Figure 3

A graphical representation of the species boundary. The numbers of natural hybrids known between pairs of species (from Table 1) are plotted on a logarithmic scale against the average uncorrected DNA divergence estimated from data for 1569 bp of mtDNA [30]. If backcrosses are also known from wild specimens, a halo around the point is shown. Comparisons reflect only species that have zones of overlap; average distance measures are given in Additional File 3. There are no known hybrids between species groups, and no estimates of divergence have been included for intergroup comparisons (Neruda and Laparus are here treated as part of the melpomene-cydno- silvaniform group to which they are closest in mtDNA divergence). A least-squares exponential fit of the species data alone is shown. (To display species pairs which lack known hybrids on the log-linear plot, they have been assigned 0.1 hybrids each, but the fitted line is based on a non-linear regression with untransformed data). Because the comparisons are non-independent, especially where branches of the same phylogeny or even the same species are used twice, a simple statistical analysis is not appropriate (under an assumption of independence, there is a highly significant negative correlation between in rates of hybridization and genetic distance: N = 180, P = 0.0022, although the proportion of the variance explained is not high, r² = 5%, because of the large number of species pairs for which no hybrids are known). Intraspecific hybridization also approximately fits this scheme; smaller square points in blue represent the equivalent numbers of intraspecific hybrids in world collections (not used in curve fitting). These were estimated by counting the numbers of intraspecific hybrids (between morphologically divergent subspecies) in the 2001 catalogue of the W. Neukirchen collection, and dividing by the fraction of total interspecific hybrids in the Neukirchen collection over the total known worldwide.