Visual Mate Choice in Poison Frogs Kyle Summers, Rebecca Symula. Mark Clough and Thomas Cronin Department of Biology, East Carolina University, Greenville, North Carolina 27858, USA. Email: summersk@mail.ecu.edu We wish to thank Kyle for his permission to use this paper. It has been edited to remove the statistical information and the list of references. If you require these please contact the Editor. Anuran amphibians have played a central role in research on mating strategies and mate choice. Studies of mate choice in frogs have provided key empirical evidence for both sensory exploitation and the Handicap Hypothesis. This research has all concerned mate choice for acoustical cues. It remains an open question whether visual cues, which play a large role in mate choice in fish, and birds, are utilised by female frogs in any species. While nocturnal species might not be expected to use such cues, many species of frogs, particularly tropical, are partially or completely diurnal. Here we report evidence for mate choice based on visual cues in a poison frog (Dendrobates pumilio, family Dendrobatidae) from the Bocas del Toro Archipelago in Panamá, Central America. Infra-specific differentiation in the poison frog D. pumilio of the Bocas del Toro is one of the most dramatic cases of divergence among populations ever discovered. Different populations of this frog display completely different colouration and pattern from island to island and among isolated populations on the mainland. Myers and Daly argued that these different populations (with the exception of the Valiente Peninsula) are all members of a single species, because of similarities in calling characteristics. Recent genetic analysis supports their interpretation. We investigated whether females from two islands in this archipelago. Nancy Key (orange morph), and Pope Island (green morph) prefer members of their own population in the context of mate choice, when only visual cues are available. Material and Methods Females were captured in the field, and kept in terraria at the Smithsonian Tropical Research Institute in Panamá City. Females that were the subject of mate choice experiments were placed in single ten-gallon terraria supplied with leaf litter and plants and allowed to acclimatise to their environment for a week. Four such "choice chamber" terraria were maintained simultaneously. Mate choice experiments were carried out by placing one frog of each morph underneath an overturned glass at one end of the terrarium, on opposite sides (approximately 10cm. apart). The frogs were matched for snout-vent length (within 1 mm). During the trials, advertisement calls recorded in the field at Nancy Key were played from speakers either outside of the terraria (Nancy Key females) or inside the terraria behind the glasses containing the frogs (Pope Island females). For unknown reasons, the Pope Island (green) females were less responsive to calls played outside of the terraria than the Nancy Key (orange) females. Subject female response to each object frog was measured as the amount of time the female spent oriented toward each object frog once she had approached within 15 cm of the frogs, and as the amount of time she spent in close proximity with each object frog, during a one hour trial. To define close proximity, each overturned glass was placed in the centre of an overturned Petri dish, and the subject female was considered to be in close proximity if she had at least two limbs on the Petri dish. Responding females clearly perceived and oriented toward the object frogs under both lighting conditions. In most cases they would try to make contact with the object frogs through the glass, attempting to stroke them in a manner similar to that seen previously during courtship in this species. In one case the subject female actually oviposited at the base of the overturned glass containing the object frog, indicating that the female was in fact attempting to mate with that frog. In the first three trials (with orange morph females) male frogs were used as the object frogs in the choice experiments. Unfortunately, males would call when approached by the subject female, and their calls were clearly audible through the glass. Orange females displayed a preference for their own morph, but it was not clear that male calling behaviour was not influencing their choice. For this reason, subsequent experiments employed female object frogs, and the results reported here exclude the results using males. Object females under the glass did not reciprocate courtship by the subject female. The sides for each morph were switched between consecutive trials (that produced scoreable results), and the speakers were switched between trials for experiments with the green morph females (in which the speakers were in the terraria). Experiments under white light employed 40watt overhead fluorescent lights and 25-watt incandescent lights (referred to as white light). Control experiments for the effects of visual cues employed 60 watt incandescent lights shone through two types of spectral filters (BES Optics): overhead lighting using a BG 12 blue green filter, and side lighting using a BG 23 blue-green filter. The irradiance for these conditions could not be measured in Panama, but the conditions were recreated in the lab at the University of Maryland Biotechnology Centre and measured with a spectrometer. All trials were videotaped. Trials in which the female responded were scored for orientation and proximity by