It is shown that nonlinear Volterra, polynomial autoregressive, and bilinear filters have the same layered implementation procedure. Using the layered structure, the order of nonlinearity can be increased by adding more layers to the structure. The structure is modular and consists of the simple moving average (MA) or autoregressive (AR) filters which can be added to the structure to achieve a desired degree of complexity. In addition, the modular layered structures admit very large scale integration (VLSI) implementation of the polynomial nonlinear filters.