A method is described for revealing patterns of social interaction. The procedure involves ordering groups by some principle, such as the total initiation rate of members within groups, and then examining the clusters of covariation. Two types of patterns are expected: those of like or common responses, and those of complementary responses. The analysis used to illustrate the method indicates both patterns. Limitations of the procedure are discussed and the method is judged to be crude but possibly the necessary first step before more detailed analyses will yield interpretable results.