The Reconstruction of Ancestral Character States

Abstract

during the past decade evolutionary biology has seen a resurgence of interest in this approach (e.g., Ridley, 1983;Coddington, 1988;Carpenter, 1989;Brooks and McLennan, 1991;Baum and Larson, 1991;Harvey and Pagel, 1991). This increased appreciation for the role of historical processes can be attributed at least in part to recent major conceptual and methodological advances in the field of systematics (Hennig, 1966;Eldredge and Cracraft, 1980;Nelson and Platnick, 1981;Wiley, 1981;Felsenstein, 1982;Farris, 1983;Ax, 1987;Sober, 1988;Hillis et al., 1994).Responding to this trend, Frumhoff and Reeve (1994) examine both the usefulness and reliability of phylogenetic information in the study of evolutionary adaptation. In addition to scrutinizing specific methodological proposals, these authors conclude that the inference of ancestral character states from the distribution of character states present in descendant species is highly error-prone.Because the inference of ancestral character states is central to the phylogenetic study of evolutionary patterns, this criticism, if accurate, has important negative T.R. Schultz, R. Cocroft, and G.A. Churchill Ancestral State Reconstruction Page3 consequences for historical studies of behavioral, ecologicat and morphological evolution.Here we focus on the issue of whether or not the distribution of character states within extant species provides information about ancestral states. We begin by arguing that evolutionary homology provides the central concept for understanding this issue. We extend the model contributed by Frumhoff and Reeve (1994) for the neglected problem of error in ancestral character-state inference, focusing on the two evolutionary mechanisms identified by those authors: random and concerted homoplasy. By imposing confidence levels, we demonstrate that Frumhoff and Reeve's (1994) reported error rates are unduly pessimistic. We discuss the parameters that must be quantified in order to judge the likelihood of the general mechanism for parallel evolution proposed by those authors, and conclude with a discussion of how the model might be made more realistic. HomologyThe difference between homology and homoplasy lies at the heart of the problem of reconstructing ancestral character states. In phylogeny reconstruction the assignment of the same character state to two or more species constitutes a hypothesis of homology, i.e., the character states are hypothesized to be similar due to an unbroken chain of inheritance from a common ancestor that also possessed that state. When a number of such presumed homologies are combined into a phylogenetic analysis, a cladogram is identified in which the number of evolutionary origins of hypothesized homologies is minimized or in which their joint probability is maximized (as the sole criterion in "maximum parsimony" methods, or under additional a priori constraints in various other methods).T.R. Schultz, R. Cocroft, and G.A. Churchill Ancestral State Reconstruction Page4When most a priori homology assessments are correct, dis...

The Role of Subjectivity in Reconstructing Ancestral Character States: A Bayesian Approach to Unknown Rates, States, and Transformation Asymmetries

“…Put simply, in the extreme case in which there are a large number of descendant species and in which transitions A ® B and B ® A are equiprobable (symmetrical exchange), when enough time t has passed, we expect to observe a roughly 50% mixture of As and Bs in the N descendant species rather than a uniform xation of one or the other state. With this approach and a given rate matrix, Schultz et al (1996) found the probability to be nearly 1.00 that the ancestral state is the same as the descendant state when N = 8 (i.e., the reconstruction under both parsimony and maximum likelihood methods). Using the same model and parameter values, but assuming that higher rates of change were plausible, Frumhoff and Reeve (1994:Fig.…”

“…1). Various probabilistic methods (including maximum-likelihood approaches) for determining the relative reliability of ancestral state assignments have been proposed, all depending on various model assumptions about character evolution (e.g., Frumhoff and Reeve, 1994;Maddison, 1995;Schluter, 1995;Schultz et al, 1996;Schluter et al, 1997;reviewed in Cunningham et al, 1998). These general models require plausible values for the parameters describing the probability of change between character states, and these values can either be estimated from the data or derived from independent sources.…”

The Assumptions and Challenges of Ancestral State Reconstructions

“…These latter three assumptions of parsimony reconstructions are more subtle or implicit assumptions that for the most part have been neglected (but see Felsenstein, 1973;Maddison and Maddison, 1992;Swofford and Maddison, 1992;Collins et al, 1994;Frumhoff and Reeve, 1994;Schultz et al, 1996;Omland, 1997;Cunningham et al, 1998;Ree and Donoghue, 1998). There are probably many characters that do not meet one or more of these assumptions.…”

“…A related point is that the time between successive nodes must not be too great for a given rate of character evolution. Schultz et al (1996) showed that more-rapidly evolving characters are reconstructed with less accuracy. Schluter et al (1997) con rmed the dependence of accuracy on rates by using maximum likelihood approaches.…”

“…But although ancestral reconstructions have been embraced by both systematists and evolutionary biologists, the reliability and assumptions of these methods need continued examination. An increasing number of authors have discussed the assumptions and challenges of character mapping (Swofford and Maddison, 1992;Frumhoff and Reeve, 1994;Maddison, 1994;Ryan, 1996;Schultz et al, 1996;Cunningham et al, 1997;Omland, 1997;Cunningham et al, 1998).…”