Closely jointed greywacke rock masses are widespread throughout both the North and South Islands of New Zealand and much of New Zealand's infrastructure is constructed upon greywacke rock masses. This thesis deals with determining the rock mass strength of unweathered closely jointed New Zealand greywacke rock masses. Currently, the estimation of rock mass strength and deformability is reasonably well predicted through the use of such empirical failure criteria as the Hoek-Brown failure criterion and empirical expressions to predict deformability. However, previous studies upon predicting the strength and deformability of unweathered closely jointed New Zealand greywacke rock masses has shown that existing empirical methods of determining strength and deformability are unsatisfactory. The problem with predicting rock mass strength and deformability moduli of New Zealand greywacke and the lack of adequate data to calibrate a failure criterion was the starting point for this work. The objective of this thesis was to increase the knowledge of intact and defect properties of closely jointed greywacke, develop reliable rock mass data with which to calibrate a failure criterion and improve the ability to estimate the rock mass strength of greywacke rock masses. A review of existing failure criteria for rock masses was conducted and of these criteria, the Hoek-Brown rock mass failure criteria was selected to calibrate to both the intact rock and rock mass failure data, because of its broad acceptance in the rock mechanics community. A database of greywacke properties was developed based on previous studies upon unweathered greywacke around New Zealand and is attached to the thesis as an Appendix. The database included descriptions of greywacke defect properties and mechanical properties of the intact rock and joints. From this database, inputs could be justified for numerical modelling and later analyses of failure criteria. Records from the construction archives of the Benmore and Aviemore hydroelectric power projects in the South Island of New Zealand were reviewed to obtain information and results from a series of shear tests carried out on unweathered closely jointed greywacke in the 1960s. Data on rock mass strength at failure and rock mass deformability were extracted from these records to assess the predictability of the failure criterion and deformability expressions. Problems experienced during the shear tests at the Aviemore dam site created doubt as to the actual rock mass strengths achieved at failure. The behaviour of these tests was studied using the finite difference code FLAC. The work was aimed at investigating the potential for transfer of shear force between the two concrete blocks sheared in each test and the impact shear force transfer had upon the likely normal stresses beneath each block at failure. The numerical modelling results indicated that a combination of preferential failure occurring in one direction, and doubt in the actual normal load applied to the concrete blocks during testing lead to premature failure in the blocks sheared upstream. The blocks sheared in the opposite direction failed at normal stresses that are reflective of the strength of an unweathered greywacke rock mass, but these results could be explained by failure occurring along defects therefore not satisfying the assumptions of homogeneity typically required of a rock mass failure criterion. The Hoek-Brown failure criterion for intact rock was investigated by fitting it to the largest intact greywacke datasets. For a full set of test data (i.e. including tensile data), the Mostyn & Douglas (2000) variant of the Hoek-Brown failure criterion gave the best fit for a full set of rock mass data. A multiple regression method was developed which improved the fitted curve to intact data in the tensile region and gave the best estimate of tensile strength if no existing lab results for tensile strength were available. These results suggest that the Hoek-Brown failure criterion is significantly limited in its applicability to intact NZ greywacke rock. Hoek-Brown input parameters different to those suggested by Hoek et al (2002) are recommended for using the Hoek-Brown failure criterion for intact NZ greywacke. For closely jointed NZ greywacke rock masses, the results from the shear tests at Aviemore and Benmore were separated into different GSI classes and Hoek-Brown envelopes fitted to the datasets by multiple regression. Revised expressions were proposed for each Hoek-Brown input parameter (mb, s, ab) as a function of the GSI. The resulting revised Hoek-Brown failure envelopes for NZ greywacke offer a significant improvement on the existing criterion used to predict the strength of NZ greywacke intact rock and rock masses. The differences in the behaviour of the reaction blocks that failed before the test blocks and the reduction in rock strength due to sliding along defects from that predicted could be reasoned from recorded observations and the behaviour of the concrete blocks during the shear tests. This study has clearly illustrated the need for continued research in this area. This includes (1) a means of assessing the role of defects upon the shear strength of closely jointed greywacke rock mass into a failure criterion, (2) further modelling of the in-situ shear tests by a discrete element procedure to expressly determine the role of the defect on failure, (3) more testing on rock masses to obtain more data to calibrate a rock mass failure criterion, and (4) more studies on predicting the strength of extremely disturbed rock masses.