A new class of statistical models is proposed for the analysis of longitudinal data, especially those from growth studies. The models are all derived from a simple univariate two-level polynomial model. It is shown that they make efficient use of available data, and can handle a very wide range of problems. They have several important advantages over existing procedures.