Abstract
Theoretical considerations and a comparison using a single set of vegetational data are used to compare techniques developed by workers at the University of Wisconsin with those described in Part I. The similarity coefficient (Sorensen''s) used by these workers is shown to be non-euclidean and dependent on species number. Their use of reference stands does not provide an adequate representation of the variation between stands of vegetation. It is concluded that the previous techniques are inefficient for extracting ecological information and that principal components analysis of a new coefficient (weighted similarity coefficient) is more efficient.