Birth weight is known to be closely related to child health, although as many infants in developing countries are not weighed at birth and thus will not have a recorded birth weight it is difficult to use birth weight when analysing the determinants of child illness. It is common to use a proxy for birth weight instead, namely the mother's perception of the baby's size at birth. Using DHS surveys in Cambodia, Kazakhstan and Malawi the responses to this question were assessed to indicate the relationship between birth weight and mother's perception. The determinants of perception were investigated using multilevel ordinal regression to gauge if they are different for infants with and without a recorded birth weight, and to consider if there are societal or community influences on perception of size. The results indicate that mother's perception is closely linked to birth weight, although there are other influences on the classification of infants into size groups. On average, a girl of the same birth weight as a boy will be classified into a smaller size category. Likewise, infants who died by the time of the survey will be classified as smaller than similarly heavy infants who are still alive. There are significant variations in size perception between sampling districts and clusters, indicating that mothers mainly judge their child for size against a national norm. However, there is also evidence that the size of infants in the community around the newborn also has an effect on the final size perception classification. Overall the results indicate that mother's perception of size is a good proxy for birth weight in large nationally representative surveys, although care should be taken to control for societal influences on perception.