A method is presented of simulating a class of nonstationary Gaussian random processes by passing a Gaussian white noise through a system introducing desirable nonstationarity at some phase of simulation. The method might have wide applications in stochastic mechanics such as in analysis of aircraft structures subjected to severe, random aerodynamic loading and in earthquake engineering. The relationship between the proposed nonstationary Gaussian process and the filtered Poisson process is examined in detail. Examples show that, if the convolution integrals, involved in the output of the system and also of the structure, can be avoided by particular choice of the impulse response functions, then a set of numerical simulations of the nonstationary process and the structural response can be performed rapidly on a computer.