> Something with an exp(-x^2) in it just seems perversely hard to integrate! The popularity of the gaussian in this context almost surely derives from the heat kernel. --bw