In a multidatabase system, schemas of the component databases are usually integrated into a global schema to provide users with transparencies on data distribution and schema differences. To process queries against the global schema, query modification, which maps the query into one against local schemas, is necessary. In this paper, an approach which transforms and decomposes the modified query into subqueries to optimize distributed query processing is presented. Each subquery references data contained at a local site and, therefore, can be locally processed. The partial results produced from these subquery executions are then collected to compute the query answer. This approach reduces query response time due to parallel processing. Moreover, queries which can actually be processed at a single site can be identified.