Our work is motivated by the desire to develop methods to comprehend large vocabularies and large schemas of object-oriented databases. The ability of a user of a database participating in a federated system to retrieve information from the other database systems will be greatly enhanced by acquiring a better comprehension of these systems. The authors are trying to develop both a theoretical paradigm and a methodology to analyze existing large schemas. Their approach to achieve comprehension is based on combining two concepts: informational thinning (i.e. concentration on the specialization hierarchy of the schema) and partitioning. They present a new technique for modeling which is called disciplined modeling. Based on the rules of disciplined modeling we develop a theoretical paradigm to support the existence of a meaningful forest hierarchy within the specialization hierarchy. Such a hierarchy functions as a skeleton of the schema and supports comprehension and partitioning efforts.