General models of rangeland degradation suggest that soil and nutrients are lost and conversion of rainfall into primary productivity is diminished. These models are supported by studies on 'non-resilient' landscapes but have not been tested on 'robust' or 'resilient' landscapes. Results are presented as one large example (2000 ha) of a resilient landscape typical of the arid shrubland of Western Australia. The experiment was necessarily replicated in time not space, which limits confident extrapolation of the results to other areas.Over twelve years, when rainfall was mostly average or below-average, rainfall-use efficiency was 29% greater and net primary productivity 15% greater on a non-degraded landscape than on a similar degraded landscape where there were far fewer shrubs. Although our data conform with the general model of landscape degradation, that is degraded landscapes are less efficient in converting rainfall into biomass, in four of the 12 years, rainfall-use efficiency was similar to or greater on the degraded landscape than on the non-degraded landscape. There was increased production from ephemeral species on the degraded landscape, particularly in years of higher rainfall, and primary productivity became more variable from season to season. These responses are reported to characterize resilient landscapes.These data suggest that both resilient and non-resilient landscapes conform to a similar general model of landscape degradation. That is, averaged over time, primary productivity and rainfall-use efficiency are still reduced on degraded resilient landscapes.