In quasieleastic neutron scattering spectrometers one usually faces a trade-off between energy resolution and counting statistics. If the resolution is improved the intensity at the detectors reduces and vice versa. It is not immediately clear how to weigh both factors against each other. In this paper it is proposed to use the maximum time obtainable by Fourier transform of the spectra as the quantity to be optimised. It is shown that this leads to a well-defined criterion for the choice of the resolution.