(1) Rate of increase was measured for feral donkey populations in parts of northern Australia as the average exponential rate of increase, r, in a population subjected to substantial reduction. (2) The annual rate of increase was estimated to be r = 0.21. This estimate was compared with the exponential rate of increase in another recovering donkey population and found to be in close agreement. (3) Ultimate and proximate factors regulating the abundance of feral donkey populations were examined by monitoring and sampling two populations; one at or close to equilibrium density, the other below equilibrium density and recovering from reduction. (4) The size of the population at equilibrium density remained stable over the 12 months of monitoring while the size of the recovering population increased by 20 percent (r = 0.18/yr). (5) Growth and body condition were significantly depressed in the population at equilibrium density suggesting that donkey populations are limited by the food resources available to them. (6) Breeding occured over a discrete season, with births occurring between September and February. 11 (7) Fecundity was high, with more than 75 percent of mature females breeding in each year, and was independent of population density. (8) Adult and juvenile mortality were density dependent, with mortality over the first six months of life the most important demographic factor influencing rate of increase in donkey populations, and hence population abundance. (9) Implications of the estimated rate of population increase for the cost of long-term control of feral donkey populations were examined by constructing numerical models predicting the relative cost of ongoing control. (10) These models were constructed using functions to describe density dependent variation in population productivity and the cost per donkey removed. (11) The cost of removing donkeys at various population densities was estimated using predator-prey theory. The cost, measured in hours of helicopter time per removal, was found theoretically and empirically to take the form of an inverted functional response curve, with cost saturated at high donkey densities. (12) The utility of models predicting the cost of continuing pest control is illustrated by comparing the relative costs of two potential strategies for feral donkey control.