Correlation in the fitness landscape corresponding to an optimization
problem can tell much about the landscape's topology. It has previousl
y been suggested that a long random walk is sufficient to gather this
statistical information. In this paper we will show that fitness lands
capes for constrained optimization problems are statistically anisotro
pic. This means a single random walk over this type of landscape is in
sufficient for determining correlation.