Although anthropometrics have traditionally been utilized in the assessment of protein-energy malnutrition, little is known of the reliability of these measurements in elderly patient populations. In this study, the reliability of an anthropometric measurement protocol, designed for ease in measurement of both independent and functionally debilitated elderly patients, is evaluated. An analysis of variance technique was used to calculate the test-retest variance for each of three observers. Although the experience of the three observers was equivalent, the precision of their measurements varied considerably. After repeat practice sessions, improved measurement reliability was demonstrated. For all the observers, the test-retest variance was found to be proportional to the size of the measurement and to be influenced by the accuracy in which the site for each measurement was located. Based on this study, we conclude that when using anthropometrics it is important to know the reliability of each observers' measurements. Because of the correlation between the test-retest variance and the measurement value, we suggest that the variance be expressed as a percentile. A method for transforming the variances is presented.