, , , , ,

So, I got my hands on that mouth watering Nature paper I mentioned a couple days ago. It is titled, “Cladistic analysis of continuous modularized traits provides phylogenetic signals in Homo evolution,” and it is probably the biggest anthropology news of this week. I’ve read it and it is dense. It really shouldn’t be so dense because what the authors ultimately did was a cladistic analysis on 17 of the most complete hominin fossil skulls. What really makes this study different is in the traits they have quantified and compared for their study. I’ll do my best to translate it into plain English.

First, let me define the methodology behind this paper, cladistics, also known as phylogenetics, is a way to organize entities. It can be used to organize really anything — such as tools like hammers, wrenches, pliers, etc. In this case, cladistics was applied on hominin species. Cladistics ultimately organizes things based on upon their ‘shared derived characteristics.’ Shared derived characteristics is an ambiguous term. It really means the traits that are seen in members of the same species, which are unique from other species, i.e. all orangutans have red hair whereas chimpanzees do not, making them different. When dealing with fossils, researchers can only compare and contrast measurements from the bones, such as the size of the brain case or the length of a bone.

I’ve written on several occasions how cladistic analysis works great for grouping different organisms that are really derived. It gets dicey when cladistics tries to resolve organization of organisms that show a lot of variation within one another, as seen in a recent discussion on H. floresiensis. The authors of this paper acknowledge this as a primary motivation for their research. They write,

“However, completion of cladistic analysis becomes problematic because, at lower taxonomic levels, most of observable variation is expressed as continuous changes of size and shape rather than in discrete identifiable structures.”

This problem arises when analyzing hominin fossils, especially fossils thought to be members of the genus Homo. All members of the genus Homo share a large number of derived morphological similarities, that make them different from australopithecines and other apes. What are they? A strictly bipedal adult stature and the largest brain size of all primates are a couple. But, because collections of Homo fossils show a lot of variation in the traits, it makes it difficult to validate such questions like if Homo neanderthalensis was more bipedal that H. heidelbergensis. This has made discerning what traits are unique to each species of Homo troublesome. The authors also acknowledge this. In their own words,

“these [discretization methods] disregard the continuous nature of many complex morphological traits.”

I’ll get to what that really means later in the next paragraph, but first let me also outline another related shortcoming in using cladistics to figure out what’s going on within Homo. See, in the process of trying to find discrete traits in each fossil hominid, researchers have often relied on quantifying multiple traits and presenting them as a univariant measurement. According to the authors, this “disregards the multivariant and geometric nature of form.”

Yesterday, I shared a perfect example of this problem — Yoel Rak identified several traits in the Neandertal mandible thought to be unique to them. He was pownced upon for disregarding that the Neandertal mandible size may have increased as a function of the expanding skull. This problem is not isolated to just Rak, it seems to be prolific within paleoanthropology. Functional traits are often separated from developmental traits, when in fact they are really integrated. The authors address ‘a logical approach’ to overcome this problem,

“treat integrated features as a single phylogenetic complex, and to treat the complex as if it were an independent character.”

The authors applied their approach, selecting four measurements of the skull that integrate functional and developmental traits. The four measurements the authors quantified are the flexure of the cranial base, facial retraction, neurocranial globularity, and the shape and position of the masticatory apparatus. If you’re curious to see exact what these measurements are, the authors provided a figure that illustrates them. I’m putting it right below this paragraph. These traits are believed to be constrained to evolve in a coordinated fashion, rather than independent from one another.

17 skulls were analyzed. John Hawks expressed beef with the sample size, but they are all very complete skulls. Folding in incomplete specimens woulda relied on inferring more missing measurements — which I don’t particularly think is the most conservative approach to figuring out hominid evolution. The 17 skulls are A.L. 444-2, Sts 5, KNMER-406, OH 5, SK 48, WT 17000, KNMER 1813, KNMER 3733, Zhoukoudian, D2700, Steinheim, Kabwe a.k.a. Broken Hill 1, Atapuerca 5, Gibraltar 1, La Chappelle-aux-Saints, and La Ferrassie 1. A skull of a gorilla, chimpanzee, and modern human were also included in the analysis.

