Abstract: Small mammals play a very important role in late Cenozoic biostratigraphy and biochronology. Especially since screen-washing technology began to be applied in paleontological investigations, it has often been possible to acquire large samples of small mammal fossils, which makes them a more effective proxy for biostratigraphy and biochronology than large mammals. Accordingly, classification and identification of these large samples have become a fundamental paleontological task. However, when dealing with them using empirical or intuitive traditional morphological methods, it is impossible to avoid paying too much attention to some slight and inconsistent differences and “oversplitting” taxonomic units, or conversely neglecting some consistent differences that are imperceptible to human eyes and lumping several taxonomic units together. In either case, subjective factors will introduce a degree of arbitrariness into the classification and identification of the specimens. To avoid this, a quantitative geometric morphometric method of classification and identification based on analysis of large samples is explored in this paper. The subjects of this case study are 1284 isolated molars of Mimomys gansunicus, Heteromimomys zhengi and Villanyia fanchangensis from the lower Lower Pleistocene Renzidong site, and 163 specimens ofMimomys bilikeensis from the Lower Pliocene Bilike site are used as a point of reference. 7-14 2D homologous landmarks were defined on the occlusal surface of each of the six molars and used as the basis for a linear discriminant analysis. The results confirm that there are three different arvicoline species in the Renzidong sample, and the linear discriminant functions produced in the analysis can describe the consistent differences that exist among the species in this large sample. Furthermore, the same functions can be used as a basis for identifying newly recovered fossil specimens of related arvicoline species. Because geometric morphometrics can deal with both discrete landmark data and continuous outline data pertaining to significant morphological characters, and is suitable for use in both 2D and 3D, this method can be generally applied to the classification and identification of small mammal fossils.

摘要： 小哺乳动物化石在晚新生代生物地层学和生物年代学研究中具有重要作用，尤其是筛洗法在古生物调查中得到广泛应用后，通常可以采集到数量可观的标本，使得其地位较大型哺乳动物化石更加彰显。因此，小哺乳动物化石标本的分类鉴定也成了一项十分关键的工作。然而，传统形态学方法在对大量标本进行分类鉴定时，往往容易受主观因素影响而将不稳定的细微性状变异作为依据建立新种，或者忽视一些肉眼难于察觉的形态学差异而将两个甚至多个类群合并到一起，导致基于形态学的化石分类鉴定随意性增加，失去客观性。此外，对于不具鉴定意义的非关键性单个牙齿，很难凭借肉眼或显微镜观察进行区分。针对这些问题，本文选取了安徽繁昌人字洞早更新世早期三种䶄类甘肃模鼠Mimomys gansunicus, 郑氏异模鼠Heteromimomys zhengi和繁昌维蓝尼鼠Villanyia fanchangensis的1284件臼齿并以早上新世内蒙古比例克的比例克模鼠Mimomys bilikeensis的163件臼齿作为参考，采用几何形态测量学方法在各个臼齿咬合面上分别选取了7~14个同源landmark对咬合面的形态特征进行了线性判别分析，建立了针对这4个种的臼齿咬合面形态差异判别函数。分析结果表明，根据这些landmark所提供的形态差异信息，人字洞的1284件臼齿标本中的确存在3个可以明显区分且形态学性状稳定的类群，先前的分类鉴定得到了验证。与之不同地点不同时代的比例克模鼠也可以很好区分。因此，建立在大样本基础上的这4种䶄类臼齿咬合面同源形态特征的判别函数可以用来描述这些种类较为稳定的形态特征差异，并用来作为今后对样本较少的标本进行分类鉴定时的判别依据。因为几何形态测量学的方法不仅可以对二维的离散landmark数据以及重要形态学特征的轮廓线进行分析，甚至还可以扩展至三维空间，所以上述方法对于较为容易获得大样本标本的小哺乳物化石分类鉴定具有普遍意义。