A survey of published descriptions of 32 of the largest, least eroded terrestrial impact structures reveals that the amount of melt at craters in crystalline rocks is approximately 2 orders of magnitude greater than at craters in sedimentary rocks. In this paper we present a model for the impact process and examine whether this difference in melt abundance is due to differences in the amount of melt generated in various target materials or due to differences in the fate of the melt during late stages of the impact. The model consists of a theoretical part for the early stages of impact, based on a Birch-Murnaghan equation of state, a penetration scheme after Shoemaker (1963), and an attenuation model modified from Gault and Heitowit (1963), and a descriptive part for the later stages of impact, based on field observations at the large terrestrial craters. The impacts of iron, stone, permafrost, and ice meteorites 1 km in diameter into crystalline, carbonate, dry sandstone, ice-saturated sand, and ice targets are modeled for velocities of 6.25, 17, and 24.6 km/s. Tables of calculated crater volume, depth of penetration of the meteorite, equivalent scaled depth of burst, radii to various peak pressure isobars, volume of silicate melt, and volume of water vapor (or, in the case of carbonate, carbon dioxide vapor) are presented. Simple algebraic expressions for pressure attenuation are derived: for the near field, dX/dR = 3Xn/R(1 - n), where X is the pressure normalized to an averaged bulk modulus for the target rocks, R is the radius normalized to the radius of the cavity in which energy is initially deposited, and n is the pressure derivative of the bulk modulus. For the far field the pressure attenuation is given by dX/dR ∼−3X/R. For most materials considered, n = 4–6, and therefore the near-field attenuation is proportional to R−3.65 - R−4 and the far-field attenuation is proportional to R−3. The calculations show that the volume of material shocked to pressures sufficient for melting should not be significantly different in sedimentary and crystalline rocks. Hence we conclude that shock melt is formed in the early stages of the cratering process by impacts into rocks rich in volatiles but is destroyed by the cratering process. We propose that the melt is finely dispersed by the great expansion of shocked volatiles upon release from high pressure and that suevite units are the product of this process. The fragmented silicates produced by this process may react penecontemporaneously with the hot volatiles to produce hydrated minerals such as clays. This process may produce hydrothermally altered minerals in planetary regoliths, such as the Martian regolith. The dispersion of shock melt by volatile expansion may also account for the apparent lack of lunarlike melt sheets on the surface of Mars. Because large amounts of volatiles vaporize during impact and are transferred from depth either into space, into the atmosphere, or onto near-surface ejecta by condensation, repeated impact degasses a planet, depleting some layers in volatiles and, unless the volatiles escape the planet, enriching others.