DISCUSSION AND CONCLUSIONS (continued)

2. Why are so few wing patterns actually successful in nature?

There are possible answers to this question such as "few of the extant racial varieties are able to form zones of contact in nature." While that is true, synthetic introgression among just a few adjacent races of cydno and melpomene which can and do hybridize, generate much more variety of pattern than occurs in nature in races and species of that clade. For example, Figure 4 (upper left) shows a small sample of possible varieties which quite likely have occurred at one time or another in central Costa Rica. So the fact that realized pattern diversity of described taxa is well below what we know is possible given the potential of known hybrid zones presents an interesting question. It is abundantly clear that even in relatively short evolutionary time frames, any lack of pattern variety is not due to any developmental genetic constraint on pattern evolution!

A few reasons that so few of the possible Heliconius wing patterns are found to exist in nature include

a. sensory bias in predators: Patterns that survive are the best signals in terms of recognition and learning. This hypothesis is certainly testable knowing both relevant predators and the extent of heritable pattern variety. Mimicry rings may be constrained to a few modal color patterns because of the way light environments in different habitats interact with the ability of predators to perceive, discriminate, and learn visual signals.

b. adaptive peaks are limited: In most areas at any given time common, aposematic species are few and those few would create a large selection gradient in favor of a small fraction of possible hybrid zone patterns. This idea could be tested for example, by introducing identical interracial F1 cydno into several isolated sites, each harboring abundant Passiflora host, naive birds, and no other Heliconius. Each site would receive prior releases of model phenotypes (different for each site) to experimentally create different adaptive peaks. The same variable F2 population emerging at each site should evolve a different monomorphic pattern predicted by the experimental model for that area. Such an experiment could be conducted ancillary to using Heliconius cydno in biocontrol of the pest liana of Hawaiian forests, Passiflora mollissima (Wagge et al. 1981).

c. auto adaptive peaks: The rules of meiosis, epistasis, and dominance will insure that in a hybrid zone some phenotypes will outnumber others, all other things equal. One possibility is that isolated hybrid populations may be initiated in disturbed patches temporarily free of predator pressure. Later, when predators colonize the patch they quickly learn the most common phenotype which thus forms its own adaptive peak (e.g., Kapan 1998). This possibility is testable by reviewing the genetic rules that govern pattern in interracial and interspecific crosses, checking whether successful patterns tend to be those that would be the most abundant in the absence of selection. Note that in a typical hybrid zone between mimetic races with predators constantly present, novel alleles flowing into an area are more likely to persist if they are recessive (e.g., Mallet et al. 1990), just the reverse of the scenario suggested.

d. geographical constraints: Areas ideal for the operation of evolutionary processes like biotic drift or shifting balance may be limited. Geomorphology and climatic history determine where patterns can be stable and where conditions will promote revolution.