Highly complex engineering structures such as the body-in-white of a car consist of hundreds of different parts that are assembled using various types of joints such as welded and adhesive joints. The finite element method has been used extensively in various engineering fields to predict and analyze the dynamic behavior of assembled structures. However, despite the use of well-characterized individual subcomponent models, the predicted results frequently differ from measured results. This is believed to be because of the invalid assumptions of the joint model such as the material properties and other uncertainties associated with the assembly. This study investigated the use of a model updating method (MSC Nastran SOL 200) to identify the invalid assumptions and uncertainties in adhesive jointed composite structures and to minimize the difference between predicted and measured results. Results revealed that the stresses built-up in a structure during the assembly process are permanently retained in the structure after the adhesive is completely cured and that they cause a local stress stiffening effect, which is believed to affect the resonance frequencies of the structure. Introducing the effect of the stresses as an updating parameter in the model updating process increases the prediction accuracy.