In refugee emergencies, rapid collection of nutritional data provides important information for public-health planning. In Rwandan refugee camps in eastern Zaire in August, 1994, a two-step procedure of screening for referral to supplementary feeding programmes was used--mid-upper-arm circumference (MUAC) followed by weight-for-height for children with MUAC of less than 12 cm. To assess the usefulness of this procedure, we analysed data from complete screening of 3681 children in three camps. The performance of MUAC varied with the cut-off chosen; a high cut-off of 14 cm allowed detection of 88% of children with low weight-for-height but at the cost of measuring more than 40% of children in the second step. MUAC preferentially selects younger children as malnourished, and misses older children with low weight-for-height. The groups of children chosen by low MUAC and by low weight-for-height have poor overlap, varying from 20% to 39% overlap depending on age. Thus two-step screening does not save as much time as might be expected and low MUAC cannot be used as a substitute for low weight-for-height. For decision-making in refugee settings, weight-for-height surveys or screening are probably more efficient strategies for data collection.