Simple shell measurements do not consistently predict habitat in turtles: a reply to Lichtig and Lucas (2017)

Inferring palaeoecology for fossils is a key interest of palaeobiology. For groups with extant representatives, correlations of aspects of body shape with ecology can provide important insights to understanding extinct members of lineages. The origin and ancestral ecology of turtles is debated and various shell or limb pro-portions have been reported to correlate with habitat ecology among extant turtles, such that they may be informative for inferring the ecology of fossil turtles, including early shelled stem turtles. One recently described method proposes that simple shell measurements that effectively quantify shell doming and plastron width can differentiate habitat classes among extant turtles in linear discriminant analysis, whereby aquatic turtles have low domed shells with narrow plastra. The respective study proposes unorthodox habitat predictions for key fossil turtles, including aquatic lifestyles for the early turtle Proganochelys quenstedtii and the meiolaniform Meiolania platyceps, and terrestrial habits for the early turtle Proterochersis robusta. Here, we show that these published results are the consequence of questionable methodological choices such as omission of species data which do not conform to a preconceived shell shape-ecology association. When these choices are reversed, species measurements for fossils are corrected, and phylogenetic flexible discriminant analysis applied, habitat cannot be correctly predicted for extant turtles based on these simple shell measurements. This invalidates the method as well as the proposed palaeohabitats for fossils.