In this paper the ionization associated with afterglowing nitrogen is examined more thoroughly than has yet been done. It is found that this ionization is completely cut off if the testing vessel is separated from the afterglow by a silica wall. The ionization cannot therefore be directly produced by light, at any rate within the range from $\lambda $ 1850 to the visible. No increased current is observed when the surface action of the active gas is so vigorous as to make the testing cathode (of gold) red hot. The ionization is in general nearly the same when the cathode of the testing vessel is one of the common metals or a surface coated with metaphosphoric acid. Copper, however, is an exception. While the copper is clean it gives an effect several times larger, though this soon goes off, as it becomes dull by the action of the gas. It appears therefore that though there is some surface emission of electrons, it is in most cases small compared with the volume ionization in the gas space. Comparisons are made of the number of ion pairs generated, and the number of photons emitted, per cubic centimetre at various stages as the glow dies down. At first the number of photons is many times greater, but as the glow gets down the numbers become equal, and at intensities too low for convenient observation it is likely that the number of ion pairs may become greater than the number of photons. It is certain that the light emission and the ionization do not go down at the same rate. The latter goes down much more slowly, and gains relatively. The interpretation of this is discussed. It is found that the admission of inert nitrogen which is known to increase the (instantaneous) emission of photons, also increases the (instantaneous) ionization. It is pointed out that the ionization of nitrogen requires $15\cdot 51$ V, and the presence of this ionization seems to throw much doubt on the attempts to calculate the energy of active nitrogen from the band spectrum emitted, which suggests an energy of $9\cdot 6$ V.