Feeding experiments with lizards are used to examine the function of small eyespot markings found along the wing margins of many butterfly species. Such eyespots are frequently suggested to function by deflecting the attacks of vertebrate predators away from the vulnerable body towards the wing margins, which can tear easily; the eyespots are considered to mislead predators and to act as targets for their attacks. Such misdirected attacks give the butterflies a chance to evade capture, albeit sometimes losing pieces of wing tissue. As a model prey species, we used fruit-feeding individuals of the tropical butterfly Bicyclus anynana that were attacked in a standard way in laboratory cages by the anolis lizard, Anolis carolinensis. We also manipulated the butterflies' wing patterns by pasting eyespots on different parts of the wings to examine the deflection hypothesis in more detail. Our results indicate no influence on the lizard attacks either of the presence of eyespots, or of their position on the wings. The lizards attacked butterflies in a highly stereotyped manner both when the prey were presented on matching or on contrasting backgrounds. We thus found no support for the deflection hypothesis for attacks by insectivorous lizards. Indeed, our only support to date has been obtained for naïve flycatcher birds, but even this requires further corroboration. Although effective deflection may occur rather infrequently, except perhaps under certain ecological conditions such as high-density feeding of butterflies on fallen fruit, it may still be sufficiently consistent over time to have contributed to shaping the evolution of marginal eyespot patterns.
Numerous species of butterflies and moths possess conspicuous, colourful eyespots on one or both sides of their wings. These are believed to function as an antipredator mechanism and several hypotheses exist to explain their function (Stevens, 2005). The intimidation hypothesis proposes that many dorsal eyespots mimic vertebrate eyes and are a component of startling displays to deter predators from attacking. In resting butterflies, such eyespots are hidden from view but they can be presented in a sudden way by unfolding the wings. These features may be combined with other behavioural elements to form a ritualized display (Edmunds, 1974; Vallin et al., 2005, 2006). Some authors argue that a startling type of function can also be associated with ventral eyespots on the forewing that are hidden behind the hindwings in resting butterflies, but which can be suddenly exposed by flicking up the forewing (Brakefield, 1984; Dennis, Porter & Williams, 1986). Blest (1957) showed that circular patterns consisting of concentric coloured rings were most effective in startling predators. Critics point out that intimidation may be an avoidance of novel features (neophobia; Marples & Kelly, 2001) or simply be due to the markings' conspicuousness, arguing that our interpretation of them resembling eyes need not apply (Dziuraweic & Deręgowski, 2002).
In the present study, we focus on another potential function for some eyespot markings that is implicit in the deflection hypothesis (Stevens, 2005). Many Lepidoptera possess wing eyespots that are usually not hidden from view in resting individuals; these eyespots are often smaller, less conspicuous and usually positioned close to the edge of the ventral wing surfaces. Such marginal eyespots are considered to function as deflective devices, at least sometimes misdirecting predator attacks away from the vulnerable body to the wing margin that is comparatively fragile and readily tears away (DeVries, 2002; Hill & Vaca, 2004). This type of deflective hypothesis has been prominent since at least the late 19th century when it was discussed by Poulton (1890). The idea also draws on the related phenomenon of the ‘false head’ wing pattern that is most common among the lycaenid butterflies. Here, a pattern in the anal angle of the hindwings combines eyespot-like markings, bands radiating across the wings towards the anal angle and filamentous and antenna-like extensions on the edge of the wing, with behavioural components that together seem to create the illusion of a false head. The complex pattern created by this whole suite of traits is more conspicuous than the real head, and is also considered to misdirect predator attacks thus enabling prey-escape (Poulton, 1890; Van Someren, 1922; Robbins, 1980, 1981; Wourms & Wasserman, 1985; Tonner et al., 1993; Cordero, 2001).
