A series of rainfall simulation experiments was carried out at the Walnut Gulch Experimental Watershed, Tombstone, Arizona (31 43'N, 110 41'W), to observe the speed at which desert pavement surfaces could be re-established following disturbance. The results of these experiments, which consisted of repeated, 5 min rainfall events, demonstrate that pavements can reform within 10 events, which is compatible with observations of the recovery of surfaces under natural rainfall on an annual cycle. A model for the development of pavements by raindrop erosion processes had previously shown the importance of these processes. The rainfall simulation experiments were used to test the general applicability of this model. The model was able to reproduce the general characteristics of the regenerated surfaces and the timing of their development. However, details of the particle size fractions produced were less well simulated by the model. Testing of the sensitivity of the model to the sediment transport parameters suggests that this problem is not related to the soil characteristics, but is more likely to be an indication of a poor understanding of all the feedbacks operating in the raindrop erosion processes.