The phenetic analysis of non-nodulating Acacia species by Harrier et al. (1
997) was repeated to illustrate how different computer programs may generat
e alternative UPGMA trees for the very same data, even in the absence of da
ta input order effects (ties). For example, all Harrier et al.'s UPGMA dend
rograms produced by software from the Scottish Agricultural Statistics Serv
ice differed from those obtained by the packages NTSYS and MVSP87. Particul
arly, the positions of A. albida, A. rovumae, and A. pentagona, as well as
the relationships between Diacanthae and Triacanthae were affected by this
phenomenon. Hence, whenever: clustering techniques are used, care should be
taken to consider possible software-dependent caveats and artefacts. Never
theless, all programs provided clusterings that largely coincided with the
subgeneric and sectional groupings proposed by Vassal (1972) although the p
ositions of some species varied depending on whether morphological or molec
ular data were considered (e.g. A. albida and A. rovumae).