A numerical method for solving stochastic mechanics problems by repres
enting the solution using a small number of random parameters is prese
nted. In essence, the method is a Galerkin approximation in the sample
space. The associated projection of the solution into the space of si
mple random variables reduces the stochastic problem to a set of deter
ministic problems. Alternatively, this method can be viewed as a modif
ied-for computational efficiency-stratified sampling method. Several e
xamples are considered involving the use of the Loeve-Karhunen expansi
on for a stochastic field approximation. The examples deal with the de
termination of the natural frequencies and of the seismic response of
a beam with random rigidity.