1. The relative potencies of amiloride (5 and 20 mg) and spironolactone (25 and 100 mg) for plasma and erythrocyte electrolytes were investigated in a double-blind, randomised, balanced, crossover study in 12 normal men treated concomitantly with hydrochlorothiazide 100 mg daily for 1 week. 2. Participants satisfied an a priori requirement for a fall in plasma potassium concentration of at least 0.5 mmol l-1 after 7 days of treatment with hydrochlorothiazide alone. 3. After hydrochlorothiazide alone, plasma potassium and sodium concentrations fell (P < 0.001). There were associated reductions in erythrocyte sodium (P < 0.01). Plasma magnesium concentration did not change, although erythrocyte magnesium decreased (P < 0.001). 4. Both amiloride and spironolactone attenuated the thiazide-induced fall in plasma potassium (relative potency, amiloride:spironolactone 10:1, 95% confidence interval 6.3-16.2:1). Amiloride but not spironolactone was associated with a dose-related increase in plasma magnesium; a relative potency estimation was precluded. There was little evidence of influences of amiloride or spironolactone on erythrocyte electrolytes. 5. On a weight basis, amiloride is ten times more potent than spironolactone as a potassium-sparing agent in diuretic-treated subjects but neither agent had major effects on erythrocyte potassium. The drugs may have divergent actions on magnesium handling; hydrochlorothiazide alone had no influence on plasma magnesium.