The constitution and configuration of taxonomic groups resulting from applying three classificatory programs-two intensely clustering and the third of variable intensity-to two sets of data relating to a base group of 54 grass genera scored for 62 attributes are compared. All programs produced much the same groupings of genera at the 10-group level. Addition of identical taxa to the original sample resulted in increasing isolation of the replicated genus (Panicum) without affecting the composition of the remaining groups. Addition of near-identical taxa had a more generalized influence on the dendrogram. Dendrograms arrived at by intensely clustering programs may not represent taxonomically acceptable hierarchies. The flexible program causes less distortion between dendrograms based on different sample sizes, but may produce weakly structured, less informative classifications.