Symbolic inductive learning systems explore a space of hypotheses to derive the definition of concepts. Due to the size and complexity of this space, sequential algorithms use various heuristics that limit the classification accuracy of the generated definitions. Parallel search algorithm can rectify such limitations. In this paper we present the PARIS parallel symbolic inductive system. PARIS uses the branch and bound search algorithm. It has been implemented on the Thinking Machines CM- and the MasPar MP-1 computers, and has been tested on the domain of financial analysis.