Plasma arginine vasopressin was assessed by radioimmunoassay in dehydrated elderly patients and healthy young and elderly control subjects. The elderly control subjects' AVP levels were higher than the young control subjects (Kruskal-Wallis rank sums z = 3.29; P less than 0.001). Despite similar plasma osmolalities, after correction for the osmotic contribution of urea, the patients' AVP levels were higher than the elderly controls (z = 4.16; P less than 0.001) and were in the range reported to evoke a maximal antidiuretic response. It is concluded that the dehydration which often accompanies acute illness in the elderly is not primarily a consequence of inadequate AVP release.