a researcher unfamiliar with these frogs, using a black and white video monitor so that differences in colour could not be perceived. The total amount of time spent oriented toward and in proximity to each frog was combined for analysis. Relative responsiveness was calculated as the response of the subject female to its own morph divided by the total time spent oriented toward or in the proximity of either frog. Results Females responded to the object frogs in only 27 out of 144 experiments (19%). In most cases, the subject female did not react to the calls, and did not approach the object frogs. This is consistent with the pattern of parental investment in these frogs, in which females provide the majority of parental care, and with previous research demonstrating that females are highly selective during mate choice in the field in this species. First, we compared the preference females displayed for frogs from their own island under white light, relative to the preference females displayed under blue light. Considering only those experiments where results were obtained for the same female under both white and blue light, females displayed a significant preference for frogs from their own island under white light (relative to blue light) in paired tests. Green females from Pope Island displayed a significant preference for their own morph. Orange females from Nancy Key did not display a significant preference for their own morph, although sample size was small and there was a trend in the predicted direction. Considering data from all experiments, including experiments under white light with females that did not respond under blue light, a significant preference was displayed by females from both islands combined, and by females from Pope Island. The females from Nancy Key also displayed a significant preference. Subject females displayed a significant preference for their own morph under white light, relative to a baseline of 50%, the result expected if they were choosing at random. In contrast, all subject females combined displayed a significant preference for the opposite morph under blue light. Separate analysis did not reveal a significant preference under blue light by females from Pope Island, or from Nancy Key. There was no significant tendency for females to prefer the frog placed on a particular side of the tank. For the females from Pope Island, there was no significant tendency to prefer the frog placed in front of a particular speaker in the tank. Finally, there was no significant relationship between the size (snout-vent length) difference between the two object frogs, and the preference expressed by the subject. Discussion These results demonstrate that females from each island prefer to court frogs from their own population, compared with members of another population. Under control conditions in which the width of the spectrum of visible wavelengths available to the frogs is reduced, females no longer exhibit this preference. This demonstrates that female preference under white light is made on the basis of visual cues. The most obvious difference between these morphs is colour. Unfortunately light intensity declined dramatically with the use of the blue filters, and we were not able to avoid this disparity. Hence, it is possible that females utilise some other visual cue to recognise members of their own population, which they could not detect under the darker conditions of the blue light experiments. Nevertheless, our experiments suggest that differences in colour are the most likely visual cues used by these frogs. Recent evidence indicates that Dendrobates pumilio has four different types of cones, and is capable of perceiving differences colour differences. This is the first experimental demonstration that visual cues are important to frogs during courtship. Visual stimuli were a likely candidate for mate choice cues in this species because they are diurnal, and mating involves a long, complex courtship sequence in which the female appears to follow the male using both visual and acoustic cues. If females do in fact prefer to court members of their own population based on colour differences, as suggested by the results of this study, then why should this be the case? There are two likely explanations: either divergence in colour drove divergence in female preference, or divergence in female preference drove divergence in colour. Divergence in colouration could drive differences in female preference through a process of sensory bias [Ryan, 1990 #108]: females may become attuned to the colour of the males from their local population, and may evolve a preference for that colour because they are more easily able to detect it. There is some circumstantial evidence to support the hypothesis that female choice may have driven colour divergence in these populations of D. pumilio, rather than the other way around. Compared with D. pumilio, Minyobates sp. and P. lugubris, two closely related species of dart-poison frogs that are sympatric with D. pumilio in the Bocas del Toro archipelago, show little variation in colour or pattern across their ranges. However, both display the bright, aposematic coloration characteristic of the dart-poison frogs. Variation in colour and pattern across different Minyobates and Phyllobates species provides evidence of the evolutionary potential for colour and pattern variety in both of these genera. Owing to common descent (family Dendrobatidae) and shared geography, Minyobates sp. and