In most studies, feed efficiency is defined as the ratio of output (e.g. milk yield or gain) to feed consumption or its inverse. Several investigations have shown that selection for higher milk yield improves gross feed efficiency. It has been estimated that correlated response to direct selection for increased milk yield probably results in 70–95% of the potential improvement that could be achieved through direct selection for feed efficiency. However, in most studies, cows were fed according to production and this leads automatically to a high correlation between milk yield and gross feed efficiency. Variation between animals in appetite, digestion and nutrient absorption, maintenance requirement, utilization of metabolizable energy for production, nutrient partitioning and output composition makes gross feed efficiency an imprecise measure of feed efficiency. Evidence on genetic variability of these different characteristics is scanty, and to a certain extent conflicting. Feed intake in cattle is moderately heritable and selection for its improvement would theoretically be effective. Animals of high genetic merit for milk have greater intakes, and use more of their body reserves in early lactation than those of low merit. In the first part of lactation intakes of nutrients, especially energy and protein, do not meet requirements of the high-yielding dairy cow. Investigations suggest that selection on milk yield would not automatically increase feed intake of dairy cows in the first part of lactation. In several species, differences between animals in using metabolizable energy for maintenance and production have been measured by residual feed consumption. Results in cattle show moderate heritabilities for this trait.