Five techniques were used to predict the potential biogeography of the four alien plant species, Acacia longifolia, Acacia mearnsii, Opuntia ficus-indica and Solanum sisymbrifolium. Prediction was based on five environmental factors, median annual rainfall, co-efficient of variation for rainfall, mean monthly maximum temperature for January, mean monthly minimum temperature for July and elevation. A geographical information system was used to manage the data and produce the predictive maps. The models were constructed with presence and absence data and then validated by means of an independent data set and chisquared tests. Of the five models used, three (the range, principal components analysis and discriminant function analysis) were linear while the other two (artificial neural networks and fuzzy logic) were non-linear. The two non-linear techniques were chosen as a plant's response to its environment is commonly assumed to be non-linear. However, these two techniques did not offer significant advantages over the linear methods. The principal components analysis was particularly useful in ascertaining the variables that were important in determining the distribution of each species. Artifacts on the predictive maps were also proved useful for this purpose. The techniques that produced the most statistically accurate validation results were the artificial neural networks (77% correct median prediction rate) and the discriminant function analysis (71% correct median prediction rate) while the techniques that performed the worst were the range and the fuzzy classification. The artificial neural network, discriminant function analysis and principal component analysis techniques all show great potential as predictive distribution models.