In this study, we explored the spatial and temporal relations between boulders and their original in-situ locations on sandstone bedrock cliffs. This was accomplished by combining field observations with dating methods using cosmogenic isotopes (10Be and 14C) and optically stimulated luminescence (OSL). Our conclusions bear both on the landscape evolution and cliff retreat process in the hyperarid region of Timna and on the methodology of estimating exposure ages using cosmogenic isotopes. We recognize three discrete rock fall events, at 31 ka, 15 ka, and 4 ka. In this hyperarid region, the most plausible triggering mechanism for rock fall events is strong ground acceleration caused by earthquakes generated by the nearby Dead Sea fault (DSF). Our record, however, under represents the regional earthquake record implying that ongoing development of detachment cracks prior to the triggering event might be slower than the earthquake cycle. Cliff retreat rates calculated using the timing of rock fall events and estimated thickness of rock removed in each event range between 0.14 m ky-1 and 2 m ky-1. When only full cycles are considered, we derive a more realistic range of 0.4 m ky-1 to 0.7 m ky-1. These rates are an order of magnitude faster than the calculated rate of surface lowering in the area. We conclude that sandstone cliffs at Timna retreat through episodic rock fall events that preserve the sharp, imposing, landscape characteristic to this region and that ongoing weathering of the cliff faces is minor. A 10%-20% difference in the 10Be concentrations in samples from matching boulder and cliff faces that have identical exposure histories and are located only a few meters apart indicates that cosmogenic nuclide production rates are sensitive to shielding and vary spatially over short distances. However, uncertainties associated with age calculations yielded boulder and matching cliff face ages that are similar within 1 ??. The use of external constraints in the form of field relations and OSL dating helped to establish each pair's age. The agreement between calculated 14C and 10Be ages indicates that the accumulation of 10Be at depth by the capture of slow deep-penetrating muons was properly accounted for in the study. ?? 2005 Elsevier B.V. All rights reserved.