There is a real and growing danger of detonation of unconfined fuel-air clouds formed by massive industrial gas spills, and the danger increases with the size of the spill. A simple flame is probably not enough to initiate detonation of an unconfined gas cloud no matter how large the scale; but small contributory explosions such as a bursting pressure vessel or a muffler backfire can initiate detonations if strategically located, and gas clouds burning under confinement can undergo spontaneous transition to detonation and initiate open-air clouds. Large scale spills present a greater likelihood of detonation than do smaller spills simply because they entail more opportunities for confined burning to occur or for the cloud to reach an initiation source. Their burning phenomena do not appear to be intrinsically different from those of smaller spills. The most serious hitherto-unrecognized hazard in a refinery is the widespread existence of confined spaces in which a gas cloud can burn to detonation. They include spaces under floors and behind panels, the sewer system, stacks of empty pipe lengths and numerous others. Theoretically, a large spill of flammable gas could mix with the stoichiometric amount of air and possess the explosive power of a small atomic bomb; but experience has shown that the actual yield seldom exceeds 5% of the potential yield - although gas which does not participate in an explosion can still contribute to damage by participating in a subsequent firestorm. Techniques are available to make a quantitative assessment of the damage potential of a given spill in a given situation, and such an assessment would provide top management with the information needed to make an informed risk decision. Techniques are also available to reduce the risk of explosion in the event of a spill. They include limiting the scale of processing, limiting the growth of spill clouds by pilot lights, elimination of confined spaces in which a flame can grow to detonation, greater use of flame arrestors and the installation of gas quench systems. This report is a scouting and planning study intended to give an appreciation for the kind and amount of information that exists and to assess what might be accomplished by a full-scale safety R & D program. It presents a recommended program in sufficient detail for management evaluation and action; with the data and evidential basis for the recommendation.