Objective: To develop a nutrition screening tool for use in older South Africans. Design: A cross-sectional validation study in 283 free-living and institutionalised black South Africans (60+ years). Methods: Trained fieldworkers administered a 24-hour recall and the Mini Nutritional Assessment (MNA) screening tool, and performed anthropometric measurements and physical function tests. Cognitive function was assessed using a validated version of the Six-Item Cognitive Impairment Test. Biochemical indicators assessed included serum albumin, haemoglobin, ferritin, vitamin B12, red-blood-cell folate, cholesterol and vitamin C. The MNA was used as the gold standard against which a novel screening tool was developed using a six-step systematic approach, namely: correspondence analysis; identification of key questions; determination of internal consistency; correlational analyses with objective measures; determination of reference cut-off values for categories of nutritional risk; and determination of sensitivity and specificity. Results: The new screening tool includes nine separate concepts, comprising a total of 14 questions, as well as measurement of mid-upper arm circumference. The new tool score was positively associated with level of independence in either basic activities of daily living (r=0.472) or the more complex instrumental activities of daily living (r=0.233). A three-category scoring system of nutritional risk was developed and shown to significantly characterise subjects according to physical function tests, level of independence and cognitive function. The new tool has good sensitivity (87.5%) and specificity (95.0%) compared with the MNA scoring system. It has a very high negative predictive value (99.5%), which means that the tool is unlikely to falsely classify subjects as well nourished/at risk when they are in fact malnourished. Conclusion: A novel screening tool has been shown to have content-, construct- and criterion-related validity, and the individual items have been shown to have good internal consistency. Further validation of the tool in a new population of elderly Africans is warranted.