A set of mathematical procedures for grouping individual items into classes. The technique used is cluster analysis, which produces groupings of items based on their degree of similarity. There are different ways of measuring the similarity between items, and different techniques of producing clusters from such measurements. Agglomerative techniques start with the most similar items and repeatedly add new members to existing clusters as the standard of similarity is lowered; divisive methods, on the other hand, start with the entire collection to be classified and repeatedly subdivide into smaller groups on the basis of certain attributes. The results of the analyses can be shown in the form of a dendrogram, but the interpretation of the groupings produced will depend on a detailed assessment of the archaeological data itself. Numerical taxonomy is also the multivariate analysis of many measurable features (taxonomic characters) to produce a biological classification. Because of the complexity of the analysis, the use of a computer is virtually mandatory. No attempt is made, as in evolutionary taxonomy, to weight characters on the basis of their presumed roles in natural selection. For this reason, numerical taxonomy produces a classification that reflects phenetic distances i.e., degrees of similarity. Such classifications are rejected by many conventional taxonomists who feel that the relationships expressed in a classification should be strictly evolutionary. The numerical evaluation of the affinity or similarity between taxonomic units and the ordering of these units into taxa on the basis of their affinities is used often in archaeology.