Theoretically, discriminating an amplitude modulated tone (AM) from a quasi-frequency modulated tone (QFM) is an ideal task for measuring the bandwidth of phase sensitivity because the stimuli have identical amplitude spectra but different phase spectra. The stimuli are perfectly discriminable at narrow bandwidths, but become indistinguishable at wide bandwidths. Measurements, however, are thought to be compromised by auditory distortion products, particularly a cubic distortion tone which interacts with the lower sideband of the stimulus to create an intensity cue. The results and implications of using a roving level procedure to eliminate distortion product effects are discussed.