In egg-type poultry, variation in egg mass, body weight and change in body weight account for about 50–70% of the variation in feed consumed. In meat-type poultry, about 40% of the variation in feed consumption is attributable to variation in change of body weight (gain), and body weight itself. Past selection, primarily for egg numbers and egg size (egg mass) in egg stocks and for gain in meat stocks, has resulted in a correlated improvement in feed efficiency. Heterosis from breed or strain crosses contributes an average of 6% to improvement of feed efficiency in both egg and meat stocks. The success of direct selection for feed efficiency depends upon variation in feed consumption not attributable to economic traits in present poultry selection programs. There is evidence that such variation exists which could be successfully exploited in poultry. In egg stocks, the evidence is mainly from studies of variation within and between strains. In meat stocks, there is evidence from uni-trait selection studies as well as studies of within and between strain variation. The degree to which such selection would be successful in a multi-trait selection program and the economic viability of such success as compared to alternatives needs to be determined. Improvement of feed efficiency is possible by reducing feed requirements for maintenance and activity, by reducing behaviors causing feed wastage, or by reducing body fat to require less feed energy per unit gain. Feed wastage can be reduced by management techniques. Thus, the exploitation of feed-wasting behavior is not presently necessary. The best prospects are those associated with maintenance energy requirements in both egg and meat stocks of poultry, and the reduction of body fat in meat stocks. It is improbable that the opportunities for breeding to improve feed efficiency have been exhausted, much less identified. Key words: Breeding, feed efficiency, poultry, selection