This paper investigates the flocking problem of networked nonlinear Euler-Lagrange systems with parametric uncertainties, and they are assumed to interact on directed graphs with a directed spanning tree. We propose an adaptive controller to achieve the flocking objective, and the resultant closed-loop networked system bears the cascade structure. Using a new similarity decomposition approach, a critical-characteristic-root based approach, and the input-output stability analysis, we demonstrate the convergence of the position/velocity synchronization errors among the uncertain nonlinear Euler-Lagrange agents. We also show that the velocities of the Euler-Lagrange systems converge to the weighted average velocity value. Simulation results are provided to demonstrate the performance of the proposed controller. (C) 2013 Elsevier Ltd. All rights reserved.