Forty patients receiving respirator treatment were subdivided at random to receive tube-feeding with either a "home-made" powder product (group 1) or a ready-made sterile standard tube-feeding product: Salvimulsin MCT (group 2). Tube-feeding was administered as a continuous drip infusion and the patients were followed throughout three days, with particular attention to the development of diarrhoea. A series of cultures were taken from the tube-feeding diet, at the beginning and end of an infusion. The volumes were similar in both of the groups and averaged 860 ml on the first day and approximately 1,300 on the second and third days. Significantly greater bacterial contamination was found in the "home-made" tube-feeding diet, particularly with growth of Gram-negative rods. Seven patients (35%) in group 1 and two patients (11.1%) in group 2 developed diarrhoea. This difference is not statistically significant. No correlation was found between hypoalbuminaemia and the development of diarrhoea. It is concluded that powder mixtures and dilution of tube-feeding diets should be avoided on account of the risk of contamination and, similarly, tube-feeding should not be administered for more than 6-8 hours.