This article presents a method for directly calculating the area function of a vocal tract model by means of measured formant frequencies. The model is a concatenation of uniform tubelets whose cross-sections and lengths may vary in time. The proposed method uses a mathematical expression that relates time derivatives of tubelet cross-sections, lengths and formant frequencies. The derivatives are numerically integrated to arrive at cross-section and length trajectories. When more than one model area function is compatible with observed formant frequencies, additional constraints are used to select a unique area function. Results show that form ant-matched trajectories of tubelet cross-sections arid lengths are smooth. The agreement between observed and model-generated formant frequencies is better than 0.01 Hz