The diffusion property of a soft percolation process in three dimensions is
studied. The jump rate of a random walker between two sites distributed ra
ndomly in space is assumed to fall off with the distance r as w similar to
[1- (r/r(0))](alpha) where r(0) is a cutoff length. The diffusion exponent
d(w) is determined by a new computational approach using the exact expressi
on for mean-square displacement, [r(2)(t)], of a random walker on a single
percolation cluster. The diffusion exponent dw is shown to depend on alpha,
deviating gradually from the value for alpha = 0.