Under grazing pressure the vegetation of rangelands is often altered. Range managers traditionally have thought of these changes as retrogression or disturbance, setting climax vegetation back to a serai stage. Unfortunately, removing the grazing pressure often does not reverse the changes in the way the succession model would predict. Here four elements are sketched out of a theory which can deal with these exceptions. First, which plant growth-forms win competition often depends not only on the growth-forms but also on which is present as established adults, or on which is most abundant. Second, vegetation changes can sometimes induce soil changes. The extent to which reversing a vegetation change depends on soil restoration varies greatly, and the speed with which soil structure can be restored varies from rapid to negligible. Third, the impact of grazing on different life-forms varies greatly in different situations; in particular, the nature of and scope for selectivity is very different in dense grasslands, where much of the primary production is being removed, compared to open arid vegetation. Fourth, a new model is outlined for the adaptation of plant growth-forms to different arid climates. This model shows that different growth-forms can coexist stably in arid regions because they use different growth opportunities in complex weather sequences. Climates with the same general level of aridity can offer very different mixtures of growth opportunities, because of patterning of rainfall in time; accordingly different mixtures of growth-forms are found. These growth-forms are adapted demographically, not just physiologically. After particular weather sequences, some species may be present only as seeds or seedlings, while others are present as adults; thus above-ground vegetation can vary even while the mixture of populations present is stable. One growth-form, such as annuals, can include several distinct demographic strategies, with different responses both to drought and to grazing. These four elements of a theory, taken in different combinations, can explain both classical cases of range succession and the important exceptions.