Cross-sectional studies and longitudinal studies alike make regular use of retrospective questions about childhood circumstances. However, little is known about the accuracy with which adults can recall this kind of information. This paper seeks to address this topic by comparing retrospective reports of the number of people and the number of rooms in one’s household at age 11 provided by 50 year old members of a birth cohort study, with responses provided contemporaneously by their parents. The paper demonstrates encouraging levels of consistency between retrospective and contemporaneous reports. By examining reports of number of rooms provided by parents living at the same address in two earlier sweeps of the study (at ages 7 and 11), the paper shows that responses to contemporaneous questions may also be inconsistent, suggesting that retrospective questions of this nature may not be hugely less reliable. A retrospective measure of overcrowding at age 11 is derived using the two variables, and compared with a contemporaneous measure. The two measures lead to the same estimate of the extent of overcrowding, but when used in a model examining the odds of experiencing lung problems as an adult, the two measures behave differently. The paper also demonstrates that there are particular groups who are more likely to provide inconsistent responses than others. Around one in five participants were identified as having particularly poor recall, and the likelihood of being in this group was considerably higher amongst those whose childhood circumstances were more complex. The paper also finds that performance in a delayed memory assessment at age 50 was associated with better recall of childhood circumstances.