The cells of Halobacterium sp., strain 5, contain a large number of highly refractile bodies of the type which Petter (1932) suggested were gas-filled vacuoles. The present studies support Petter's contention, but the evidence for the exact chemical nature of the vacuole content is still indirect. It is not carbon dioxide or oxygen, but might possibly be nitrogen. “Strain 5” loses spontaneously and with a high frequency the ability to make the vacuoles. When vacuolated cells are subjected to pressure, the vacuoles disappear, but can recover upon aeration. Oxygen and the organic constituents of the growth medium stimulate the recovery, whereas 2.4-dinitrophenol inhibits it. A procedure is described for the isolation of the vacuoles. The vacuoles are bounded by a membrane which reveals itself in electron micrographs of thin sections as a 1-layered structure about 30 A thick.