In order to evaluate the design quality of the UML, understanding meta-classes is a key activity as they are the primary weapons by which the UML specifies the application domains. The paper introduces taxonomic patterns for clustering the UML metaclasses based on the observation of their evolution and fitness. The result sets of the patterns assist in finding the evidence of the concerns about the UML design and quality. It not only helps to find out problematic meta-classes, possible design defects of the UML and the inconsistency between the UML meta-models and the application domains; but also provides valuable information for guiding the development and evaluation of the UML. The work can be the basis of further quality analysis of the UML meta-models.