Six anthropometric surveys using a random cluster sampling design have been organised in the capital of Zaire between November 1991 and March 1994. The prevalence of acute malnutrition fluctuated with the seasons and followed the cyclic movements of food prices on the market. The general tendency was rising. In March '94, 10.7% (95% confidence interval: 8.8-12.9%) of the children was suffering of acute malnutrition. Nevertheless, the standardised prevalence remained stable. This may suggest that the most disadvantaged living at certain poor zones of the city disconnected, while the rest of the population could still manage to maintain their nutritional status. The hypothesis of increasing heterogeneity is supported by the raising of the design-effect.