Hypotheses about the relative productivities and impacts of current and officially recommended stocking rates were tested in a case study in southern Africa. They support the view that biased measurements of livestock productivity and inappropriate measurements of rangeland degradation can lead to under-estimates of the productivity of African communal rangelands, overestimates of the urgency of de-stocking, and misunderstanding of the strengths of measures required to effect it. By including the full range of livestock products in estimates of productivity and calculating output per unit of land area rather than per unit of livestock, more accurate estimates of productivity can be made. Estimates of degradation based on irreversible changes enable discrimination between trivial and important impacts of grazing. Modelling suggests de-stocking is not in general worthwhile at present in eastern Botswana. It may be necessary to regulate numbers of animals in the future. If livestock densities increase, grass cover could fall below the critical level and rates of degradation increase suddenly and rapidly. Adverse climatic change could reduce herbaceous productivity with the same effect. The approach used here should help decision making in those circumstaces. Choosing an appropriate stocking rate should be an ethical decision requiring negotiations among stakeholders. Practitioners should recognise that it is socially, not technically determined and that it is but one level among many possible densities.