Experimental studies of coral snake mimicry: do snakes mimic millipedes?

The prevalence of similarly ringed colour patterns in Neotropical snake species has stimulated heated controversy over the cause of this apparent convergence since it was first noted by Cope (1860) and Wallace (1870) in the last century. These ringed and banded species include venomous (neurotoxic) elapid coral snakes (Micrurus) and a number of mildly or non-venomous colubrid and anniliid snakes. The most commonly accepted explanation for this convergence in colour pattern is coral snake mimicry (Greene & McDiarmid 1981; Pough 1988). This hypothesis, wherein the non-elapid species are protected through resemblance of the true coral snakes, is now supported by comparative and laboratory and field experimental evidence (Smith 1975, 1977; Greene & McDiarmid 1981; Pough 1988; Campbell & Lamar 1989; Brodie 1993; Brodie & Janzen, in press). However, there are still untested alternatives to the coral snake mimicry hypothesis, including convergence on a common colour pattern because of crypsis or aposematism for unprofitability as prey, or even selective neutrality of ringed patterns (for a recent review, see Pough 1988). Perhaps the strongest alternative explanation is that all ringed or banded snakes (and lizards) may actually be mimics of other banded elongate animals such as millipedes (Vitt 1992). Most Neotropical ringed snakes are slender, and even as adults are similar in diameter to large millipedes (e.g. Campbell & Lamar 1989). The argument has been made that millipedes are commonly banded, produce toxic compounds for defence (Eisner et al. 1978), are locally common throughout the tropics and share predators in common with small reptiles. Experimental evidence suggests that some captive birds and mammals learn to avoid millipedes after experiencing their noxious secretions (Eisner et al. 1978). Additionally, millipedes have been present during the entire evolution of vertebrates (since the Silurian; Almond 1985), and thus could represent the original model for mimicry systems involving elongate ringed or banded prey (Vitt 1992). Before the ‘millipede as model’ hypothesis can be evaluated, some basic data must be collected. At present, no information is available on the frequency of predation by visual predators on ringed millipedes, or on the ability of such predators to distinguish between patterns exhibited by millipedes and those found on snakes. To argue that millipedes are the true model for coral snake mimicry complexes, it must be established that (1) millipedes and snakes have visual predators in common, (2) visually foraging predators avoid millipedes and (3) these predators do not distinguish between millipede and snake ringed patterns. We conducted a field experiment to determine whether millipedes are avoided by free-ranging, visually foraging predators (birds, see below) and if such predators distinguish between the ringed patterns found on snakes and millipedes. We constructed plasticine millipede-sized models that differed only in coloration. Models were then exposed to predation attempts by free-ranging birds to determine the relative attack frequency on models with different colour patterns. The techniques and materials used to produce the models are described elsewhere (Brodie 1993; Brodie & Janzen, in press). Three patterns were tested (Fig. 1): unmarked brown, red and black narrow ringed (‘millipedes’), and red and black wide ringed (‘coral snakes’). The two ringed patterns differed only in ring width, one corresponding to ring widths found on the bicolour coral snake in the study area (Micrurus multifasciatus; Campbell & Lamar 1989), the other corresponding to ring widths characteristic of ringed millipedes