Both lizards and birds can be important predators of butterflies, perhaps especially in the tropics (birds: Chai, 1996; lizards: Brockie, 1972; Young, 1979, 1980; Ehrlich & Ehrlich, 1982). Many authors have interpreted ventral eyespots as having some deflective role (Blest, 1957; Young, 1979; Brakefield, 1984; Brakefield & Reitsma, 1991), sometimes supporting their ideas with observations of wing damage (and associated behaviour) occurring in the wild (Dennis et al., 1986; Tonner et al., 1993) or found among museum specimens (Young, 1980). Such wing damage frequently involves missing pieces of wing consistent with tearing away from birds' or lizards' beaks and not with general wear and tear (Robbins, 1980). However, Edmunds (1974) and others have correctly drawn attention to the severe problems in interpreting beak-mark data. Marginal eyespots involved in deflection might tend to be positioned in parts of the wing such as the hindwings that are least sensitive to loss in terms of butterfly flight performance (Dennis et al., 1986). In addition, these wing parts may have evolved to be more fragile and to tear readily (DeVries, 2002), especially in butterflies with deflection markings (Hill & Vaca, 2004).
Although the deflection hypothesis has received substantial interest, very few definitive observations have been made on predator attacks in association with butterfly eyespots. Wourms & Wasserman (1985) offered dead Pieris rapae L. butterflies painted with different elements of the false head pattern of lycaenid butterflies to blue jays, Cyanocitta cristata L., in an aviary. They observed more attacks directed towards these elements if an eyespot-like marking was included. Our study follows on from the observations made by Lyytinen, Brakefield & Mappes (2003) and Lyytinen et al. (2004), who used both birds, Ficedula hypoleuca Pallas, and lizards, Anolis carolinensis Voigt, as model predators of the fruit-feeding butterfly, Bicyclus anynana Butler, that has marginal wing eyespots. Their work provides some weak support for the deflection hypothesis but only when naïve birds are concerned. Using the same Anolis lizards but a more standardized experimental design, we attempt here to determine whether butterfly eyespots could influence the attack behaviour of the lizards. Young's (1979, 1980) observations on tropical Morpho butterflies suggested that attacks while certain species were feeding on fruit on the forest floor could account for their conspicuous marginal eyespots. Other species more restricted to the forest canopy and not known to feed on fallen fruit showed eyespot patterns that were much less well-developed. We therefore studied the attacks of lizards on live butterflies while they were feeding on pieces of fruit.
MATERIAL AND METHODS
Bicyclus anynana (Satyrinae) is a small, brown, satyrine butterfly. Our laboratory stock was established from material collected in Malawi, Africa where warm rainy seasons alternate with cool dry seasons.Bicyclus anynana, like other species of Bicyclus found in the wet–dry tropics, exhibits strong seasonal polyphenism (Fig. 1); a conspicuous wet-season form (WSF) has large marginal ventral eyespots whereas an inconspicuous dry-season form (DSF) lacks these eyespots (Brakefield & Larsen, 1984; Brakefield & Reitsma, 1991; Brakefield & Frankino, 2007). These alternative forms can be produced readily in the laboratory by rearing larvae at a high or low temperature, respectively (Fig. 1).
Anolis carolinensis (Iguanidae) is a small, diurnal lizard found in the south-east of the USA and the Caribbean area (Roughgarden, 1995) where it inhabits the trunk and crown of trees (Fite & Lister, 1981). It is a general insectivore (Sexton, 1964), and males are larger than females.Anolis lizards have acute vision, relying heavily on movement to find prey (Goodman, 1971). They filter out irrelevant motions in the surroundings (Fleishman, 1986). Their eyes are bifoveal (Fite & Lister, 1981) with a field of vision of approximately 340°. Movements made through as little as 0.22° attract lizard attention and elicit the visual grasp reflex (Fleishman, 1986, 1992). They are able to discriminate colours (Shafir & Roughgarden, 1994; although see Tiemann, 1972), and their vision probably extends into the ultraviolet (UV) (Fleishman, Loew & Leal, 1993).Anolis lizards have taste buds (Schwenk, 1995) and can recognize obnoxious prey, learning to associate colour with unpalatability (Sexton, 1964; Shafir & Roughgarden, 1994; Stanger-Hall et al., 2001; Sword, 2001). However, their sense of smell is not highly developed or important behaviourally (Stanger-Hall et al., 2001). Although the distributional ranges of our two model species do not overlap, we emphasize that this study focuses on the general occurrence of eyespots on butterfly wings. Marginal eyespots are characteristic of satyrine species, many of which, such as species of Euptychia of similar size to B. anynana, live in the natural habitat of Anolis lizards (D'Abrera, 1988). Furthermore, lizards of the families Codylidae, Chameleontidae, Gekkonidae and Agamidae live in the natural habitat of B. anynana in Africa and are wholly or partially insectivorous (Webb, Wallwork & Elgood, 1978; Hedges, 1983; Branch, 1988). The latter two families of lizards employ a similar sit and wait strategy as A. carolinensis (Pianka, 1977; Stamps, 1977).
