In this study the intraspecific variation of wild Labeo rohita was investigated on the basis of morphometric characters using the truss network system constructed from the fish body. Altogether 435 fish samples were collected from six drainages of the Ganga basin in India. Data were subjected to principal component analysis, discriminant function analysis and univariate analysis of variance. The first principal component (PC1) explained 47.88% of the total variation, while PC2 and PC3 explained 17.22 and 8.33%, respectively. The step-wise discriminant function analysis retained three variables that significantly discriminated the populations. Using these variables, 62.3% of the original groups were classified into their correct samples and 53.1% of the cross-validated groups omitting one procedure were classified into their correct samples. Misclassification was higher for samples from the River Gomti (28.6%). Of the total of 31 transformed truss measurements, 30 exhibited significant differences among populations. These findings indicate the presence of six different stocks of L. rohita in the Ganga basin.