1. 288 arterial blood samples were obtained at successive times, through indwelling catheters, from conscious rats subjected to selective dietary potassium restriction for up to 5 weeks. 2. In control rats with unrestricted access to potassium [HCO-3]a = 25.4 mmol/l, pHa = 7.47, PaCO2 = 34 mmHg, PaO2 = 94 mmHg and [K+]a = 4.3 mmol/l were steady and within the range of values reported in conscious rats. Five weeks of isolated potassium restriction resulted in significant hypokalaemia ([K+]a = 2.15 mmol/l) and metabolic alkalosis ([HCO-3]a = 34.1 mmol/1, pH 7.57). This alkalosis, due mostly to H+ transfer into cells, was accompanied by a significant increase in PaCO2 to 36.4 mmHg and decrease in PaO2 to 88 mmHg, contrary to previous reports in man and dog. The administration of neutral sodium phosphate in addition to potassium restriction enhanced both the alkalosis ([HCO-3]a = 42.3 mmol/l, pH 7.61) and its respiratory compensation (PaCO2 = 40.8, PaO2 = 82 mmHg), without altering the PaCO2: [HCO-3]a relationship. 3. The opposite variations of PaCO2 and PaO2 were significantly correlated to the increase in plasma bicarbonate concentration and described best by crossed sigmoid curves. The equations of both curves were calculated; their point of inflection occurred at the same bicarbonate concentration (36 mmol/l). The maximum intensity of respiratory compensation (0.62 mmHg PaCO2 for each mmol [HCO-3]a/l) observed at this point fell within the range of values yielded by previous estimations.