Cattle trampling of endangered plants, certain animal species, and gro
und nests may be a management concern on rangeland, Researchers need t
heoretical models of trampling lass to assist in design of studies and
interpretation of results, Managers can use such models to assist in
grazing management decisions, We present null (random background) mode
ls for predicting probability of trampling loss, explore the effects o
f failure of assumptions underlying these models, and develop alternat
ive models for dealing with nonrandom grazing and nonrandom placement
of vulnerable objects, The null models predict that if time-based stoc
king rate (head-days ha(-1)) is held constant and 1 pasture is grazed
under several rotation schedules (a study design used to simulate rota
tional grazing), or if 1 pasture is divided into n paddocks through wh
ich 1 herd rotates, the probability of trampling is operationally cons
tant. This qualitative prediction holds when grazing is nonindependent
and nonrandom, competing risks exist, and objects subject to tramplin
g are dispersed nonrandomly, Quantitative predictions of the null mode
ls do not hold under nonrandom grazing, which is expected to reduce pr
obability of trampling, Researchers can use predictions of the models
as a priori hypotheses, If empirical results deviate from the predicti
ons, then researchers should search for the underlying cause-effect me
chanisms, For management, the models indicate that trampling varies wi
th livestock density and time grazed but is independent of herd rotati
on.