Even though the large body of knowledge of relational query optimization techniques can be utilized as a starting point for object-oriented query optimization the full exploitation of the object-oriented paradigm requires new, customized optimization techniques-not merely the assimilation of relational methods. This paper describes such an optimization strategy used in the GOM (Generic Object Model) project which combines established relational methods with new techniques designed for object models. The optimization method unites two concepts: (1) access support relations and (2) rule-bused query optimization. Access support relations constitute an index structure that is tailored for accessing objects along reference chains leading from one object to another via single-valued or set-valued attributes. The idea is to redundantly maintain frequently traversed reference chains separate from the object representation. The rule-based query optimizer generates for a declaratively stated qu...