In Kapalowe rural health district, hospitalisation of malnourished children is restricted to complicated cases; once the complication is under control or eliminated, the child's treatment is continued at home, based on a 13 weeks contract, between parents and health centre. The parents commit themselves to feed their child four times a day (two porridge and two family dish portions), to consult once a week at the health centre and to welcome a weekly home visit. The objective of this visit is to support the parents, to detect possible problems and to reach the roots for this particular child. During the contract period, cost of medical treatment and recommended soya flour, is borne by the parents through a lump sum contribution. In this article, data concerning the first 95 children home rehabilitated (1989-1991) in Kapalowe are analysed. Characteristics of these children are classical regarding malnutrition; for example, age distribution is similar to that of weaning and of defunction of children at the hospital during the same year. Approximately half of them are still breastfed at the beginning of the contract. Most of them are correctly immunized and have been seen at the health centre at least two times in the last six months. Seventy-four children finished the contract; there were 17 abandons and 4 deaths. Weight gain is inferior to that observed in specialized feeding centres which do benefit from external resources, which is not the case here. It was not possible to show a significant catch up for the height for age indicator after the three months contract. These anthropometrical results are less important than the global and subjective improvement in the child's general health status observed at the end of the contract. None of the children reached the target weight after 13 weeks but important changes were observed in their behaviour, in their resistance to infection and in the attitude of their parents. The parents generally followed the instructions quite well. The middle of the contract seems to be a key period when either significative changes happen or when the attention is released. Treatment instructions have been amended to avoid monotony and overload, and to stimulate staff creativity and self-satisfaction. Payment was not a problem for the parents as malnutrition is not linked to extreme poverty. Mother's attitude and confidence and child initial weight for height status are two important contract success determinants. Abandons are more frequent when the mother is pessimistic and in case of kwashiorkor. Despite this, most of these children had gained more than one kilo before the contract was interrupted. Some didn't fulfill the W/H inclusion criteria (-2 standard deviations) and should probably not have been under contract. The four deaths were linked to insufficient treatment instructions for usually banal diseases that have another meaning in case of malnutrition, such as diarrhoea, fever, etc. An evaluation performed three months after the end of the contract in 26 children show 13 further improvements, 8 statu quo, 4 relapses and 2 new deaths. Conclusions are that home nutritional rehabilitation is possible where a health district is fully operational, that anthropometric data are useful to monitor rehabilitation but not to be pursued only as sole and ultimate objectives, and that adequate follow up after the first intensive stage is essential. The paper also shows how such a research result can have direct consequences on the organization of health activities.