A hypothesis is generated that despite high reported rates, suicide among elderly Japanese females is substantially underestimated due to misclassification of drowning suicides (ICD-9 E954) as unintentional drownings (ICD-9 E910). Data are adapted from 1979-1981 age-, sex- and cause-specific mortality tabulations for Japan, the United States, Australia, France, New Zealand, Norway, Sweden and the United Kingdom. Between ages 55 and 74 years, unintentional drowning rates for males and females in Japan begin to diverge sharply from those of comparison countries. By ages 75 and older, the rate for Japanese females is 13.5 per 100,000, which exceeds comparison rates by 7- to 15-fold. Although drowning suicide rates in this population are also high, its ratio of drowning suicides to unintentional drownings declines precipitously beyond ages 35-44. Excess drowning suicide underestimation among Japanese females is suggested by the absence of a similar change among the males and evidence of both a lack of drowning witnesses and sex differentials in life expectancy, living arrangements and suicide methods. A preliminary test of the drowning suicide hypothesis is proposed which incorporates psychological autopsies.