Signal loss and absolute quantitation errors in 1H-MRS (localized proton MR spectroscopy) because of physiologic brain motion are analyzed quantitatively. Cardiac and respiratory related motion lead to substantial phase dispersion when using a standard, short echo-time STEAM sequence. The loss in signal area varies from 6-7% with TM (middle interval time in a STEAM sequence) = 13.7 ms, to 25-39% with TM = 100 ms. The variation in signal area because of motion-related phase dispersion is up to 16% for TM = 100 ms. The signal phase as a function of the position in the cardiac cycle is shown to be reproducible. Maximal differences in the signal phase are over 180 degrees for long TMs. ECG-gating reduces the phase dispersion considerably but introduces problems with variable repetition times. Using a phase calibration curve recorded with the water suppression turned off, it is possible to retrogate subsequent untriggered acquisitions with the water suppression activated, if the time points in the cardiac cycle are recorded for each acquisition. The gain in signal intensity is between 3 and 21%. For absolute quantification via brain water, this phase analysis has the important consequence that reference scans must be phased individually before co-adding, otherwise metabolite concentrations may be severely overestimated.