The study of geographic disjunctions is inextricably linked with the timing of lineage splitting. This presentation will summarize some of the principal issues and methods pertaining the inference of evolutionary rates and timing of lineage divergence on the basis of molecular sequence data, and in the context of an explicit phylogenetic hypothesis. The adequacy of substitution models should be assessed, particularly regarding the need to include parameters to accommodate among-site rate variation. Rate variation among lineages is an important source of error in age estimates. The assumption of rate constancy should be tested either through relative rate tests, or through global tests, which may rely on a comparison of observed vs. estimated branch lengths, or on the likelihood values of models of evolution that differ only in the presence/absence of rate constancy. Information from geological events or the fossil record can be incorporated into a phylogeny to provide a general chronological calibration, or minimum- or maximum-age constrains to nodes. Fossils should be evaluated as to whether they provide the stem group or crown group age of a given clade. Reconstruction of divergence times among lineages under rate constancy may be straightforwardly achieved through predictions from a weighted linear regression analysis of inferred distances against calibration times, or through the use of maximum likelihood and a rate-constant substitution model on a given phylogeny. Promising approaches to estimating divergence times in the absence of rate constancy have been recently developed. A non-parametric method, based on the assumption of rate autocorrelation, minimizes local shifts in rates from branch to branch. A semi-parametric method combines a different substitution rate in each branch with a penalization for excessive variation. Examples of some of these methods and of the effect of node constrains, based on lineage divergence among major seed plant clades, will be presented.

Key words: branch lengths, fossil record, lineage-effects, non-parametric rate smoothing, penalized likelihood, rate variation