Although the theoretical steady-state equilibrium relationship between potential and temperature in an electrically heated conductor has been established since the beginning of the century (Kohlrausch 1900) the important case in which equilibrium breaks down has until recently been overlooked. In part I (Bowden & Williamson 1958) this was discussed experimentally with reference to gold contacts; in part II (Greenwood & Williamson 1958) it was shown theoretically that for suitable materials, equilibrium is only possible when the current is below a critical value. In the earlier experiments the metal contacts were free from contaminating films: this paper will describe the behaviour of iron sufaces which are contaminated by an oxide layer. It will be shown as before that conditions become unstable when the current is too large. Short pulses of current were passed through iron contacts. Currents below a critical size were found to have no effect; larger currents caused a growth of contact area and permanent decrease in contact resistance. When no contaminating films were present the product of the final contact resistance R$_0$ and the current I was constant. For iron contacts coated with varying amounts of oxide it was found that, instead of remaining constant, IR$_0$ increased steadily with the amount of contamination. The value observed for the cleanest contacts was close to the theoretical estimate for iron, while for badly contaminated ones it was two to three times as large. Experiments in which the length of the current pulse was varied demonstrate the dependence of the rate of heating on the contact area. Direct evidence for the unstable temperature rise is given by oscilloscope records of transient voltage across the contact during the passage of a current pulse. The observations are compared with a new theoretical estimate of the rate of temperature development in non-equilibrium conditions.