Desertification is the diminution or destruction of the biological potential of land, and can lead ultimately to desert-like conditions. The vegetation of southern Africa is claimed to have altered over the past 100 years and much of the change is attributed to pastoral practice. In recent years however there has been much debate around the issue of the deterioration and loss of productivity of the natural rangelands, specifically those under communal management. It is one thing to claim that the vegetation has changed but quite another to produce data and analyses to show this unequivocally. Furthermore it is generally difficult to determine the nature and extent of change in natural ecosystems, as one does not know what the optimal base-line conditions should be. For this reason emphasis has been placed on developing models of potential or expected vegetation. By comparing a model of potential or expected vegetation with that of the contemporary vegetation, areas that deviate from expectation can be identified, in so doing providing evidence of the direction of change in the rangelands under various management treatments. The objective of this study was to determine shifts in the vegetation under different land-use treatments, by developing a technique to predict the potential vegetation of an area. In order to explore the nature and extent of degradation at the landscape scale a study site was selected where a range of land-use and rangeland management practices could be studied in parallel. The mid-Fish River valley consists of three markedly different units of land management, namely commercial rangelands, communal rangelands and nature conservation areas. The vegetation within the mid-Fish River valley falls within the Thicket biome and consists of three main vegetation types namely, Short Succulent Thicket, Medium Succulent Thicket and Mesic Bushclump Savanna. The creation of this potential vegetation model was dependent on the direct gradient analysis approach of relating the community patterns with environmental variables. To achieve this, floristic information was collected at sites along a topographical-moisture gradient. A Canonical Correspondence Analysis (CCA) between the environmental variables and the plant communities produced a classification from which the conditions normally associated with the major plant communities were predicted. When projected as a digital map, the qualifying sites provided a testable hypothesis of the potential vegetation. The results of this study showed a definite grazing gradient, which reflects a change from a more mesic environment towards a more arid environment with an increase in utilisation pressure. The predictive vegetation model proved to be useful for predicting the occurrence of the valley thicket communities within the Eastern Cape.