Mutual information clustering (MIC), which is a conceptually simple method for hierarchical clustering of data, is suggested. It uses mutual information (MI) as a similarity measure and exploits its grouping property. \par The authors show that the MI can be used not only as a proximity measure in clustering, but that it also suggests a conceptually very simple and natural hierarchical clustering algorithm if it is easy to estimate the MI between objects. In applications, MIC might not give optimal clustering, but it is a simple method and it seems promising.