In this paper, a distinction has been drawn between the demographic analyses that have been performed on fertility (at both the macro-structural and micro-structural levels) and the social analysis of fertility, at the micro-structural level. To stress the distinctive objectives of the latter, I have labelled the dependent variable in such analyses `family formation' rather than `fertility'. Along with this re-labelling, an increased stress was placed on the need for finding new measures of family formation, that would be integrative of past demographic measurement and take account of the costs or dependency of children. The latter seemed important to any socio-economic or social analysis of family formation. A general measure of family formation processes was proposed - the `child-years-of dependency' (CYD) index. This (1) integrates demographic measures of quantity and tempo of fertility; (2) appears to discriminate among family formation patterns where some of the conventional demographic measures (especially quantity measures) do not; (3) is readily calculable from fertility histories, which are currently used in conventional demographic surveys of fertility; (4) is very flexible in taking account of `deviant' or irregular family formation patterns, such as adoption, child mortality, etc., where most conventional quantity and tempo measures in demographic studies are not; and (5) has a face validity that appears greater for social analysis of family formation than conventional demographic measures. The stress of the CYD on the dependency burden of children led to a discussion of more valid cost estimates than the simpler counting of children present in the basic CYD measure. Among the issues treated were group differentials in average costs, declining marginal costs of child-rearing, and age gradients of costs. Costs were defined minimally, in terms of dollar expenditures per child for consumption, and increased household work time entailed in child-rearing. Some existing data were used to illustrate levels and social class differentials of these various costs, recognizing the general inadequacy of available data. The need was stressed for longitudinal cost data, following families through their successive stages of family formation to see the actual and perceived levels of costs involved in rearing additional children. Finally, the inclusion of CYD measures in a general cost-benefit model of family formation was suggested. In part, this is required to balance the perspectives derived from some macro-structural cost-benefit models of economists, in understanding the persistence of high fertility, especially in underdeveloped areas. [ABSTRACT FROM AUTHOR]