Although common among the hospitalized frail elderly, malnutrition is often unrecognized by clinicians, and its identification is frought with difficulty due to inadequate nutritional assessment methods and standards. This study compared the use of percent ideal body weight (%IBW) and anthropometry in the assessment of malnutrition in the hospitalized frail elderly. Approximately 45% of patients studied had at least two anthropometric measurements below the 5th percentile, a level reflecting severe malnutrition. However, only 28% of patients were found to be less than 90% IBW, a level reflecting only mild to severe changes in body weight. Serum albumin was below normal ( < 35 g/l) in 30% of patients. The total lymphocyte count was below normal ( < 1,500 cells/mm3) in 53%, and was severely depressed ( < 800 cells/mm3) in 24%. Thus, despite the use of stringent anthropometric criteria indicating severe malnutrition, anthropometry appeared more sensitive than %IBW as a measure of malnutrition in the hospitalized frail elderly. Furthermore, acute illness causes changes in commonly employed blood measures which make them unreliable in the assessment of malnutrition in this population. Anthropometry may prove to be the most stable, easily performed, and sensitive measure of malnutrition in the hospitalized frail elderly. However, further studies are clearly needed, including the development of appropriate anthropometric reference standards for the very old, a population that commonly suffers malnutrition.