An implantable mechanical-chemical device was constructed to act as a feedback mechanism in controlling blood pressure. It consisted of a balloon connected to a rubber catheter ending in a slit valve. Flow-pressure curves were derived from in vitro testing for three valve thresholds (120, 140 and 170 mmHg). Five fast-acting hypotensive drugs subsequently filled the device during 40 nonadrenaline infusions in 25 dogs, with the balloon in the abdominal aorta and the catheter in the inferior vena cava. The results were as follows: (1) Following a short initial increase in systolic aortic pressure, significantly lower (p less than 0. 001) than in control experiments, the device prevented any pressure rise above its threshold (ii) The time needed for pressure lowering at the device's threshold depended on the drug used, being 3. 34 plus or minus 0. 84 (mean plus or minus s. e. in minutes) for sodium nitroprusside, 5. 99 plus 0. 96 for phentolamin, 11. 63 plus or minus 2 plus for hydralazine, 14. 52 plus or minus 2. 43 for a methyl-dopa and 23. 32 plus or minus 2. 07 for diazoxide.