Testing fundamental evolutionary hypotheses

Abstract

Sober and Steel (J. Theor. Biol. 218, 395–408) give important limits on the use of current models with sequence data for studying ancient aspects of evolution; but they go too far in suggesting that several fundamental aspects of evolutionary theory cannot be tested in a normal scientific manner. To the contrary, we show examples of how some alternatives to the theory of descent can be formulated in such a way that they lead to predictions that can be evaluated (and rejected). The critical factor is a logical formulation of the alternatives, even though not all possible alternatives can be tested simultaneously. Similarly, some of the limits using DNA sequence data can be overcome by other types of sequence derived characters. The uniqueness (or not) of the origin of life, though still difficult, is similarly amenable to the testing of alternative hypotheses.

Introduction

Sober and Steel, in a recent contribution to this journal (2002) critically examine the “Hypothesis of Common Ancestry”—that all life on earth traces back to a single common ancestor. They rightly point out that this is accepted within biology without rigorous testing, and they present theoretical results as to why the theory may be difficult to test. Our paper is a response to their claims, and illustrates how the Hypothesis of Common Ancestry has been tested in the past, and how difficult aspects (such as whether more than one ancestral lineage contributed to modern life) can be tested further. Our conclusion is that the Hypothesis of Common Ancestry is testable in principle, and it is not intrinsically different from other scientific theories. Nevertheless, Sober and Steel's paper is an important challenge; the reliability of current methods for building evolutionary trees from DNA sequence data is rightly criticized, as is the notion of a single (lineage-based) origin of all modern life. Thus questions analysed by Sober and Steel are fundamental in evolution—issues too often neglected. What are the expected limits of reconstructing evolutionary trees from sequences? How many ancestors are there for life on earth? Are there tests that distinguish between single- and multiple-origin hypotheses? We accept fundamental points in their article but secondary problems distract from the key issues, and the problems identified are solvable by good science. We focus on four main themes:

• the theory of descent leads to testable predictions,

• science does not claim to have absolute tests of hypotheses,

• the limits to phylogeny reconstruction depend on the model, and

• are there one, or more than one, common ancestors of life?

Sober and Steel suggest that the hypothesis of common ancestry is so ingrained in the minds of biologists that, when attempting to reconstruct the relationships that link a set of species, “the typical question is which tree is the best one, not whether there is a tree in the first place” (Sober and Steel, 2002). Historically, this is certainly not the case; many forms of relationship between species are possible (Fig. 1) and there is no a priori reason to assume a Steiner tree (Fig. 2). The concept of species having a continuity through time was only developed in the late 17th century (and only after continuous spontaneous generation of complex organisms was invalidated, see Farley, 1977). Higher life forms were no longer thought to “transmute” into different kinds during the lifetime of an individual. Many proposals relating these new entities (species) are shown in Fig. 1 and/or discussed in Bowler (1984). It took over 2000 years (from the time of the ancient Greeks), and over 150 years from the concept of permanent species, before a rooted Steiner tree was proposed by Charles Darwin. Some of the earlier ideas had common ancestors for subgroups, others did not. By providing a mechanism (natural selection, and descent with modification), Darwin could suggest a scientific model (pattern and mechanism) for species relationships. Darwin's evolutionary tree was neither obvious, nor easy to find. We claim that any alternative (as in Fig. 1) is testable individually, but that it is logically impossible to compare any one hypothesis against “all possible alternatives” (including those not yet specified). Rejecting all possible alternatives is logically equivalent to “proving” the theory.