We used 24 A. carolinensis lizards obtained via Dutch retailers more than a year before the experiments. Groups of three individuals (one male and two females) were kept in cages (46 × 33 × 51 or 57.5 cm; length × width × height) in a climate room at the Leiden laboratory and labelled on their back for identification using a felt-tip pen. Temperature varied in the range 25–27 °C up to approximately 31 °C directly under lamps positioned over the cages; these lamps also emitted UV light. Relative humidity was approximately 70% with a 12 : 12 h light/dark photoperiod. Water was available ad libitum and the plants in the cages were misted twice daily. Bark chips covered the cage floor and three bamboo perches were provided. Twice a week before and between observational periods, the lizards were fed buffalo worms, Alphitobius laevigatus F., dusted with calcium or vitamin supplement and sometimes complemented with live crickets, Gryllus bimaculatus De Geer.
Observations were made in a large (244 × 90 × 85 cm; length × width × height) wooden-frame cage, covered with nylon netting, and divided into two equal halves separated by a cardboard partition. In each compartment, one group of lizards was observed at a time. The bottom of the cage was covered with a layer of soil and brown autumn leaves. The cage contained two potted plants, a shelter constructed from three bricks and a Petri-dish with water. Butterflies were presented to the lizards by using a construction consisting of a small board attached to the top of a pillar over which the lizards approached. A piece of bamboo was attached to one end of this board, serving as butterfly perch. It was orientated perpendicular to the board so that the butterflies always perched with their sides and wings facing the approaching lizard. The other end of the board extended to the leaves of a plant where the lizards often rested. Two long pieces of bamboo halfway along the board led to the sides of the cage.
Groups of lizards were placed in each compartment of the cage approximately 20 h prior to the start of an observational period to acclimatize. They had not fed for 3 days. Butterflies also had no access to food (fruit) for 12 h prior to the experiment to increase their readiness to feed. An observational period lasted 1.5 h and started with a butterfly being introduced on to the perch to which a piece of banana had already been smeared. The small adjustments in position made by a butterfly when feeding were usually sufficient to attract the attention of a lizard.
Butterflies with differing wing patterns were offered in random order. Each butterfly was replaced once it had flown away or been eaten. Observation ceased if the lizards had not attacked a butterfly during 1 h. The observer sat behind a one-way mirror to record attacks on videotape that was analysed later.
The butterflies used for these observations had manipulated wing patterns. During the pilot experiments normal WSF and DSF B. anynana were offered. We found that the butterflies were always struck by the lizard either on the body or on the wings. We used the four different wing margins to further categorize the location of attacks: the costal and outer margins of the forewing, and the outer and anal margins of the hindwing (abbreviated as FCM, FOM, HOM and HAM, respectively). To further analyse the effect of a butterfly eyespot on the lizards' attack behaviour, we pasted eyespots on to two different locations of the wings of live butterflies. We used butterflies of an artificially selected ‘Low’ line with an extreme DSF pattern to paste eyespots on to. The large posterior eyespot on the ventral forewing was cut from the wings of dead WSF butterflies and pasted on to the wings of Low-line butterflies using a little heated candle wax. This was done at 5 °C to minimize handling difficulties and hence wing damage. The eyespot was either attached to the anal angle of the hindwing (in addition covering the normal forewing eyespot, which is hidden when at rest, with a piece of brown wing tissue; HS), or pasted on to the forewing over the position of the posterior eyespot (and pasting brown wing tissue on to the anal angle of the hindwing; FS) (Fig. 2). Untreated and treated dead butterflies inspected under a UV-lamp at wavelengths of 254 and 366 nm showed no visible artefacts due to any manipulation.
