Commonly used to design excavations in rock, empirical rock-mass classification systems make design recommendations on the basis of experience in similar geologic conditions and are generally considered conservative. Numerical methods make it possible to develop design procedures, in which the influence of joint properties and other variables can be assessed directly. This study examined the mechanical response of a jointed rock mass around a supported circular tunnel, using the distinct element code UDEC. This response was compared to that predicted by continuum analysis, using properties derived from empirical systems. The Rock Mass Rating (RMR) system, Bieniawski 1976, and the Hoek-Brown failure criterion, Hoek, Brown 1980, were examined in this study. Five rock mass parameters were varied: orientation, spacing, condition, location of the joints, and the state of stress. Data from numerical models suggested that, for most of the rock mass models, RMR provided reasonably conservative results except in cases when the orientation of the joints caused the models to behave anisotropically. In such cases, continuum models, using material properties predicted by RMR, overestimated the rock mass strength predicted by discontinuum models.