Introduction

In a very influential paper Avise et al. [1] introduced the term ``phylogeography'' to refer to evolutionary studies lying at the interface of population genetics and systematics. An important property of molecular sequences is that the degree of difference among them contains information about their relatedness. Avise et al. proposed combining information derived from the phylogenetic relationship of molecular sequences with information about where the sequences were collected from to infer something about the biogeography of relationships among populations within species. Figure  provides an early and straightforward example.

Figure: A phylogeographic analysis of 75 bowfins, Amia calva, sampled from the southeastern United States. A. A parsimony network connecting the 13 mtDNA haplotypes identified from the sample. B. The geographical distribution of the haplotypes.

The data are from bowfins, Amia calva, and consist of mtDNA haplotypes detected by restriction site mapping. There are two highly divergent groups of haplotypes separated from one another by a minimum of four restriction site differences. Moreover, the two sets of haplotypes are found in areas that are geographically disjunct. Haplotypes 1-9 are found exclusively in the eastern portion of the range, while haplotypes 10-13 are found exclusively in the western part of the range. This pattern suggests that the populations of bowfin in the two geographical regions have had independent evolutionary histories for a relatively long period of time. Interestingly, this disjunction between populations west and east of the Appalachicola River is shared by a number of other species, as are disjunctions between the Atlantic and Gulf coasts, the west and east sides of the Tombigbee River, the west and east sides of the Appalachian mountains, and the west and east sides of the Mississippi River [2].

Early analyses often provided very clear patterns, like the one in bowfins. As data accumulated, however, it became clear that in some species it was necessary to account for differences in frequency, not just presence versus absence of particular haplotypes. We saw this in the application of AMOVA to mtDNA haplotype variation in humans. These approaches have two critical things in common:

• Haplotype networks are constructed as minimum-spanning (parsimony) networks without consideration as to whether assuming a parsimonious reconstruction of among haplotype differences is reasonable.¹

• The relationship between geographical distributions and haplotypes contains information about the history of those distributions, but there is no formal way to assess different interpretations of that history.

Nested-clade analysis (NCA) has become a widely used technique for phylogeographic analysis because it provides methods intended to assess each of those concerns [4].² In broad outline the ideas are pretty simple:

• Use statistical parsimony to construct a statistically supportable haplotype network.

• Identify nested clades, test for an association between geography and haplotype distribution, and work through an inference key to identify the processes that could have produced the association.

As we'll see, implementing these simple ideas poses some challenges.³