Anthropometry is widely used to monitor infant growth and to estimate child nutritional status. Current evidence suggests that existent growth curves are not adequate for use with all infants and researchers sought to identify another data set suitable for development as a new international growth reference. In this article we cast about unconditional limits for growth monitoring from raw data on age, sex, height and weight. Anthrompometric data from children aged 1 to 9 years from two studies on malnutrition in Brazil was analyzed. Data on age, sex, weight, height and body mass index from 141 Amerindian children was used to develop mathematical models to predict percent of NCHS medians for weight-for-age, weight-for-height, and height-for-age using multiple linear regression. Data from 251 children in a non-indian seaside village was used for cross-validation. Six age-specific equations were obtained with coefficients of determination greater than 0.96. Coefficients of correlation between NCHS-derived and model-derived values into the validation data set were greater than 0.96 for weight-for-age, greater than 0.99 for weight-for-height, and near 1.00 for height-for-age. It remains to be seen if one can achieve universal linear models from more representative samples, using the approach described here. Perhaps establishing a mathematical relation among anthropometric data would result in absolute individual limits for growth monitoring. They may even be as important to infant nutritional assessment as growth reference values.