The quantal theory of negative ions has now been developed considerably (Massey and Smith 1936; Massey 1938), but on account of difficulties of computing it is usually necessary to assume rather than to prove that a given ion exists, and then to discuss the probability of its formation by different processes. The work described here is a contribution to the experimental side of this subject. It had its origin in a projected attempt to measure the capture cross-section of mercury atoms for electrons, as a verification of Massey and Smith's (1936) then unpublished theory of this process. In considering this it became clear that the experiment would be one of unusual difficulty. Before proceeding with it, therefore, it was decided to verify the existence of Hg - as a stable entity, which is assumed in the quantal theory. this was in doubt since Stille (1933), in some careful experiments, had recently failed to obtain it from the plasma of various forms of discharge through mercury vapour, although it is known that negative ions tend to accumulate in such regions (Emeleus and Sayers 1938). Whilst one of us was repeating his experiments, with modifications which led to essentially the same results and will be describes elsewhere, Arnot and Milligan (1936 b ) reported that they had obtained Hg - by bombardment of metal surfaces with Hg + . A comparison of their work with our own showed only one essential point of difference, namely that the construction of their apparatus did not permit of degassing in situ , a condition satisfied with our tubes. Both for this reason and because of the intrinsic importance of their discovery it was thought desirable to repeat part of their work, with apparatus geometrically similar in electrode construction, but capable of being degassed in a furnace under vacuum. We were again unable to obtain Hg - after the apparatus had been degassed and in operation for a short of that of Hg - was obtained. There were, however, always present several light negative ions, which had the excess energies found by Arnot and Milligan (1936 b ) with Hg - . The conditions under which these were formed led us to suppose that they were produced by bombardment of the metal surfaces by mercury positive ions (Press 1937; Sloane 1937) and not capture of electrons by positive ions of the same species, the process suggested by Arnot and Milligan (1936 b ). The existence of a process of this type, which may be conveniently termed "sputtering" (Sloane and Press 1938), is also implied by some earlier work by J. S. Thompson (1931) which has, so far as we know, never been published in detail. It is, however, impossible to decide definitely ion (e. g. CO - ) when it hits the surface, so long as one is producing the negative ions on a metal surface in a plasma or ionization chamber. One cannot overlook the possibility of the negative ion being formed from its own positive ion, since the latter may be present in the plasma or ionization chamber, and CO + , for example, was in fact shown in our work to be there from the positive-ion mass spectra, although in quantity only a fraction of 1% of Hg + . An unambiguous decision on this point can only be reached by first isolating a particular positive ion by a mass spectrograph, then bombarding a surface by this in a high vacuum , and finally making a mass spectrographic and energy distribution analysis of the resultant negative ions. We have built a double mass spectrograph for this purpose and find that negative ions of one kind possessing energies in excess of that imparted to them by the accelerating fields can be produced by bombardment with positive ions of another kind (Sloane and Press 1938). An account of the experiments with the single mass spectrograph is given in 1. the experiments with the double mass spectrograph are described in 2.