This paper argues that most queries in object-oriented databases require traversing from one object to another in the aggregation hierarchy. Thus, the connections between objects through object identifiers are essential to the efficiency of query processing and should be represented separately from the database. We introduce the concept of path dictionary and describe how it supports queries of different types. We evaluate the storage overhead, query and update costs of the path dictionary. Compared to the path index, the path dictionary has better overall query and update performance and lower storage overhead.