Nutritional statements that depend on observation or anecdote should be given serious consideration, but consideration should also be given to the physical and psychological quirks of the observer. The significance attached to an experimental conclusion depends, in part, on the scientific credentials of the experimentalist similarly, the significance of selected observations depends, again in part, on the preconceptions of the observer. Regimes that are proposed by people who do not look as if they enjoyed their food, and who do not themselves have a well-fed air, may not be ideal for normal people. Graham Lusk, who combined expert knowledge with a normal appreciation of good food, describes how he and Chittenden, who advocated a low-protein diet, spent some weeks in Britain eating the rations of the 1914-1918 war and then got more ample rations on board ship. Lusk attributed his sense of well-being to the extra meat he was eating Chittenden attributed it to the sea air. When young animals are reared for sale as meat, the desirable amount of protein in their food is a simple matter of economics. Protein is expensive, so the amount given is increased up to the level at which the increased rate of growth is offset by the increased cost of the diet. As already mentioned, the efficiency with which protein is used to build the body diminishes as the percentage of protein in the diet increases. In practice, the best diets seem to contain between 15 and 25 per cent protein. It is not certain that maximum growth rate is desirable in children some experiments with rats suggest that rapid growth is associated with a shorter ultimate expectation of life. There are practical and ethical obstacles to human experiments in which the effect of protein can be measured. Children do not grow as fast as the young animals in which there is a commercial interest. Their need for protein is therefore presumably smaller, but there is no evidence that the desirable protein level, after weaning, is less than 15 per cent. An argument against this percentage of protein is that in human milk only 13 per cent of the solid material is protein. That protein is, however, of better quality than any protein likely to be given to infants that are not weaned on cow's milk. Furthermore, milk, like other products of evolution, is a compromise. Mothers are not expendable. A species would not long survive if mothers depleted their own proteins so much in the course of feeding the first child that the prospects of later children were seriously jeopardized. Human milk is no doubt a good food, but the assumption that it is necessarily ideal is stretching belief in the beneficence and perfection of Nature too far. When considering nutritional statements, apart from statements that depend on observation, we should also consider______.