Fourier transform NMR spectrometers have two very common instrumental imperfections that give rise to frequency-dependent phase shifts in the spectra-finite pulse intensity and the delay between excitation and acquisition. In most practical situations this phase error is largely eliminated by applying a phase correction that is linear with respect to offset from resonance. This has become a routine adjustment for the operator; alternatively it can be accomplished automatically by computer when there are well-defined signals in suitable regions of the spectrum. Many years ago when Fourier transform methods were first applied to nitrogen15 spectroscopy, using a sample of (enriched) ammonium nitrate with proton decoupling, it seemed quite natural to adjust the frequency-dependent phase correction to give two pure absorption-mode resonances. Unfortunately this obscured the really interesting piece of information that the NH4 resonance was in fact inverted due to a negative nuclear Overhauser enhancement (reflecting the negative gyromagnetic ratio of nitrogen15). Long experience with well-behaved spectra would normally alert the NMR operator to this type of problem, but we can hardly expect the computer to recognize these occasional anomalies. We propose here an alternative approach aimed at eliminating the necessity for frequency-dependent phase correction routines after data acquisition. It should prove useful in spectrometers which record many different spectra unattended (overnight) in cases where automatic phase correction routines are hampered by poor signal-tonoise. When high-resolution spectra contain some broad lines, these are distorted when frequency-dependent phase corrections are used since each individual resonance should be assigned a single phase angle, not a distribution of phase angles. Applications of phase-corrected pulses can also be visualized in multidimensional spectroscopy or in magnetic resonance imaging. It is well-known (I, 2) that a 90” pulse applied off resonance creates a phase error that is virtually a linear function of resonance offset over a wide range of offsets. We propose to replace this with a “reversed rotation pulse” [ 90” ( -X) ] -‘, a concept first introduced by Levitt (3, 4). This is a composite radiofrequency pulse that has the overall effect equivalent to a rotation in a sense opposite to that defined by the gyromagnetic ratio of the nucleus in question. Thus if a 909(+X) pulse, applied off resonance, rotates magnetization vectors clockwise about a tilted effective field in the X 2 plane, then the inverse pulse [ 90”( -X) I-’ rotates such vectors counterclockwise