Overcoring and hydrofracturing measurements show that the horizontal normal stresses at or near the earth's surface are nearly always compressive, with magnitudes which occasionally reach tens of megapascals. This is surprising because simple theoretical arguments suggest that the crustal extension and cooling which occur during unroofing (exposure of rocks by uplift and erosion) should frequently produce large tensile stresses near the surface. One likely cause of this discrepancy is the growth of microcracks, driven by the grain scale internal stresses produced by cooling and depressurization during unroofing. The unroofing process has been simulated for a simple two-dimensional model of a rock in which internal stresses arise only from cooling. These calculations show that crack growth leads to compressive changes in macroscopic stresses by increasing the elastic compliance of the rock and by decreasing its linear thermal expansion coefficient. The growth and opening of cracks also produces an expansion which must be canceled by increased horizontal compression in order to satisfy the lateral constraints imposed by the presence of adjacent rocks. The resulting stress changes (relative to the stresses expected in a crack-free rock) could easily reach tens of megapascals at the surface in some rocks, especially granites and marbles. This effect tends to make the near-surface stresses relatively insensitive to the mean geothermal gradient, in marked contrast to the behavior of uncracked rocks. Crack growth driven by internal stresses is also likely to be triggered by drilling, tunneling, or quarrying in rocks subjected to substantial in situ compressive stresses and could lead to macroscopic deformation quite different from that expected on the basis of a simple elastic analysis. The importance of such effects in a given rock depends upon whether the internal stresses are sufficient to produce significant crack growth, which in turn depends upon the relative values of two key parameters that are poorly known at present. These quantities are the temperature at which ductile deformation effectively ceases during unroofing and the minimum stress-intensity factor capable of producing significant crack growth during unroofing. Better estimates of these parameters would permit more quantitative conclusions regarding the effects of crack growth in particular rock types or geologic settings and would facilitate the development of constitutive laws describing the macroscopic deformation resulting from crack growth.