The stability of submarine gas hydrates is largely dictated by pressure and temperature, gas composition, and pore water salinity. However, the physical properties and surface chemistry of deep marine sediments may also affect the thermodynamic state, growth kinetics, spatial distributions, and growth forms of clathrates. Our conceptual model presumes that gas hydrate behaves in a way analogous to ice in a freezing soil. Hydrate growth is inhibited within finegrained sediments by a combination of reduced pore water activity in the vicinity of hydrophilic mineral surfaces, and the excess internal energy of small crystals confined in pores. The excess energy can be thought of as a capillary pressure in the hydrate crystal, related to the pore size distribution and the state of stress in the sediment framework. The base of gas hydrate stability in a sequence of fine sediments is predicted by our model to occur at a lower temperature (nearer to the seabed) than would be calculated from bulk thermodynamic equilibrium. Capillary effects or a build up of salt in the system can expand the phase boundary between hydrate and free gas into a divariant field extending over a finite depth range dictated by total methane content and pore-size distribution. Hysteresis between the temperatures of crystallization and dissociation of the clathrate is also predicted. Growth forms commonly observed in hydrate samples recovered from marine sediments (nodules, and lenses in muds; cements in sands) can largely be explained by capillary effects, but kinetics of nucleation and growth are also important. The formation of concentrated gas hydrates in a partially closed system with respect to material transport, or where gas can flush through the system, may lead to water depletion in the host sediment. This freeze-drying may be detectable through physical changes to the sediment (low water content and overconsolidation) and/or chemical anomalies in the pore waters and metastable presence of free gas within the normal zone of hydrate stability. ¿ 1999 American Geophysical Union