Abstract
The possibilities of using the statistiral technique of factor analysis in describing variations in plant communities are explored. This method enables the variations to be treated as continuous, instead of resulting in a separation of the stands studied into a limited number of discrete associations or other synecological categories. It further provides a means for testing whether such separation can be objectively justified. It may often facilitate the recognition of the complexes of environmental factors which mainly determine differences in vegetation, and provides a means of estimating the relative value of the various species as indicators of these environmental complexes. In the present paper, the "principal axes" technique of factor analysis is applied to the analysis of data for percentage cover for 14 species in the Victorian Mallee. It is shown that their distribution, in so far as it does not depend on factors peculiar to individual species, can be represented in terms of at most five orthogonal "factors". The two most important "factors" are interpreted in terms of catenary changes in the vegetation. Other less common species not included in the analysis show high correlations with these "factors". In units of 1.28 ha there is no evidence that more than one continuously varying population is represented in the area; but in units of 25 sq.m. the majority of quadrat records fall into one or other of two principal categories, representing the valley and ridge communities. The potential value of factor analysis in plant sociology, and difficulties in its application to this field, are discussed.