colón

Appendix 4

See Fig. 7.

Fig. 7

Spectral reflectance curves for most common backgrounds on (a) Isla Solarte and (b) Isla Colón

Appendix 5

See Tables 3 and 4.

Table 3 Pairwise comparisons for colour contrasts on distinct background for frog and bird vision

Table 4 Pairwise comparisons for brightness contrasts on distinct background for frog and bird vision