Three experiments were performed, during experiment 1, the lizards were offered FS and HS butterflies in random order in the observational cage. For experiment 2, we obtained an additional nine young lizards (three males and six females). Although the lizards were field collected they had had no recent experience of butterfly prey and, thus, it was assumed that they were all naïve or less experienced with such prey. The animals were allowed to acclimatize for three weeks in somewhat smaller cages (31 × 25 × 55 cm; length × width × height) in the same climate room as the other lizards, with every cage again containing one male and two females fed as in the main group. After the acclimatization period, they were also fed FS and HS butterflies in random order. For experiment 3, the experienced lizards were offered FS and HS butterflies in front of a matching brown background in case the eyespots were then more contrasting and thus more effective as misdirecting targets. This background was constructed from a transparent plastic tray (16 × 11 × 6 cm; length × width × height) fixed vertically but slightly tilted to a metal stand and placed behind the butterfly perch. It was filled with the leaf mixture from the bottom of the cage covered by chicken wire (the latter hidden by a few sprinkled leaves). Some leaves were also placed just in front of the butterfly so that, from the lizards' perspective, it would appear to be within a carpet of brown leaves. This construction could obscure the butterfly, depending on the position of the lizard in the cage, so that fewer attacks were expected to occur.
Anolis carolinensis lizards attack in a highly stereotyped manner, sneaking up to the butterflies slowly and then leaping at them suddenly from a rather standard distance of approximately 5 cm. Female lizards would often start bobbing their head up and down during the course of approaches to the prey and, although the literature refers to it as ‘wavering’ (Greenberg & Noble, 1944), the function of this behaviour in a feeding context is unknown. There was a sexual difference in attack behaviour with females tending to approach more haltingly with the ‘wavering’ behaviour, whereas males usually attacked more directly. Prey were always consumed whole and head first.
The lizards took butterflies in experiment 1 at a similar rate as in the pilot experiments (chi-square test: χ2 = 1.98, d.f. = 1, not significant; Fig. 3). They also reacted to the prey during experiment 1 in a similar manner, thus showing no aversion to the candle wax. All prey taken were eaten without a single rejection. The two manipulated forms (FS and HS) were almost always grabbed by the lizards at the same location, namely the costal or leading margin of the forewing (FCM; Fig. 4). Only two attacks struck the hindwing anal margin (HAM).
The naïve or less experienced lizards in experiment 2 attacked more butterflies than those with recent experience of these prey (χ2 = 5.89, d.f. = 1, P < 0.05; Fig. 3). They also behaved differently in other respects. Several ate the banana instead of attacking the butterfly (even sometimes apparently having to shove the butterfly aside to reach the banana). Others just approached close to the butterflies before retreating. Some individuals first ate banana and on later occasions attacked butterflies, others did the reverse whereas several attacked butterflies without delay. The butterflies that were attacked were all struck on the FCM, and no rejections occurred (Fig. 4).
Attacks on butterflies presented on a cryptic background in experiment 3 were also almost always aimed at the FCM. A single butterfly was grabbed at the FOM, but this individual was also in a different position relative to the approaching lizard. Significantly fewer attacks occurred during experiment 3 in comparison with experiment 2 (χ2 = 10.41, d.f. = 1, P < 0.01) or the pilot experiments (χ2 = 5.92, d.f. = 1, P < 0.05), but not when compared to experiment 1 (χ2 = 1.49, d.f. = 1, not significant; Fig. 3).
The Anolis lizards used for these observations showed a highly stereotyped way of attacking the butterflies, nearly always grabbing them at the costal margin of the forewing, independent of eyespot location (FCM; Fig. 4). Even when butterfly prey were presented against a background that generally matched the background wing colour and was likely to increase the contrasting pattern of the eyespots, nearly all attacks were directed at the FCM (a frame-by-frame playback of video records of the two attacks to the HAM made in experiment 1 suggested that these attacks were also aimed at the FCM but that the butterfly tried to escape at the final moment). Thus, our observations provide no support for the deflection hypothesis because the mode of attack appears to be completely uninfluenced by the wing pattern of the butterfly. Once a lizard detects a butterfly prey by any minor adjustment of the body or wings when feeding, it appears to attack without giving any attention to eyespots or other wing patterning. This applies equally to both experienced and less experienced individuals.
