The 11 measurable morphological indicators and 18 framework morphological indicators of Leuciscus merzbacheri were studied by methods of principal component analysis, cluster analysis and discriminant analysis, in order to reveal their morphological information and laws which can guide the identification, resource conservation and fine variety breeding of . The results of principal component analysis showed that the contribution rates of three extracted principle components were 57.71%, 9.019%, 6.502% respectively and the cumulative contribution rate of them was 73.23%. The main proportional parameters of principle component 1 gathered in head and caudal peduncle. The parameter of head included six proportional parameters such as HL/BL, ML/BL, SL/BL, /BL6. And the parameter of caudal peduncle included five proportional parameters such as CL/BL、/BL and /BL. Cluster analysis revealed that the genetic relationship of is was closed, and both of them had the father genetic relationship with . A stepwise discriminant method selected eleven parameters which also gathered in head and caudal peduncle, in order to establish the characteristic discriminant function. The data of there kinds of fish were submitted into the characteristic discriminant function to validate. It showed that it is a little bit difficult to discriminate between is easy to distinguish between them. The individual discrimination accuracy (L. baicalensis are more than 80%, and the individual discrimination accuracy ( reached more than 95%. And the comprehensive discrimination rate of no cross-validated method was 91.2% and the one of cross-validated method was 89.4%. The comprehensive discrimination rate of no cross-validated method is high and very closed to the cross-validated method. It meant that the characteristic discriminant function is credible. Three species of in Xinjiang are divided into two branches. The morphological variation of them is mainly reflected on the part of head and caudal peduncle. And three species of in Xinjiang could be discriminated by multiple regression function.