The ability of Rossavik growth models, determined from measurements obtained before 24 weeks, to predict third-trimester growth and birth characteristics in normally growing twins has been investigated. Third-trimester values for head circumference, abdominal circumference, and femur diaphysis length were predicted with an accuracy of +/- 6% to 9% (95% to 98% of percent deviations). For thigh circumference and estimated weight, the comparable values were +/- 15% and +/- 16%, respectively. The head circumference at birth was predicted without bias; the random error was approximately +/- 5% (94% of percent differences). Weight, abdominal circumference, and thigh circumference were systematically overestimated (3.1%, 14.9%, and 11.3%, respectively) as a result of differences in prenatal and postnatal measurement procedures. After correction for systematic errors, these parameters could be predicted with random errors of -11.5% to 7.2% (weight), -12.8% to 5.4% (abdominal circumference), and -15.3% to 10.0% (thigh circumference). Growth Potential Realization Index values were found to have means of approximately 100% and ranges from 91% to 118%. These results are similar to those for singletons and indicate that individual assessment of growth in twins can be carried out with the same methods used for singletons.