A common procedure in palaeodemography is to use mean skeletal age to estimate the expectation of life at birth in a population. This article shows that skeletal age and expectation of life at birth are not equivalent unless the population is stationary. This assumption is not justified for most real populations. Shows that mean skeletal age is approximately equivalent to the reciprocal of the birth rate and is not correlated with the death rate. Thus, the practice of inferring changes in life span and death rates from changes in mean age at death is not reliable and most conclusions of palaeodemographic studies should be revised. On the other hand, skeletal age may provide high-quality information about fertility in archaeological populations. Several published palaeodemographic studies are reinterpreted in light of the model presented.