Discontinuities in a rock mass can intersect an excavation surface to form discrete blocks (keyblocks) which can be unstable. This engineering problem is divided into two parts: block identification, and evaluation of block stability. One stable keyblock and thirteen fallen keyblocks were observed in field investigations at the Nevada Test Site. Nine blocks were measured in detail sufficient to allow back-analysis of their stability. Measurements included block geometry, and discontinuity roughness and compressive strength. Back-analysis correctly predicted stability or failure in all but two cases. These two exceptions involved situations that violated the stress assumptions of the stability calculations. Keyblock faces correlated well with known joint set orientations. The effect of tunnel orientation on keyblock frequency was apparent. Back-analysis of physical models successfully predicted block pullout force for two-dimensional models of unit thickness. Two-dimensional (2D) and three-dimensional (3D) analytic models for the stability of simple pyramidal keyblocks were examined. Calculated stability is greater for 3D analyses than for 2D analyses. Calculated keyblock stability increases with larger in situ stress magnitudes, larger lateral stress ratios, and larger shear strengths. Discontinuity stiffness controls block displacement more strongly than it does stability itself. Large keyblocks are less stable than small ones, and stability …