The aim of this study was to assess the variability of lesion load measurements for repeated brain magnetic resonance imaging (MRI) scans judged to have either poor or good repositioning quality on the basis of subjective criteria used for clinical trials in multiple sclerosis (MS). Scan-rescan variability was also compared with the intra-observer variability assessed from three repeated volume measurements of the same scan. Nine patients with MS were studied; each of them underwent the scan-rescan procedure on the same day. Five scans were considered to have poor and four scans to have good repositioning. Between these two groups there were no differences in either the mean lesion loads or the measurement variance. For the whole group of patients, the intra-observer variability of repeated measurements gave a significantly lower coefficient of variation (COV = 4.9%) than the variability owing to repositioning (COV = 8.9%, P < 0.01, F test). Our results confirm that the effect of repositioning on brain MRI lesion load assessment is stronger than that owing to the intra-observer variability for repeated measurements and indicate that the simple visual judgment about the scan repositioning quality seems unable to predict the scan-rescan reproducibility.