Estimates of the stresses in the crust can be obtained from considerations of isostasy, and insofar as the earth's topography is generally near a state of isostatic compensation, these values reflect the strength of the crust subjected to loads on a geological scale. While gravity is a poor indicator of the stress state in the crust, some limits can be established. Lower limits can be found using Jeffreys' criterion that the compensation is such as to minimize the maximum stress differences that occur in the loaded crust. Analysis of several large gravity anomalies using this approach suggests an average value of about 500 bars, with maximum values up to 1.5 kbar for the Tibetan plateau. This approach is only valid if the region is free from comtemporary and recent tectonics, which the latter is not. An upper limit to the maximum stress differences is given by modeling the features using elastic plate theory, Such models often indicate stress differences in excess of 5 kbar, but they can be significantly reduced; beneath the large seamount, for example, stress differences need not exceed 1 kbar, Thus the limits of maximum stress differences are about 500 to perhaps as high as 1--1.5 kbar, with the maximum stress differences being confined to the upper crust above 15--20 km in the continents and shallower in the oceans.