Since relaxation times are influenced by the hydration of the tissue and the chemical environment of the water molecules, T1 and T2 measurements (quantitative MRI) could be used as an indicator for the water content and the biochemical composition of lumbar intervertebral discs. The discriminating power of quantitative MRI for tissue characterization in individuals (for clinical diagnosis) and in cohorts (e.g. for investigations on disc physiology or composition) relies on the reproducibility in relation to the expected tissue differences. We therefore investigated the reproducibility in vitro (lumbar spine phantom) and in vivo (10 volunteers). To estimate the differences between normal and pathologic tissues in vivo, 100 normal and 20 herniated intervertebral discs were examined by quantitative MRI in a first application of our method. The relaxation times were calculated from a set of 20 images obtained with five single-slice/multi-echo sequences at different TR values on a commercial whole-body system (1.5 T). We have found a satisfactory reproducibility in vitro (T1: 1.9%; T2: 6.2%), while the reproducibility was less satisfactory in vivo (T1: 16.4%; T2: 13.4%). Calculated from theses values, differences in relaxation times of various tissues must exceed 486 ms for T1 and 24 ms for T2 (tolerance limits) to allow discrimination with a 95% confidence in individuals. We observed statistically significant (p = 0.001) mean differences between normal (n = 100) and herniated (n = 20) intervertebral discs (delta T1: 196 ms; delta T2: 15 ms). Although statistical significant in cohorts, a discrimination of normal and herniated intervertebral discs is limited by quantitative MRI in individuals, since the differences are smaller than the tolerance limits necessary for a reliable clinical diagnosis. However, our results indicate that variations in the disc hydration and/or composition can be noninvasively detected by quantitative MRI in studies of cohorts with sufficient accuracy.