An iterative procedure for making a life table is described, whereby a first draft, which can be very rough, is successively made to approach a final table, not dependent on the starting point, which meets a certain criterion of fidelity to the observations. At each successive table through which the process iterates an estimate is made of what the age-specific death rates in the actual population would be if that table were applicable, taking account of the rate of increase of the population as it is likely to affect the distribution within the five-year or other age groups. These age-specific rates are then compared with those observed, and an adjustment made for the next iteration. At most three iterations are sufficient to reproduce the observations to six places of decimals.