Statistical evaluation of character support reveals the instability of higher-level dinosaur phylogeny

The interrelationships of the three major dinosaur clades (Theropoda, Sauropodomorpha, and Ornithischia) have come under increased scrutiny following the recovery of conflicting phylogenies by a large new character matrix and its extensively modified revision. Here, we use tools derived from recent phylogenomic studies to investigate the strength and causes of this conflict. Using maximum likelihood as an overarching framework, we examine the global support for alternative hypotheses as well as the distribution of phylogenetic signal among individual characters in both the original and rescored dataset. We find the three possible ways of resolving the relationships among the main dinosaur lineages (Saurischia, Ornithischiformes, and Ornithoscelida) to be statistically indistinguishable and supported by nearly equal numbers of characters in both matrices. While the changes made to the revised matrix increased the mean phylogenetic signal of individual characters, this amplified rather than reduced their conflict, resulting in greater sensitivity to character removal or coding changes and little overall improvement in the ability to discriminate between alternative topologies. We conclude that early dinosaur relationships are unlikely to be resolved without fundamental changes to both the quality of available datasets and the techniques used to analyze them.三大恐龙主支系（兽脚亚目（Theropoda）、蜥脚形亚目（Sauropodomorpha）与鸟臀目（Ornithischia））之间的演化关系，在一项基于大型全新性状矩阵及其经大幅修改的修订版的研究得到相互矛盾的系统发育结果后，受到了愈发广泛的审视。本研究借助近年系统发育组学研究衍生的分析工具，探究该冲突的强度与成因。本研究以最大似然法作为总体分析框架，检验了原始数据集与重新评分后的数据集两类材料中，各替代假说所获得的全局支持度，以及单个性状所携带的系统发育信号的分布情况。本研究发现，针对恐龙主要谱系（蜥臀目（Saurischia）、鸟臀形类（Ornithischiformes）与鸟躯类（Ornithoscelida））之间的演化关系，目前存在三种可行的解决方案，三者在统计学上并无显著差异，且两类性状矩阵中支持每种方案的性状数量几乎均等。尽管对修订后矩阵所做的修改提升了单个性状的平均系统发育信号强度，但这反而加剧而非缓解了性状间的冲突，使得分析结果对性状移除或编码修改的敏感性更高，且在区分不同系统发育拓扑结构的整体能力上几乎没有提升。本研究最终认为，若不对现有数据集的质量与分析所用的技术手段进行根本性改进，恐龙早期演化关系的争议恐难以得到解决。