A survey of patients in a general surgical ward was undertaken to establish biochemical and anthropometric standards which could be used to detect malnourished patients in hospital. Results of biochemical and anthropometric tests of nutritional status were compared with assessment by a clinician and the quick nutritional index of Seltzer. Triceps skinfold thickness and serum albumin concentrations indicated that 29% and 35% of patients, respectively, were undernourished compared with 16% by clinical assessment and 17% by the quick nutritional index. Significant correlations (p less than 0.001) between serum albumin and transferrin concentrations and arm muscle area were found for men but not for women. Poor nutritional specificity and sensitivity of some anthropometric and biochemical tests may account for the difference in the level of undernutrition found by these tests and clinical assessment. This shows the importance of the choice of test in influencing the level of undernutrition detected.