The highly predictable manner in which the lizard attacks were aimed at the costal margin of the forewing in our experiments also suggests that lizards may not be very important in the evolution of the false head patterns of lycaenid butterflies (Robbins, 1980; but see Van Someren, 1922). However, in Brockie's (1972) experiments in which caged Lacerta muralis lizards were fed Maniola jurtina butterflies, the lizards often grabbed butterflies on the hindwings, allowing them to escape with damaged wings. Thus, different species of lizard may have differing modes of attack such that, although not important with respect to Anolis, the false head patterns are effective against certain other species. An alternative explanation is that the individuals of M. jurtina were taken when the forewings with ventral eyespots had been withdrawn between the hindwings (Brakefield, 1984), a behaviour that B. anynana does not exhibit.
There was some difference in feeding behaviour of male and female lizards in our experiments. Males were more direct in their attacks, usually approaching the butterfly in one continuous movement, whereas females showed a less direct behaviour with some head-bobbing (Greenberg & Noble, 1944). In the pilot experiments, this appeared to result in the females missing significantly more butterflies than males (a parallel but nonsignificant trend also occurred in the later experiments).
We also found that less experienced lizards attacked a higher proportion of butterflies than the experienced individuals. This is probably not fully explained by an effect of novelty because both groups were offered butterflies with manipulated wing patterns for the first time. The procedure of pasting pieces of wing tissue onto DSF-butterflies had no effect on the lizards' attack rate and no rejections occurred, suggesting that our methodology did not affect prey acceptability. Individual lizards of the less experienced group varied in their response to the butterfly prey. Some never attacked the prey (even eating the banana presented with the butterflies), others learned to attack via several tentative attempts. This reluctance showed elements of neophobia (Marples & Kelly, 2001) and supports their naïve status with respect to prior experience of butterflies.
The present study, together with earlier work by Lyytinen et al. (2003, 2004), provides no indication that the attacks of insectivorous lizards on butterflies at rest are likely to be misdirected towards ‘target’ eyespots at the wing margins. It is possible that attacks of birds are more likely to be deflected by wing eyespots although Lyytinen et al. (2004) found only some weak support for naïve flycatchers, F. hypoleuca. Wourms & Wasserman (1985) found that the handling behaviour in the bill by blue jays was influenced by eyespot-like markings painted onto the wings of dead butterflies. Blest (1957) also found that yellow hammer birds misdirected attacks to eyespots painted on mealworms. Some further indirect support for the deflection hypothesis comes from beak mark studies and from measurements of wing strength in butterflies (DeVries, 2002; Hill & Vaca, 2004). Butterflies carrying beak marks can sometimes occur at high frequencies in populations.
It is clear that further research is needed to substantiate the role of the deflection hypothesis for butterfly eyespots along the wing margins. This may help to explain associations such as those between the presence of eyespots and feeding patterns in tropical Morpho butterflies (Young, 1979, 1980), or between the wet–dry-seasonal forms of tropical butterflies and their alternative ecological environments (Brakefield & Larsen, 1984; Brakefield & Frankino, 2007). Our results suggest that such research should focus on birds rather than lizards. It is not possible to estimate how frequent the misdirection of attacks from potential predators by marginal wing eyespots is in the field. It could be infrequent except perhaps under specific ecological circumstances such as in high-density feeding by butterflies on fallen fruit (Young, 1979, 1980). Nevertheless, overall, it may be sufficiently consistent and effective over time to have played an important role in the evolution of butterfly wing patterning.
Niels Wurzer and Mariël Lavrijsen maintained the lizards and provided practical help. Jeanette Bot generously made her video camera available to us. We also thank the group of Herman Berkhoudt for providing vital input on the experimental setup, We are grateful to David Field, Carlos Cordero, two anonymous referees, and John Allen, and especially Caroline Müller, for their constructive criticism.