The problem to estimate the background due to accidental coincidences in th
e search for coincidences in gravitational wave experiments is discussed. T
he use of delayed coincidences obtained by orderly shifting the event times
of one of the two detectors is shown to be the most correct.