In any cladistic test outgroups must be established. Since we know gorillas and chimpanzees existed before the emergence of hominids, they were considered part of the outgroup — or the ancestral form to compare and constrast shared derived characteristics. Australopithecines, like members of the genus Paranthropus (KNMER-406, OH 5, SK 48, WT 17000) and Australopithecus (A.L. 444-2, Sts 5), were also considered in the outgroup because the scope of this study was to resolve the phylogeny of Homo. To reiterate, the function of establishing an outgroup in phlyogenetics is that the outgroup branched from the parent group before the other two groups branched from each other. Respectively, the ingroup (all the Homo skulls) are more closely related to each other than any single one of them is to the ‘outgroup.’

Casts of the skulls were digitized and the measurements were made. A Procrustes analysis was applied to normalize the data, removing variations in translation, rotation and scaling that woulda occurred in digitizing them. The statistical test used was a principal component analysis (PCA), a test that simplifies the variation described by many variables into a few. I kinda see a flaw in this, PCA has often been criticized for quantifying multiple traits and presenting them as a univariant measurement… a concern the authors raised in their introduction.

Anyways, two phylogenetic trees were constructed from the PCA, one applying a maximum parsimony algorithm and another applying a maximum likelihood algorithm. The parsimony algorithm organizes data based upon the premise of having least number of evolutionary changes. Maximum likelihood algorithm is much more statistically viable, it focuses on integrating probability — the normal distribution of a certain variable. Both trees constructed from the data were similar, differing only in the relative position of H. sapiens in relation to the complex H. erectus, H. ergaster, and H. rhodesiensis.

I’ve put the both figures of the tree on this post for you to evaluate. The top figure, a, is the parsimony tree and the bottom figure, b, is the likelihood tree. The coding may not be very straight forward… basically the recipe for deciphering the coding is the first capital letter represents the genus and the next three lower case letters represent the species. If you have any questions with that, comment here, and I’ll do my best to help you out.

The results aren’t very surprising. H. habilis is the ‘founder’ of the Homo clade. The press went wild over the conclusion Neandertals are distinct lineage from the H. erectussapiens lineage, which is misleading — we already have a lot of evidence supporting that. Something that’s worth pointing out is that H. heidelbergenis is more related to Neandertals than the H. erectussapiens group, which kinda makes sense, no H. erectus specimens have been found east of Turkey, to my knowledge, indicating that a group of Homo were unique to Europe, prior to the influx of H. sapiens — which could have been H. heidelbergenis.

Both trees show some curious organization in the australopithecines, they fail to form a monophyletic group, which is another hot debate in paleoanthropology. This indicates the australopithecines are paraphyletic. Both trees show that Australopithecus africanus is sister group of the genus Homo. This ain’t a very surprising conclusion because A. africanus was gracile and more like Homo… it is thought to have been a direct ancestor of modern humans and this kinda confirms it.

Even though the conclusions are not Earth shattering, I do want to point out that the scope of this paper was not to totally reorganize our understanding of human evolution. In fact, the authors advocate that paleoanthropologists be more cautious in considering what we measure, factoring that some traits need to be considered as modules. It is still possible to recover and reconstruct relevant phylogenetic signals from these modules. In their own words,

“valuable phylogenetic information is recovered from data sets that consider independence on a developmentally and functionally basis, and which preserve the multivariant and continuous nature of complex phenotypes.”

    González-José, R., Escapa, I., Neves, W.A., Cúneo, R., Pucciarelli, H.M. (2008). Cladistic analysis of continuous modularized traits provides phylogenetic signals in Homo evolution. Nature DOI: 10.1038/nature06891