The subjective Global Assessment technique was evaluated with regard to its reliability and validity in a Swedish elderly population. Ninety patients, 70 years of age or older, and newly admitted to the geriatric clinic, participated in the study. The nutritional state was assessed subjectively by two independent observers using the Subjective Global Assessment form, and objectively, using a combination of anthropometry and serum-protein analysis. The subjectively assessed malnutrition rate was about 20% higher than that assessed objectively. The agreement level between the two observers was 77.8%, Kappa 0.56 (< 0.001), and the levels between the subjective and objective assessments were 57.8% (Kappa = 0.17) and 73.3% (Kappa = 0.46), respectively. The experienced observer identified 85.2% of those who were defined as malnourished using the objective criteria and the less experienced observer, 66.6%. Furthermore, a regression analysis indicated weight index as the main significant objective factor influencing the subjective assessment. The subjective technique could be regarded as reliable from a clinical and preventive perspective since the agreement between the two rates was just under 80%. It seems also to be valid since the subnormalities in the selected criteria showed a similar pattern. The necessity of being well trained in order to manage nutritional status assessment, for both clinical and research purposes, is obvious.