In order to understand the morphological differences between four populations of genus Coilia (Teleostei: Clupeiforms) and identify them conveniently, truss network data were used to conduct multivariate analysis. Nineteen morphometric measurements were made for each individual. Burnaby's multivariate method was used to obtain size-adjusted shape data. The cluster analysis and discriminant analysis were used to discriminate among populations. The results indicated that 1) the four populations were clustered into three distinct groups; the first group included Changjiang C. mystus and Taihu C. ectenes, the second one included Zhujiang C. mystus, the last one included Changjiang C. ectenes, and 2) discriminant analysis with selected 4 morphological parameters showed that the identification accuracy was between 88% and 100%, and global identification accuracy was 95%. Our result showed that populations of different Coilia species living in geographic proximity to one another are more similar than conspecifics living farther apart. Separation and adaption are important to morphological difference. The taxonomy of genus Coilia should be reconsidered. This study also showed that the method to obtain size-adjusted data is important to acquire right conclusion.