this workshop was to develop recommendations to the U.S. Geological Survey (USGS) on the use of precariously balanced rocks and related fragile geological features to improve the national seismic hazard maps (NSHM). The context of the workshop is the growing realization that it is essential to regard hazard maps as the calculated output of hazard models that should be tested, and that fragile geological features provide the only data to validate the predictions of these models at low probabilities.A fragile geological feature (FGF) was defined as a feature that might be easily destroyed by strong earthquake ground motions and is mechanically simple enough to allow for analysis of the ground motions that might cause it to be destroyed. For testing the prediction of a seismic hazard model, an FGF must also have been in essentially the present geometry for enough time to constrain past ground motions. A precariously balanced rock (PBR) is the type example of an FGF. A PBR is a rock that is balanced on, but not mechanically attached to, its pedestal, and it has a high ratio of height to width of the contact with the pedestal, so that it is easily toppled. Brune and Whitney (1992) and Brune (1996) suggested that PBRs might serve as low-resolution seismoscopes to put upper bounds on ground motions over the lifetime of the rock. FGFs were recognized by Hanks et al. (2006) for their critical contribution to evaluation of extreme ground motions at Yucca Mountain. Because PBRs have been studied more than other FGFs, most attention at the workshop focused on their distribution and analysis. However, generalization to other FGFs is generally straightforward.Particularly in arid environments where most PBRs have been found, very old PBRs test the predictions of probabilistic seismic hazard analyses (PSHA) over time intervals of greater than 10 4 years, related to probabilities ~10 -4 per year. Figure 1 shows an example of a PBR and illustrates the measurement of parameter α, where gtanα is the static acceleration needed to initiate tipping of the rock. Together with the rock dimension, α is a fundamental indicator of stability. Early studies (e.g., Brune 1996, 1999; Anderson and Brune 1999b) compared rocks such as these with hazard curves developed from the national hazard model (i.e., the input model used for the NSHM) and from alternative hazard models, and found numerous inconsistencies (e.g., Figure 2). Characterization, testing, and interpretation of PBRs have been developed considerably beyond the level of analysis shown in Figures 1 and 2. Significant advances in quantifying rock stability were described by Anooshehpoor et al. (2004) through testing in the field. Purvance, Anooshehpoor et al. (2008) and Purvance, Brune et al. (2008) demonstrate that under dynamic excitations the probability of overturning depends on two parameters: PGA and one of PGV, SA(1s), or SA(2s), as illustrated in Figure 3. Rood et al. (2010) and Balco et al. (forthcoming) advance the understanding of dating and the life cycle of a PBR.The prese...