The thesis is divided into two parts; part I deals with a study of the taxonomy of the genus Stylosanthes as a whole and part II is restricted to a study of Australian introductions of the genus.
In part I the conventional criteria used to define taxa within the genus have been critically investigated, and it has been found, that they are not always constant, even in the same individual. The "axis rudiment", the most important character used to define taxa, has been found to be variable in its expression in the inflorescence of individual plants of a number of species. Likewise, there are some species in which individuals may have one or two inner bracteoles. Hence the accepted subdivision of the genus into two Sections on the basis of these criteria may result in mis-classifications. Furthermore, in one instance, two species (S. humilis and S. sundaica) are placed in different Sections because one has an "axis rudiment" and the other does not, although the species are otherwise difficult to distinguish.
A study of the taxonomy based on numerical methods has been made using 27 characters extracted from the literature. Eight numerical procedures have been employed, which resulted in different groupings of the species. The procedure of the non-metric coefficient with flexible sorting, which had given the "best" grouping in a comparison of methods of numerical taxonomy ('t Mannetje 1967a), gave a grouping of species identical to the conventional treatment of the taxonomy of the genus, with the exception that two species, S. humilis and S. ingrata. of the section Stylosanthes were placed with species of the section Styposanthes. This appears to be a more sensible arrangement, as it puts S. humilis and S. sundaica next to each other in the same section. S. ingrata is the only species in the section Stylosanthes, according to Mohlenhrock (1957), which has two inner bracteoles, a character usually associated with the presence of an "axis rudiment".
By applying the same numerical techniques to the inverse data-matrix, relationships "between characters have also been determined. This showed, that there are four characters that can he used to distinguish between the two sections in the genus, instead of the two used in conventional taxonomy. However, if the numerical treatment of the taxonomy of the genus would be acceptable it would mean, that the section to which a new species would have to be added would depend on the overall similarity of this new species to either of the sections, as no single character would be sufficient criterion.
On the basis of the natural distribution of the species and of their chromosome numbers it would follow that species of the section Styposanthes would be more recent in origin than those of the section Stylosanthes, because all the known polyploids are in the former section and all the known diploids, except one, are in the latter section. An apparent contradiction would seem to be that species of this more recent section have an "axis rudiment".
Part II deals with a study of Rhizobium-affinities. flowering behaviour, morphology and cytology of a number of Australian Stylosanthes introductions. Relationships between these taxa based on Rhizobium-affinities were determined by numerical analysis using data obtained experimentally by subjecting 21 Stylosanthes accessions to infection by 28 Rhizobium strains. Morphologically similar accessions belonging to the same species had high Rhizobium-affinities, and morphologically dissimilar accessions classified as S. guyanensis had low Rhizobium-affinities.
The flowering behaviour of 15 Stylosanthes accessions was determined in two phytotron experiments and one glasshouse experiment. Results from the phytotron experiments also showed that growth habit and leaflet size and shape in some species are dependent on photoperiod and temperature. For this reason these features cannot be used as taxonomic characters.
Morphological and cytological data were obtained for 14 Stylosanthes accessions and these together with data on flowering behaviour were used in a numerical analysis of phonetic relationships between the introductions. The analysis revealed that the accessions of S. guyanensis were dissimilar and belonged to several groups, thus supporting the grouping obtained with Rhizobium-affinities.
Although the grouping obtained with Rhizobium-affinities and that based on morphology, flowering behaviour and cytology were not identical, a comparison of the results of the two approaches indicated that Rhizobium-affinities may be accepted as evidence in taxonomic studies.
From the agronomic point of view it would be desirable to be able to refer to definite taxa when dealing with the different forms of S. guyanensis, because they differ greatly in terms of physiology and morphology. Past taxonomic treatment of this species has allowed for varieties within the species, but the Australian material is too limited and the descriptions of these varieties are too vague to propose changes in the presently accepted taxonomy. Pending further work on this complex a tentative key has been presented to distinguish between four forms amongst the Australian introductions of S. guyanensis.
Some general implications of these studies relating to the taxonomy of the genus, to plant introduction programmes and to agronomic use have been discussed.