Cathodes of sintered nickel and barium oxide have been used as ion sources in simple mass spectrometers and have been found to emit negative ions of atomic oxygen for at least 24 hours during their activation at 1250-1350° K. The evidence suggests that the oxygen is produced by the dissociation of barium oxide and is removed from the cathode by positive ion bombardment. At voltages greater than the ionization potentials of the residual gases, the variation of the oxygen ion current I with time t could be expressed as the sum of two exponentially decreasing terms: I = a exp ( - pt) + b exp ( - qt), where p and q were of the order of 2 and 0.2 per hour respectively. Similar results have been obtained with a commercial cathode containing triple carbonates. There appear to be two processes operating in parallel, corresponding to the two terms of the equation. It is postulated that the first process is the diffusion of oxygen through the bulk of the crystals to the nickel part of the cathode's surface, and its removal from there by bombardment with positive ions of oxygen and nitrogen. The second process is the diffusion of oxygen through the pores of the cathode, and its removal from the ends of the pores by bombardment with positive ions of barium as well as of the residual gases. The first process has the slower rate of replacement of oxygen and in an hour or two this process is insignificant compared with the second one. If the cathode's surface is temporarily poisoned with oxygen from an external source then it rids itself of the extra oxygen within a few seconds.