A new solution to the old problem of partitioning a matrix of social p
roximities into groups is proposed. It draws on a heuristic developed
in computer science, the simple genetic algorithm. The algorithm is de
scribed and its utility is demonstrated with applications to three sta
ndard data sets.