In the first half of each intercensal decade in the thirty-year period 1925 to 1955, the age-adjusted death rates for nonwhite persons have declined relatively more rapidly than for white persons. In the second half of each intercensal decade just the opposite situation prevails, an excess of white over nonwhite decline. There are no exceptions to this curious pattern for either males or females in this period. Partial explanation of this phenomenon is suggested in terms of increasing underestimates of the nonwhite relative to the white population in the two intercensal periods 1930–40 and 1940–50.