Fluid inclusions break, or decrepitate, when the fluid pressure exceeds the least principal lithostatic stress by a critical amount. After decrepitation, excess fluid pressure is relaxed, resulting in crack arrest; subsequently, crack healing may occur. Existing models of decrepitation do not adequately explain several experimentally observed phenomena. We developed a linear elastic fracture mechanics mdoel to anlayzed new data on decrepitation and crack arrest in San Carlos olivine, compared the model with previous fluid inclusion investigations, and used it to interpret some natural decrepitation microstructures. The common experimental observation that smaller inclusions may sustain higher internal fluid pressures without decrepitating may be rationalized by assuming that flaws associated with the inclusion scale with the inclusion size. According to the model, the length of the crack formed by decrepitation depends on the lithostatic pressure at the initiation of cracking, the initial sizes of the flaw and the inclusion, and the critical stress intensity factor. Further experiments show that microcracks in San Carlos olivine heal within several days at 1280 to 1400 ¿C; healing rates depend on the crack geometry, temperature, and chemistry of the buffering gas. The regression distance of the crack tip during healing can be related to time through a power law with exponent n=0.6. Chemical changes which become apparent after extremely long heat-treatments significantly affect the healing rates. Many of the inclusions in the San Carlos xenoliths stretched, decrepitated, and finally healed during uplift. The crack arrest model indicates that completely healed cracks had an initial fluid pressure of the order of 1 GPa. Using the crack arrest model and the healing kinetics, we estimate the ascent rate of these xenoliths to be between 0.001 and 0.1 m/s. ¿American Geophysical Union 1990