P. lugubris serve as natural controls in an analysis of divergence between populations of D. pumilio in the Bocas del Toro archipelago. Molecular phylogenetic and biogeographical evidence suggests that divergence in colour and pattern has been rapid among D. pumilio populations. Comparative analysis of inter-population mtDNA sequence divergence in D. pumilio, Minyobates sp. and P. lugubris did not indicate any pronounced differences in population histories that might explain the marked differences in intraspecific colour and pattern differentiation between the species. Hence, it is unlikely that differential time of divergence explains the dramatic difference between D. pumilio and the other two species in the degree of divergence among populations in colour and pattern. Dendrobates, Phyllobates and Minyobates all produce toxic alkaloids. The bright coloration characteristic of many Dendrobatids is believed to be aposematic in function, probably serving to warn visually oriented predators that have colour vision, such as birds. The D. pumilio populations differ from one another in the number and type of toxins they produce, but toxin characteristics do not appear to be correlated with the brightness of colour found in particular populations. The extreme island to island variation in coloration and pattern found in D. pumilio populations is puzzling given its presumed aposematic function, because natural selection should favour individuals that closely resemble other toxic frogs exposed to the same individual predators, whether they are from the same species or not (Mullerian mimicry). The lack of correlation between toxicity and coloration may occur because of a correlation between toxicity and diet. Recent research indicates that at least some dendrobatid toxins are accumulated through dietary uptake. However, this does not explain why D. pumilio populations have evolved marked colour differentiation, while P. lugubris and Minyobates sp. populations have not. It also does not explain the lack of a correlation between coloration and toxicity. Ecological differences seem unlikely to explain the difference between D. pumilio and the other two species in the amount of colour and pattern divergence, because all three species are diurnal, active frogs which live on the forest floor, and are frequently found within a few cm of each other. An alternative possibility is that differences between the species in parental investment and sexual selection are involved. In D. pumilio and closely related species, females carry tadpoles to small pools of water and return periodically to lay trophic eggs which the tadpoles require for survival. Males in D. pumilio provide relatively little parental care compared to females; males do not carry or feed tadpoles, but may periodically attend clutches for brief periods, shedding water on them. Phyllobates lugubris, like other members of its genus, has uniparental male care. Recent field research on Minyobates minutus, a closely related sister species of Minyobates sp., revealed that parental care is performed by the male (Summers, unpublished data). Sexual selection theory predicts that female choice for male ornamentation will be particularly strong in species in which females provide more investment in offspring than do males. In species with uniparental male care, female choice is constrained because male parental care exerts direct selection on female preferences, and rejection is more costly to females. In species where males provide little or no parental investment, these constraints are relaxed. Darwin first presented the argument that females may select for elaborate male traits. In turn, rapid divergence in sexually selected traits can occur because females may prefer different forms of the trait in different populations. One potential problem with this explanation is that most sexually selected characteristics are sexually dimorphic, whereas there is no obvious dimorphism in colour or pattern between male and female D. pumilio. However, there are two possible explanations for a lack of pronounced dimorphism in D. pumilio. First, it is possible that female colouration is genetically correlated with that of males, causing selection on male colouration to produce a correlated response in female colouration. Second, if bright colouration and contrasting patterns serve as warnings to predators in D. pumilio, then selection within populations may favour females with the same colour and pattern as males. Sexual selection associated with female parental care may have caused rapid divergence in colouration, and rapid speciation, in cichlid species flocks in Lake Victoria and Lake Malawi in East Africa. Species from these flocks are similar to the divergent populations of D. pumilio in that they show low degrees of genetic divergence, they inhabit ecologically similar or identical environments, and they differ primarily in colour and pattern. Molecular phylogenetic and biogeographical evidence suggests that divergence in colour and pattern has been rapid among D. pumilio populations. Divergent female preferences could account for this differentiation, and would not be expected to play a role in P. lugubris or Minyobates sp. populations, which show low divergence in colour and pattern. The results of this study, which suggest that females of each population prefer their own colour morph, are consistent with the hypothesis that sexual selection may have driven colour divergence among these populations of D. pumilio.