A total of 744 elderly women with fractured neck of femur were classified into three groups according to anthropometric measurements on admission: group 1, well nourished; group 2, thin; group 3, very thin. Group 1 ate well and had a low mortality and a short rehabilitation time. The thinner the patients the lower their voluntary food intake, the higher their mortality and the longer their rehabilitation time. A series of 122 patients from groups 2 and 3 were entered postoperatively into a randomised controlled trial of overnight supplementary nasogastric tube feeding (4.2 MJ (1000 kcal), including 28 g protein) in addition to their normal ward diet. This treatment was associated with improvements not only in anthropometric and plasma protein measurements but also in clinical outcome, especially in the very thin group 3 patients. Rehabilitation time and hospital stay were shortened. Mortality in group 3 was less in the tube fed patients (8%) than in the controls (22%) but this difference did not reach statistical significance. One in five patients could not tolerate the nasogastric tube, but in the remainder the treatment caused no side effects and did not seriously diminish voluntary oral food intake by day.