The validity of estimates of unintended childbearing has often been questioned, especially given their almost exclusive reliance on responses to survey questions that ask women to recall their intentions about past pregnancies. An opportunity to compare prospective and retrospective descriptions of intendedness was provided by a follow-up survey in four Indian states in 2002-2003 of rural woman originally interviewed in the 1998-99 National Family Health Survey-2. The results demonstrate a pronounced tendency for births prospectively classified as unwanted to be retrospectively described as having been wanted or mistimed. The main reason seems to be either that mothers adapt to the reality of a new birth or are reluctant to describe an existing child as having initially been unwanted. Our findings suggest that retrospective accounts of the wantedness of a birth, such as those obtained by current Demographic and Health Surveys, may lead to significant underestimates of true levels of unwanted childbearing.