This paper proposes the interoperation environment which enables an expert system to get information available to improve its performance from others. First, we have given a method library of reusable templates in order to provide a correspondence between specification and implementation of inference structures. Next, a cooperation method has been presented, using the diérence arising in the context of the correspondence between inference primitives of an originator and those of recipients. The wrapper with conversion facilities has been also provided, using a common domain ontology developed manually. After designing and implementing such an interoperation environment, experiments have been done among four heterogeneous expert systems. Furthermore, it has been shown that an expert system finds a way to perform a given task better by the interoperation with other three expert systems.