Liquid potassium was circulated between 1200 and 1600°F in 31 Type-316 stainless-steel thermal convection loops and one forced circulation loop. Each loop contained a string of niobium-1% zirconium (Nb-1% Zr) alloy and stainless-steel test specimens positioned along the entire heated leg. To follow corrosion as a function of time and temperature, the test specimens were examined at 500 to 2500 h intervals. Controlled additions of interstitial impurities to the potassium were made in some thermal convection loops at the start of the test. Oxygen additions to the potassium sharply accelerated the initial rate of Nb-1% Zr surface removal but produced no identifiable oxide film or microstructural changes. The initially high weight-loss rates, observed in oxygen addition loops, decayed rapidly with time, returning essentially to normal rates (in the absence of further oxygen additions) after 2500 h. Oxygen additions produced very little effect on the stainless-steel corrosion rates, presumably due to rapid gettering of the added oxygen by the Nb-1% Zr. Similar tests in a forced circulation loop, with potassium velocities past the test specimens 18 times higher than in the thermal convection loops, showed that any effects of velocity on the Nb-1% Zr corrosion rate were far overshadowed by effects that are assumed to be related to oxygen in the potassium.