WAGNER, P.J.; The Field Museum, Chicago: Ancestral species in the fossil record

Given that sampling and species durations both are finite, it is inevitable that we will sample ancestor-descendant pairs from the fossil record. How many we expect to find depends on sampling rates, which vary considerably among and within taxa, and other factors such as modes of speciation. From a phylogenetic perspective, adding unsampled ancestors to phylogenies inevitably increases the likelihood of a tree. This is because positing an unsampled ancestor: a) encompasses a number of possible phylogenies spanning the number of possible character combinations and thus sums the probabilities of numerous specific phylogenetic paths, and; b) often reduces the need for parallelisms and reversals, which increases the probability of trees if rates are low. Because of this, and because adding unsampled ancestors adds parameters to a phylogenetic tree, “sampled ancestor” represents a null hypothesis relative to “unsampled ancestor.” A related issue concerns the tempo of character evolution within a clade. Here, continuous change is the null hypothesis relative to speciational (or punctuated) change because continuous change posits the same rate of change during branching and within lineages, whereas speciational change posits different rates of change for between and within lineages. One can determine the most likely trees given continuous and speciational change. One also can determine when adding unsampled ancestors ceases to significantly improve phylogeny likelihoods based on information criteria. Besides improving phylogenetic inference, these approaches test whether particular parameters used as models in other analyses (e.g., continuous change or no ancestors contemporaneous with descendants) are common patterns in the fossil record.