Propagators - advanced and retarded - are introduced for nonlinear problems and it is shown that they generalize the Green's functions used to solve linear problems. These propagators are generated by using the dual operator which plays, for the nonlinear problem, the same role as the customary adjoint operator for linear problems. The propagators obey a reciprocity relationship and satisfy a closed nonlinear integral equation or, alternatively, a linear integral equation that depends parametrically on the problem's solution. These equations are formulated in a canonical way, independent of dimensionally, boundary conditions and type of the underlying nonlinear problem. 8 refs.