Herbivores harvest vegetation, interaction (5). The distribution of this harvest across vegetation types differs with herbivore species and for anyone species with time (e.g. Noy-Meir,1974a; Coppock et al. , 1986a,b). It can be generalized that herbivores, ruminants or not, seek the 'rich' resources, herbage that is highest in quality (digestibility, N, etc.).When faced with 'poor' resources, i.e. herbage of low quality, herbivores, particularly domestic stock, show considerable selectivity (finding green plant parts) or flexibility (switching plant types) in their diets (Breman and de Wit, 1984; Sinclair and Fryxell,1985; Coppock et al.,1986a; Moore, 1987) with the objective of maximizing the nutritional value of the intake (Coppock et al., 1986b).

   In persistent, arid pastoral systems the efficiency of harvest is low (Noy-Meir,1974a, 1985), as is the efficiency of conversion (Coughenour et al., 1985). In destabilized or desertifying systems this harvest rate can obviously reach 100% where the landscape is completely denuded of vegetation. However, the utilization rate does not have to reach that level to have a significant effect on the herbage layer. There is a vast literature describing the direct effects of repeated defoliation on the productivity and, to a much lesser extent, the longevity of perennial grasses and shrubs. The timing and cycle of defoliation in relation to the phenological stage of the plant is the critical factor (Hodgkinson and Mott, 1987). High levels of offtake involve risk. The greater the offtake, the longer is the recovery time because recovery time is a function of existing productive capacity. A long recovery time makes the perennial grassland system more vulnerable and less resilient, because of the increased probability of a drought intervening and generating additional stress.

   At lower levels of grazing there is a synergistic reaction of plant growth to grazing. The higher the plant biomass, the smaller the growth response for a given increment of rainfall. Thus a reduction of biomass by grazing can enhance the productivity of the vegetation as a whole (Noy-Meir,1975). The existence of this synergistic process has been proposed by McNaughton (1984), and the importance of it explored by Caughley et al. (1987).

   The interaction of vegetation on herbivores (6), serves as a negative feedback to grazing (5). Setting aside adaptations such as spinescence, toxins, etc., the predominant factors in this negative feedback to the herbivore are limitations to the supply of energy and nutrients (Ruess,1987 and references therein). In the African savannahs and grasslands the considerable range in species composition, each with its distinct phenophase, has resulted in a remarkable species packing of herbivores. Each herbivore occupies a tightly defined resource niche, each having different nutritional requirements.

   As the vegetation complex declines in 'richness' or quality, herbivores can respond by migration on some spatial scale. Where this is not possible, the response must be in diet selectivity and flexibility. Diets low in quality (energy and nutrients) reduce individual vigour and reproductive success, thereby providing a negative feedback on herbivore numbers (Sinclair et al., 1985).

   The strength of the reciprocal coupling of mammalian herbivores to vegetation varies within arid ecosystems, being stronger in the higher-rainfall savannahs than in the lower-rainfall grasslands. At the arid extremes there is almost no coupling (Shmida et al. , 1986). The strength of this coupling is determined not by the variability in the quality and quantity of the vegetation but by its predictability in time and space. Stochastic systems can only be utilized by sedentary herbivores with either short reproductive times (e.g. insects) or the capacity to store forage (e.g. termites--Braithwaite et al., 1988) or by opportunistic nomadism. Where predictability is high, i.e. pronounced and reliable rainy seasons, then this coupling of herbivores to vegetation by energy and nutrient flow can be tight and complex, and has been interpreted as stabilizing the population of both (McNaughton, 1985; Belsky,1986) .