Maximum Voluntary Isometric Contraction (MVIC) is a standardised, objective and sensitive tool for the measurement of muscle strength. The purpose of this study was to investigate different aspects of reliability of MVIC and to determine if a learning effect existed in a relatively new user of the system. Two clinical investigators participated in the study. The inter- and intra-rater reliability of MVIC of 11 muscle groups was tested on healthy subjects (n=35). Intra-class correlation co-efficients (ICCs) were calculated and the statistical methods described by Bland and Altman were applied to the data. ICCs were higher for the more experienced investigator and a learning effect was demonstrated in a relatively new user of the system. Inter-rater reliability was acceptable but lower than intra-rater reliability. Upper limb tests generally yielded higher ICCs and lower ranges of error. The ICC was similar regardless of whether the maximum or average of the two values was taken in a single session. Utilising the statistical methods proposed by Bland and Altman allows estimation of the magnitude of error of MVIC and gives additional information to the ICC. These methods may be useful in the training of investigators and in clinical interpretation of MVIC values.