News and ViewsMorphometrics and phylogenetics: Principal components of shape from cranial modules are neither appropriate nor effective cladistic characters
Introduction
Recently, González-José et al. (2008) proposed that phylogeny estimation using principal components (PCs) from shape analyses of cranial ‘modules’ (modular cladistic approach: MCL) is an advance that could provide a benchmark for future studies of evolutionary relationships. Unfortunately, MCL ignores serious practical and theoretical issues concerning the use of PCs as characters in cladistic analyses, and there is no evidence that the method is an improvement over alternative methods. Here we address three interrelated issues with the proposed approach. First, we show why rank-order scores along PC axes do not provide appropriate characters for the phylogenetic analyses used, question the rationale behind selecting a subset of PCs from each module, and consider why the use of maximum likelihood does not overcome these issues. Second, we evaluate the findings against those from methods that are not expressly cladistic, grouping by overall rather than shared derived similarity. Third, we consider the effects of subdivision into modules. Our reanalyses show that despite the fact that they take no account of synapomorphy, topologies of unweighted pair group method with arithmetic mean (UPGMA; Rohlf and Sokal, 1981) phenograms from the whole skull and the modules differ no more from the published maximum parsimony (MP) and maximum likelihood (ML) cladograms than these differ from each other. Likewise, the vast majority of analyses of arbitrary subsets of landmarks (i.e., not based on a priori biological knowledge) from whole skulls and from arbitrary subdivisions of the landmarks from the whole into four modules support a cluster including all representatives of Homo, the “most remarkable result” of MCL (González-José et al., 2008: 776). This arises because there is such a strong pattern of overall similarity that even approaches such as UPGMA of arbitrary ‘modules’ yield the same result. Therefore, the MCL approach has not been shown to be phylogenetically more informative than groupings based on overall similarity using arbitrary modules.
Section snippets
Rank-order scores along principal component axes do not provide appropriate characters for phylogenetic analysis
Geometric morphometric shape variables provide quantification of biological form that conserves the geometry of the landmark configuration. As such, it is natural to consider whether such variables can be used to estimate phylogenetic trees (Macleod and Forey, 2002). However, there are important methodological reasons to be cautious when using such variables for phylogenetic inference. The primary analytical problem is how to use cladistic linear parsimony (whose input is typically a set of
MCL trees arise from a strong pattern of superficial similarity
Beyond the theoretical considerations above, here we demonstrate the lack of evidence of efficacy of MCL through reanalyses of the data in González-José et al. (2008: their supplementary material). A UPGMA phenogram (Rohlf and Sokal, 1981, Rohlf, 2008) based on the whole set of landmarks, which ignores the concepts of ‘character’ and ‘character polarity’ and instead is based on shape distances between OTUs, shows groupings which are very similar to the clades apparent in the MP and ML trees (
Modules are elusive and arbitrarily defined modules perform as well as those of González-José et al. (2008)
The MCL method is based on the identification of relatively independent modules. However, evolutionary covariation among face, cranial base, and vault (the independent modules postulated by González-José et al. [2008]) has been shown to be much stronger than previously expected by developmental modular organization (Bookstein et al., 2003, Lieberman et al., 2008, Mitteroecker and Bookstein, 2008, Perez and Monteiro, 2008). This is likely because of the elusive nature of evolutionary modules in
Conclusion
While theoretical considerations alone raise severe objections to the modular cladistic approach proposed by González-José et al. (2008), our reanalyses of their data confirm the lack of efficacy of the MCL approach. Our conclusion is that results based on the new method approximately agree with results from alternative approaches, and to some degree with previous phylogenetic hypotheses because of a strong pattern of superficial similarity. Indeed, the geometric morphometric analyses presented
References (32)
Can biometrical shape be a homologous character?
- et al.
Cranial integration in Homo: singular warps analysis of the midsagittal plane in ontogeny and evolution
J. Hum. Evol.
(2003) - et al.
Partial-warps, phylogeny, and ontogeny: a comment on Fink and Zelditch (1995)
Syst. Biol.
(1998) - et al.
Combining and comparing continuous morphometric descriptors with a molecular phylogeny: the case of fruit evolution in the Bornean Lithocarpus (Fagaceae)
Syst. Biol.
(2001) - et al.
Does the skull carry a phylogenetic signal? evolution and modularity in the guenons
Biol. J. Linn. Soc. Lond.
(2008) - et al.
Phylogenetic and environmental components of morphological variation: skull, mandible, and molar shape in marmots (Marmota, Rodentia)
Evolution
(2005) Quantitative characters, phylogenies, and morphometrics
- et al.
Phylogenetic analysis of ontogenetic shape transformations: a reassessment of the piranha genus Pygocentrus (Teleostei)
Syst. Biol.
(1995) - et al.
Continuous characters analyzed as such
Cladistics
(2006) - et al.
Cladistic analysis of continuous modularized traits provides phylogenetic signals in Homo evolution
Nature
(2008)
Epigenetic interactions and the structure of phenotypic variation in the cranium
Evol. Dev.
Spatial packing, cranial base angulation, and craniofacial shape variation in the mammalian skull: testing a new model using mice
J. Anat.
Phylogenetic analysis of the correlated evolution of continuous characters: a simulation study
Evolution
The evolutionary role of modularity and integration in the hominid cranium
Evolution
Why morphometrics is special? The problem with using partial warps as characters for phylogenetic inference
Syst. Biol.
Cited by (56)
New in morphometry: Geometric morphometry of the external female genitalia of Triatominae (Hemiptera: Reduviidae)
2022, Acta TropicaCitation Excerpt :PCA is widely used for exploratory analyses of shape (Rohlf et al., 1993). The technique commonly applied in geometric morphometry allows the use of multivariate data to estimate macro-evolutionary events (Valente et al., 2001; Adams et al., 2011; Anderson et al., 2007). The results showed that the shape variables of the EFG are decisive for the differentiation between the taxa.
Geometric morphometric discrimination between seven populations of Kawakawa Euthynnus affinis (Cantor, 1849) from Peninsular Malaysia
2022, Journal of King Saud University - ScienceCitation Excerpt :Body shape and size is the main tool for documenting, in particular morphological variations, variations in shape and size, and for assessing taxa relationships based on changes in body shape, even among closely related species of fish (Imtiaz and Md Naim, 2018). The biological shape is quantified by geometrical morphometric variables that preserve geometric configurations of landmarks (Adams et al., 2011). The results of this study showed that morphological homogeneity existed between communities from Peninsular Malaysia's east coast (Terengganu, Kelantan, and Johor).
Variability of Australopithecus second maxillary molars from Sterkfontein Member 4
2015, Journal of Human EvolutionCitation Excerpt :This evidence broadly supports Curnoe (2010), who assigned both specimens to the genus Homo. The extent to which results of a PCA based on Procrustes shape or form coordinates can be used to address taxonomic issues is surely differently regarded by different authors (see discussions in Adams et al., 2011; Bookstein, 2014, 2015). In this study we use geometric morphometrics in order to quantify, and possibly characterize, the morphology of the Australopithecus M2 sample.
EvoPhylo: An r package for pre- and postprocessing of morphological data from relaxed clock Bayesian phylogenetics
2023, Methods in Ecology and EvolutionMultiple modes of inference reveal less phylogenetic signal in marsupial basicranial shape compared with the rest of the cranium
2023, Philosophical Transactions of the Royal Society B: Biological Sciences