Anthropometric indicators were used to assess the nutritional status of rural households in the former KwaZulu homeland of South Africa. Using these measures, households with stunted children were identified and the characteristics of these households examined, focusing on income sources, agricultural production and household demographics. The article then applies multivariate regressions and a logit maximum likelihood model to explore the relationship between household nutritional status (or child nutrition) and agricultural production among the sample of rural households. This study was a first attempt to link agricultural survey data with an assessment of household nutritional status in South Africa and proved to be successful. The results show that agricultural activities make a significantly positive contribution to household nutrition. Consequently, designing effective programmes for improving agricultural productivity in the less-developed areas of South Africa could have a potentially positive impact on household and child nutritional status.