Undernutrition among elderly patients is a significant problem in nursing care. The aims of this study were to develop and test an instrument for identifying actual and potential undernutrition among elderly patients in clinical nursing care. A Likert-type scale consisting of 15 items was constructed. A consecutive sample of 56 elderly patients (>65 years) in a geriatric rehabilitation ward in western Sweden were interviewed with the instrument. The data were mainly analysed with non-parametric statistical methods. The results showed that the instrument was a fairly reliable scale with a Cronbach's alpha coefficient of 0.72. Evidence of validity concerning face validity, criterion-related validity-including concurrent and predictive validity-and construct validity was shown in the study group. Further testing is required if the instrument is to be used in clinical nursing care and research.