A structural domain represents a volume of rock masses with similar mechanical and hydrological properties. This paper presents a new multivariate method, built upon the classical method proposed by Miller (1983), for identifying structural domains by considering a combined effect of multiple joint characteristics. Firstly, joints are placed into corresponding cells according to a comprehensive categorisation of joint characteristics. Secondly, based on the frequencies of joints in each cell, a contingency table analysis and appropriate chi-square test are used to identify structural domains. In this method, the joint frequency in each cell is not only determined by joint orientation but also by joint trace length and aperture. A group of simulated joint populations is used to check the validity of the proposed method. Seven joint populations with a total of 731 joints collected in Southeast Dalian Port, China are evaluated using the proposed method. Calculations show that the differences in both distribution forms and sequences of joint characteristics produce apparently different identification results. A combined effect of multiple joint characteristics is found to play an important role in the identification of structural domains. Reliable results for the multivariate method are obtained, given the objective of analysis and practical application.