From the Vlasov equation written as a nonlinear equation for the distribution function ƒ, a perturbation expansion of the fluctuating part of ƒ, ƒ̃̃=ƒ — 〈ƒ〉, is carried out in terms of the linear solution using a diagram representation similar to that of Wyld. From this a diagram expansion is obtained for the correlation function U=〈ƒ̃ƒ̃̃〉 and similar diagram series for two auxiliary functions, the averaged linear response function G=〈Ĝ〉 and the generalized vertex function Γ. Approximate asymptotic equations for U imply the summation of appropriate infinite subsets of diagrams. To derive such approximations two criteria are applied: 1) Consistency with the basic conservation laws (particle number, momentum, energy); 2) Selection of diagrams analogous to those leading to good approximations in Kraichnan’s random oscillator problem, the principal results of which are regained. Contrary to Wyld’s summation procedure diagram summation are not carried out directly in the expansion of U, but in the expansions of the functions 〈ƒ̃ƒ̃ƒ̃̃〉 and 〈ƒ̃̃ Ĝ〉, which appear on the right hand side of the differential equations for U and G. This automatically guaranties the validity of the conservation laws. The first approximation we discuss, which is equivalent to the quasi-Gaussian approximation well known from hydrodynamic turbulence, leads to the kinetic wave equation of weak turbulence. The next approximation consists of a coupled set of equations for U and G which are identical with the equations of the symmetrical random coupling model of Orszag and Kraichnan. The third approximation derived implies the generalized vertex function Γ. The first and the third of these approximations cannot be obtained in Wyld’s summation scheme. On the other hand it does not seem possible to derive the third approximation for the nonlinear Vlasov equation using Kraichnan’s diagram technic.