A biased fossil record can preserve reliable phylogenetic signal

Abstract

The fossil record is notoriously imperfect and biased in representation, hindering our ability to place fossil specimens into an evolutionary context. For groups with fossil records mostly consisting of disarticulated parts (e.g., vertebrates, echinoderms, plants), the limited morphological information preserved sparks concerns about whether fossils retain reliable evidence of phylogenetic relationships and lends uncertainty to analyses of diversification, paleobiogeography, and biostratigraphy in Earth's history. To address whether a fragmentary past can be trusted, we need to assess whether incompleteness affects the quality of phylogenetic information contained in fossil data. Herein, we characterize skeletal incompleteness bias in a large dataset (6585 specimens; 14,417 skeletal elements) of fossil squamates (lizards, snakes, amphisbaenians, and mosasaurs). We show that jaws + palatal bones, vertebrae, and ribs appear more frequently in the fossil record than other parts of the skeleton. This incomplete anatomical representation in the fossil record is biased against regions of the skeleton that contain the majority of morphological phylogenetic characters used to assess squamate evolutionary relationships. Despite this bias, parsimony- and model-based comparative analyses indicate that the most frequently occurring parts of the skeleton in the fossil record retain similar levels of phylogenetic signal as parts of the skeleton that are rarer. These results demonstrate that the biased squamate fossil record contains reliable phylogenetic information and support our ability to place incomplete fossils in the tree of life.

Body size estimation from isolated fossil bones reveals deep time evolutionary trends in North American lizards

“…These include anatomical measurements from a total of 283 fossil lizard specimens, with 44 recognized genera and at least 33 recognized species, from the Paleogene record of the U.S. Western Interior (Table 1 and S1 Dataset). Only fossil lizard specimens with complete cranial bones or associated limb bones were sampled (e.g., Fig 3), for the following reasons: 1) the lizard fossil record is mostly comprised of cranial bones and, fortunately, most extant phylogenetic morphological characters for lizards are found in cranial bones [62,63], making it possible to determine at least coarse taxonomic identifications (family or higher) for fossil lizard specimens from cranial material [63]; 2) historical fossil identification may lack fidelity at the genus or species level [64][65][66]; 3) cranial and limb bones can be used to estimate SVL in lizards [9,34]. A few measured specimens included complete skulls (Fig 3A and 3I) or even skeletons (Figs 4A and 5).…”

Cutting the long branches: Consilience as a path to unearth the evolutionary history of Gnetales

“…We believe, however, that integrating fossil Gnetales in the phylogeny of the extant species is a necessary step for contributing to our understanding of the timing, character history, and macroevolutionary dynamics of the group. Recent methods and analyses have shown that even fragmentary fossil material, both reproductive and vegetative, can still carry a substantial amount of phylogenetic and evolutionary signal (Coiro et al, 2018;Mongiardino Koch et al, 2021;Woolley et al, 2022). This is particularly true when data are analyzed with techniques that keep in consideration the uncertainty implicit in phylogenetic analyses (Dávalos et al, 2014;Coiro et al, 2018Coiro et al, , 2020Erdei et al, 2019;Klopfstein and Spasojevic, 2019).…”

Examining paleobotanical databases: Revisiting trends in angiosperm folivory and unlocking the paleoecological promise of propensity score matching and specification curve analysis

“…Consider a study that compares hypotheses J and K, finding support only for hypothesis K. It may be the case that hypothesis K holds no more explanatory power than hypothesis J, and the empirical support for hypothesis K is simply a result of the structure of the fossil record (Holland, 2017) or of biases in the dataset (Raja et al, 2021). Various methodological contributions have shown that it is possible to account for the structure of the fossil record when testing paleontological hypotheses (Wagner and Marcot, 2013;Woolley et al, 2022), but paleontology has not yet reached a point where all studies account for such structure and bias when interpreting analytical results.…”