Although clustering analysis is a promising methodology for studying the perceived structure of text there are practical problems in adapting existing techniques to the special requirements of text research. The purpose of the present study is to develop a clustering methodology for text that will make large‐scale clustering analyses feasible, To this end, several sources of text proximity data are compared and the properties of a natural measure of structural relatedness are examined. As a result, a highly efficient single‐linkage clustering algorithm is formulated. The algorithm is then applied to segmentation data from a long passage on biological classification. Subsolutions for several sections of the passage are presented.