In this paper, we present a new approach towards a parallel resolution procedure which explores another dimension of parallelism in addition to the AND/OR formulation and special hardware constructs. The approach organizes the input clauses of a problem domain into a connection graph. The connection graph is then partitioned and each partition is worked on by a different processor of a multiprocessor system. These processors execute the resolution procedure independently on its partition, and exchange intermediate results via clause migrations. Preliminary test results and qualitative assessments of this procedure are also given.