With 1530 nucleotides, haplotypes connected by up to 17 substitutional steps had a 95% probability of a parsimonious connection. The minimum-spanning network contained five clades connected by greater than 17 substitutional steps, three within coastal populations and two within inland populations (Fig. 3). To analyze the haplotype network using nested-clade methods, clades must contain both genetic and geographic variation. Of the two inland clades encompassing 95% prob-able connections, only nesting clade 4-1 exhibited sufficient geographic variation to conduct statistical tests of patterns of association (Figs. 3, 4). Within clade 4-1, clades 2-3, 2-4, 2-5, 1-2, and 2-2 contained no geographic variation and were excluded from individual analysis; the haplotypes they contain were incorporated at higher levels of nesting (e.g., haplotypes L, K, and BB were incorporated into clade 3–1). The nested-contingency analysis, which did not incorporate geographic distance, indicated clustering of genetic variation with geographic location for clades 1-1 and 4-1 (P 5 0.014; P , 0.001). When geographic distance was included in the analysis, haplotypes C, A, and D within clade 1-1 showed restricted geographic ranges (Fig. 5; P 5 0.016, P 5 0.004, and P 5 0.014, respectively). Haplotype A also showed a limited distance from the geographic center of the clade (Fig. 5; P 5 0.017) and haplotype E was distant from the clade center (Fig. 5; P 5 0.003). Clade 1-3 had no significant values of Dc or Dn. Within clade 3-2 tip clades were more restricted in their distribution than interior clades (Fig. 5; P 5 0.029). Clade 3-1 exhibited a restricted range and was distant from the center of clade 4-1 (Fig. 5; P 5 0.001 and P 5 0.037, respectively).
The AMOVA indicated that both coastal (populations 15– 23) and inland (populations 1–14) groups exhibited genetic structure; 88.7% of genetic variation within the coastal group and 82.7% of genetic variation within the inland group was partitioned among populations (P , 0.001 and P , 0.001,respectively). Significant differentiation occurred even within clade 4-1, which encompasses 14 populations yet has only 0.003 substitutions per site mean divergence (FST 5 0.571, P , 0.001).