The taxonomic affinities of the Neottioideae (Orchidaceae) and, in particular of its Australian members, have been examined employing numerical taxonomic methods involving some 100 taxa and 80 attributes.
It was considered unwise to accept the literature uncritically and, consequently, the data on which the results are based represent largely original observations on living and preserved material. As far as possible, preconceived ideas about the taxonomic importance or otherwise of attributes have been abandoned, and many of the attributes used (especially those concerned with anatomy) have not been previously considered in taxonomic work on the Orchidaceae.
It became apparent during this study that much of the objectivity gained by the use of numerical methods is destroyed by subjective procedures in the selection of attributes and the coding of data. Several important, but not readily obvious, problems which add to the subjectivity at these levels of the analysis were uncovered and examined. In each instance the difficulty lay in the manner in which the attributes were to be coded.
Several different hierarchical clustering strategies were used and, from a careful consideration both of the properties of these strategies and of the results obtained, it was decided that the Euclidean-based Group Average Strategy was best suited to the requirements of this study.
Two analyses were carried out, the first involving a sample from the entire family and the second involving the subfamily Neottioideae as defined by the first analysis. The results of the family analysis indicated that the original sample was heterogeneous in that there were two groups in the sample worthy of familial status - the Apostasiaceae (Apoetasia and Paphiopedilum) and the Orchidaceae (the remaining genera). The Orchidaceae in this sense comprised three subfamilies - Orchidoideae, Neottioideae and Epidendroideae.
The analysis of the Neottioideae indicated that this subfamily comprised two tribes - Diurideae, including the tuber-forming Australian taxa and the Neottieae which do not possess tubers and are cosmopolitan in distribution.
A few taxa in both analyses appeared to be intermediate between the major groups present. The best examples of intermediate taxa occurred in the family analysis and involved Vanilla, Galeola and Nervilia which are apparently intermediate between the Epidendroideae and the Neottioideae. Sue taxa have not been adequately accommodated by conventional taxonomic systems, and the numerical methods employed in this study have also failed to resolve the problem unambiguously.
In summary, the thesis considers some aspects of numerical taxonomic methodology, and adduces evidence concerning the taxonomic affinities of the previously neglected Neottioideae.