A method of grouping journals within a wide discipline area into clust
ers is proposed, based on a algorithm that attempts to re-order a cita
tions matrix so that it is block diagonal, or block recursive. The alg
orithm is based on a penalty function which allows one to account for
the level of citation, not just the distribution of citations between
journals. A case study involving eight economics journals is presented
which illustrates the principles involved, but which also highlights
the computational problems associated with extending the analysis to l
arger numbers of journals.