Plant breeding programmes rarely take explicit practical account of the two sources of genetic variance, namely between and within full-sib families, even though existence of these two sources of variation has long been recognised . This paper refers to inbred and clonal crops, not to outbred, seed-propagated species . Theory suggests that the two variances should be of similar size, sometimes very similar. Good comparisons have never been made because variance within families is laborious to estimate . It is proposed that sets of families be assayed as to means in formal trials and only the best few, judged against standards as having potential for superior segregates, should be exploited thoroughly. The calculation as to approximate equality of genetic variances between and within families is important . The vast majority of families should probably be discarded without further ado, and at considerable economy. Sensible decision-making requires an economic component in order to exploit the trade-off between the cost of the initial trial and families discarded without further cost . The object of this paper is a critical review of practical principles, not a general review of a large and diffuse literature .