It is shown that inappropriately constructed random phase surrogates can gi
ve false-positive rejections of the surrogate null hypothesis. Specifically
, the procedure erroneously indicated the presence of deterministic, nonlin
ear structure in a time series that was constructed by linearly filtering n
ormally distributed random numbers. It is shown that the erroneous identifi
cation was due to numerical errors in the estimation of the signal's Fourie
r transform. In the example examined here, the introduction of data windowi
ng into the algorithm eliminated the false-positive rejection of the null h
ypothesis. Additional guidelines for the use of surrogates are considered,
and the results of a comparison test of random phase surrogates, Gaussian s
caled surrogates and iterative surrogates are presented.