Abstract

A nested series of phylogenetic analyses provide a hierarchical framework with which to test the evolution of phenotypic covariance matrices. Phylogenetic relationships of the Neotropical rodent subfamily Sigmodontinae and tribe Phyllotini are reviewed, based on published phylogenetic analyses of morphological characters conducted in conjunction with this dissertation. These analyses present a working hypothesis of sigmodontine relationships, and identified a Phyllotis darwini species group for more detailed analysis. Phylogenetic relationships and species limits within Phyllotis are estimated using nucleotide sequence data from a 973 base-pair region of the mitochondrial cytochrome b gene. Ten taxa belonging to eight biological species were sampled in Phyllotis, along with five outgroup species. Geographic variation and species limits were assessed by sequencing multiple individuals from different populations for most species. Parsimony analyses were conducted using six different combinations of weighting schemes that incorporated information on transition/transversion bias and codon position. Criteria for choosing among alternate analyses and sequence-based weighting schemes are discussed. The results support a clade including Phyllotis and Auliscomys but not Graomys. The analyses also support the darwini species group and a northern species-pair, P. amicus and P. andium. Similar to prior morphology-based phylogenies, interspecific relationships within Phyllotis are not well resolved. However, the sequence-based phylogeny and morphometric data indicate that P. xanthopygus rupestris is polyphyletic, and that Pacific slope populations from southern Peru currently assigned to that taxon are specifically distinct from altiplano and eastern Andean representatives of P. xanthopygus. Correlation and variance/covariance matrices are calculated for twenty-eight populations belonging to six Phyllotis species. Pairwise matrix correlations of correlation matrices are consistently very high (>0.90) and show no significant association between matrix similarity and phylogenetic relatedness. Hierarchical decomposition of common principal components analyses applied to each clade in the phylogeny rejects the hypothesis that common principal components structure is shared in clades more inclusive than subspecies. Several subspecies also lack a common covariance structure as described by the common principal components model. A modified minimum-evolution approach is used to estimate covariance matrices for hypothetical ancestors in the species-group phylogeny. Branch lengths are estimated as the mean disparity in corresponding ancestor-descendent covariances. Branches are longest leading to terminal populations and subspecies, while interspecific branches are relatively short, indicating a general conservation of covariance structure among species despite a high degree of intraspecific variability. Absolute deviations in covariances are not correlated with phenotypic divergence. In sum, the analyses suggest that deviations in covariance structure are most strongly associated with the formation of diagnosably distinct taxa and stochastic sampling of genotypes at the population level. There is no evidence for significant accumulation of divergence in covariance structure beyond that attained by reproductively isolated taxa.