In the series of experiments employing a total of 603 birds, we have tested specific predictions of the eye-mimicry and conspicuousness hypotheses, two competing hypotheses that have a bearing on our understanding of how eyespots may have evolved in nature. The data do not unequivocally favor either hypothesis. Figure 3 summarizes results from all experiments. However, the results of our study question the generality of existing paradigms and augment our understanding of what properties make eyespots effective against predation.
In the first experiment, models with a pair of eyespots were attacked fewer times than the eyespot-less model. Birds also took a relatively longer time to attack the model with a pair of eyespots. These results strongly corroborate the intimidation hypothesis. However 10 eyespots with the same total area as the pair of eyespots did not confer significantly different protection compared to the spotless models.
The second experiment explicitly tested the effect of the size of the signal, where birds were offered a choice between a pair of natural sized eyespots and a pair of smaller eyespots. The results showed no significant difference in predatory attacks between models, and birds attacked both models equally quickly. Stevens et al. [22] found that the total area of the conspicuous signal determines eyespots’ effectiveness, irrespective of the number of eyespots (1, 2 or 3). They hence concluded that their data support the conspicuousness hypothesis, which predicts intimidation to increase with salience of the signal. Results of our first two experiments are in marked contrast, and provide no support for this prediction. In Experiment 5, where a single large eyespot was pitted against a pair of eyespots with the same total area, models with smaller eyespots experienced lesser predation, again in discord with findings of Stevens et al. [22].
A previous study [26] using wings of real butterflies and birds as predators, reported no difference in latency to attack between prey with four and two approximately equally sized eyespots. We note that both studies that have not found an effect of size or total area have employed real eyespots with a single bird species in controlled, indoor experiments, whereas an effect of size was found only in the outdoor experiments with highly simplified patterns and where the effect observed is that of the predator community composed of unknown predators. Further studies are needed to understand under what conditions and predatory systems the total area of eyespots makes a significant difference.
The third experiment explicitly investigated the importance of structural resemblance to eyes for deterrence. We found that models with a pair of eyespots and a pair of fans suffered similar number of attacks. With both signals being equally conspicuous, the eye-mimicry hypothesis predicts that the effectiveness of eyespots will be reduced once the shape of the eye is lost. Therefore, data from this experiment do not support the eye-mimicry hypothesis.
The results of the fourth experiment indicate that shape asymmetry does not hinder the effectiveness of eyespots. Results from previous studies that tested the effect of asymmetry have been mixed. A negative effect of asymmetry was found in Forsman et al.’s [55] study, where artificial prey were presented to chicken. However, field experiments by Stevens et al. [21] demonstrated no effect of asymmetry. Given these confounding results, more work is needed to understand the relevance of symmetry of eyespot-like patterns.
The fan-like structures used in the current study were made from cut-out sectors of the original circular eyespot, which might have decreased their possibility to appear markedly different from the original ones. Furthermore, many birds took a very short time (typically less than a second) to peck on the food once they had sighted it. This could possibly result in “blurring” of the patterns. Thus we cannot discount the possibility of the differences between the eyespots and fans being too subtle to evoke different reactions from birds.
Interestingly, in the first four experiments, all models that rendered effective deterrence possessed a pair of stimuli, eyelike or non-eyelike, arranged bilaterally symmetrically on the two wings. The last experiment was aimed to determine if being in a pair is pivotal. Indeed, the results showed that birds avoided the model with paired signals more often than they did a single, large eyespot. Scaife [56] reported that birds appeared to avoid paired, eye-like stimuli more than they did a single stimulus. However, all stimuli in his experiment were equally sized, and therefore the former had a larger total stimulating area. In a subsequent experiment, Jones [57] found that two eyelike stimuli elicited more avoidance compared to one or three stimuli of the same size, and concluded that pairedness of eye-like stimuli is critical, perhaps by ‘completing the eye-gestalt’. Jones and other authors [57-59] have argued that two is a biologically significant number in pattern signalling. Although recent studies by Stevens and colleagues [22] have contradicted this, our study strongly supports the importance of pairedness of conspicuous eyespot-like patterns for deterrence.
