Was Australopithecus anamensis ancestral to A. afarensis? A case of anagenesis in the hominin fossil record

Abstract

We tested the hypothesis that early Pliocene Australopithecus anamensis was ancestral to A. afarensis by conducting a phylogenetic analysis of four temporally successive fossil samples assigned to these species (from earliest to latest: Kanapoi, Allia Bay, Laetoli, Hadar) using polarized character-state data from 20 morphological characters of the dentition and jaws. If the hypothesis that A. anamensis is ancestral to A. afarensis is true, then character-state changes between the temporally ordered site-samples should be congruent with hypothesized polarity transformations based on outgroup (African great ape) conditions. The most parsimonious reconstruction of character-state evolution suggests that each of the hominin OTUs shares apomorphies only with geologically younger OTUs, as predicted by the hypothesis of ancestry (tree length = 31; Consistency Index = 0.903). This concordance of stratigraphic and character-state data supports the idea that the A. anamensis and A. afarensis samples represent parts of an anagenetically evolving lineage, or evolutionary species. Each site-sample appears to capture a different point along this evolutionary trajectory. We discuss the implications of this conclusion for the taxonomy and adaptive evolution of these early-middle Pliocene hominins.

Introduction

Recent debate concerning the shape of the hominin phylogenetic tree—the relative importance of anagenesis vs. cladogenesis in the generation of morphological diversity—has focused on early Pliocene and late Miocene discoveries that have extended the hominin fossil record back to the 5–7 Ma interval in which DNA evidence suggests chimpanzee and human lineages diverged (Stauffer et al., 2001). Much of the record prior to 3.5 Ma remains poorly sampled, however, creating the opportunity for widely divergent interpretations of lineage diversity near the base of the hominin clade (e.g., Tattersall, 2000, Wood and Richmond, 2000, Senut et al., 2001, Leakey et al., 2001, Brunet et al., 2002, Wood, 2002, White, 2002, Haile-Selassie et al., 2004). A theoretical commitment to cladogenesis as the major, if not the only, basis for macroevolutionary change supports interpretations of recent discoveries as evidence of high taxonomic diversity early in the hominin fossil record, whereas the view that anagenesis is capable of generating species-level differences interprets the early record of hominins as supporting only one or at most two lineages prior to the widely accepted proliferation of lineages observed close to the Plio-Pleistocene boundary. Progress toward resolution will come from a more thoroughly sampled fossil record over time and space and more completely documented ranges of sample variation, accompanied by the rigorous testing of phylogenetic hypotheses.

One area where existing samples can already be examined in this context is the hypothesized ancestor-descendant relationship between Australopithecus anamensis (4.2–3.9 Ma) and its successor in eastern Africa, A. afarensis (3.6–3.0 Ma).¹ The A. anamensis hypodigm consists of approximately 80 (mostly craniodental) fossils from two Kenyan sites, Allia Bay and Kanapoi (Leakey et al., 1995, Leakey et al., 1998, Ward et al., 2001). The A. afarensis hypodigm comprises more than 400 fossils, the great majority of which are from Laetoli, Tanzania, and Hadar, Ethiopia, with more than 90% of the total coming from the latter site (Johanson et al., 1982, Kimbel et al., 1994, Kimbel et al., 2004).² The two hypodigms contrast strikingly in comparable aspects of mandibular, maxillary, and dental anatomy (Ward et al., 2001), and there is little doubt that, on purely phenetic grounds, they can be distinguished effectively at the species level. This conclusion is underscored by the fact that morphological variation is well documented in the extensive Hadar sample, but this variation does not diminish the observed differences between A. anamensis and A. afarensis (Kimbel et al., 2004).

It is widely accepted that A. anamensis was directly ancestral to A. afarensis (e.g., Leakey et al., 1995, Wolpoff, 1999, Ward et al., 2001, Gibbons, 2002, White, 2002). This view is attributable to two factors: the known temporal range of A. anamensis completely antedates that of A. afarensis, and, insofar as the two differ anatomically, A. anamensis is described as the more plesiomorphic taxon (Ward et al., 2001). Each of these factors warrants closer scrutiny.

Although the temporal positions of the two species' hypodigms do not overlap, each hypodigm itself consists of two temporally disjunct site-samples (Fig. 1). The A. anamensis sample from Kanapoi (∼4.20–4.17 Ma) predates that from Allia Bay (∼3.9 Ma), while the Laetoli sample of A. afarensis (∼3.7–3.5 Ma) predates the one from Hadar (∼3.4–3.0 Ma). With the four site-samples thus ordered in time, it is possible to examine the sequence for temporally vectored morphological change. Leakey et al. (1995; see also Ward et al., 2001) initiated such an examination and found some differences among the samples, noting in particular that A. anamensis tends to resemble the older Laetoli sample of A. afarensis phenetically more than it does the younger Hadar sample. Lockwood et al. (2000) identified temporal trends in A. afarensis that lend statistical weight to the differences between the Laetoli and Hadar site-samples.

These observations constitute a testable hypothesis about ancestry and descent among the site-samples. To test this hypothesis, we conducted a formal phylogenetic analysis that treated each site-sample as an operational taxonomic unit (OTU) bearing polarized morphological character-states. If the hypothesis that A. anamensis was ancestral to A. afarensis is true, then character-state changes between the temporally ordered site-samples should be congruent with hypothesized polarity transformations based on outgroup conditions. That is, when employing the site-samples as OTUs in a cladistic analysis, each of the three geologically older site-samples should have only younger samples in its sister group, and no sample save the terminal one, from Hadar, should have autapomorphic states.³ Nor should there be derived states shared by some of the in-group OTUs as homoplasies.

Our combined stratigraphic-phylogenetic approach to testing a hypothesis of ancestry and descent derives from common principles underlying the identification of ancestors in a phylogenetic framework. Specifically, it is equivalent to the procedure outlined by Smith (1994: 131–132) for using cladistic practice to determine whether older, plesiomorphic phena in the fossil record can be considered to represent populations from which the ancestors of progressively younger, more derived phena were drawn. Some researchers might expect ancestors never to be identified if cladogenesis is the primary mechanism for generating macroevolutionary change, but if that is true in this case, we would expect to see evidence in the form of autapomorphic traits in older OTUs.

Failure to reject the hypothesis would support the view that anagenetic change in a lineage is sufficient to explain morphological differences among these site-samples. On the other hand, the discovery of autapomorphic (or homoplastic) states in older, putatively ancestral site-samples would suggest a more complex, “bushy” evolutionary history. Either result is important for determining the appropriate taxonomy for these samples, for resolving the relationships of A. anamensis and A. afarensis to other hominins, and for interpreting the process of adaptive evolution in these Pliocene hominins.