Different methods exist for an absolute quantification of in vivo proton MR spectra. According to the principle of reciprocity a simple method has been suggested previously, comprising the multiplication of the spectral signals with the unlocalised reference amplitude of the measurement as a correction for the different coil load. Afterwards, the result can be compared to measurements on phantoms in order to obtain in vivo metabolite concentrations. Thus, only one calibration would be sufficient in principle for all measured spectra. However, during a previous serial study on humans and phantoms we observed significant changes in transmitter amplitudes over time, although metabolite peak areas remained quite constant. Hence, the method led to incorrect metabolite concentrations. Since these effects are mostly underestimated or even neglected in clinical spectroscopy, the purpose of the present study was to investigate these instabilities over time in order to find a new correction method for quantitative MRS. According to this new method, the spectra corrected by the coil load have to be modified additionally by dividing by the reference amplitude of a phantom.