We stress that we have only tested asymmetry in shape, but not positional asymmetry. For example, the birds may respond differently to two eyespots which are not centered together on the same latitudinal or longitudinal axis. Furthermore, the influence of the position of an eyespot pair relative to the body will be worth investigating.
Latency to attack the second model after the first
In Experiment 1, where spotless models were compared against models with a pair of natural eyespots, the duration between first and second attacks was greater when the second model had eyespots, indicating that eyespots increased the hesitation to attack. The remaining tests (all comparing models with conspicuous patterns) demonstrated no significant differences in latency to attack the second model (Additional file 1: Figure S2). We surmise that once a bird encounters eyespots, it will become more hesitant to attack, at least momentarily. This is corroborated by our visual observations on the behaviour of birds during trials.
Implications for the evolution of eyespots
The current study is the first study to demonstrate that artificial, eyespot-like structures can reduce predation as a result of innate aversion by birds. Chicks used in the study had been reared from birth in the poultry farm and hence naive in terms of exposure to predators. It follows that the aversion towards eyespot-like structures is innate, rather than based on learning and association of eyespots with danger. Another study that used naive laboratory reared birds has reported aversion to eyespots, but those found on real butterfly wings [26].
Although our data are seemingly in support of the eye-mimicry hypothesis, we stress that we have found no strong evidence for either hypothesis. We surmise that conspicuousness, while being very important under many conditions, is not necessarily the single most important factor determining the effectiveness of eyespots. The results suggest that resemblance to a pair of eyes enhances the effect of conspicuous stimuli much more than the total area (or size) does. However, the structural resemblance of individual patterns to a vertebrate eye need not be perfect.
Our study has implications for the evolution of eyespots from the perspectives of development and predator perception. Large, presumably intimidating, eyespots are more abundant on the dorsal wing surfaces in many lepidopteran groups such as Junonia and Bicyclus [1,28]. Developmentally, a pair of intimidating eyespots may be easier to evolve on the dorsal surface of butterflies since both pairs of wings (i.e. left and right) are visible to the predator when the wings are held open. Therefore, the evolution of a single eyespot on one wing on the dorsal surface (either hindwing or forewing) results in a symmetric pair of eyespots (on either side of the body), which could significantly enhance the effectiveness of eyespots. When the butterfly rests with wings held together, the ventral surface of a single pair of wings (a forewing and a hindwing) are visible, making it relatively more difficult for a pair of large eyespots to evolve on the ventral wing surface.
Furthermore, it is plausible that complex eyespots started evolving as much simpler markings on the dorsal surface. Even imperfect initial markings could have provided some benefits by being present as a pair, as is suggested by Experiment 3, thereby compensating for any cost in terms of increased risk of being detected by predators. The critical question of why highly intricate, apparently three-dimensional, eyespots such as those found in Junonia have evolved remains incompletely answered. Perhaps these eyespots are also used as signals in communication between sexes, or in the context of intra-sexual conflict.
Although the results in our study agree with previous work [26], they clearly conflict with aspects of some other studies [22,34,53], highlighting the importance of receiver bias (i.e. effect of type of predator) and the environmental context. Experiments by Stevens and colleagues were conducted outdoors, where survival probability was affected not by a single predator, but a community of predators. It is likely that different predators perceive eyespots differently, and thus the parameters that influence the effectiveness of eyespots against predation vary with the predator in question. A fruitful direction for further investigation is to understand the influence of these parameters against different kinds of predators. In addition to quantifying attack frequency and latency, using other communication modalities, such as sound, could furthermore help delineate both hypotheses. Although we did not hear any audible alarm calls eliciting fear, there have been studies with evidence for fear based on alarm calls emitted by predators upon noticing eyespots on prey [18]. Under natural conditions, predation pressure is typically exerted by a predator community, and hence it is possible that different kinds of eyespots and eyespot configurations might be effective in different geographic regions or seasons [28]. This might account for some of the diversity of eyespot patterns found in nature. Similarly, the habitat of the prey may strongly influence the evolution of eyespot-like patterns.