A boundary element method (BEM) model for acoustic responses in individual ear canals was developed, and its validity was assessed and updated using a statistical approach. The BEM model was developed using high resolution computed tomography (CT) scanning from a healthy male adult. The BEM model, which was discretized adaptively for a frequency band of up to 22 kHz, included the head, pinna, ear canals, and tympanic membranes. The variability of the BEM model due to the uncertain acoustic boundary conditions and measurement locations was predicted using the eigenvector dimension reduction (EDR) method. Then, the likelihood function estimation approach was introduced in order to measure the agreement between the acoustic responses of the simulation model and experimental results. In order to enhance the simulation model performance, the acoustic boundary conditions of the BEM model were updated using a statistical calibration approach that maximizes the likelihood function value between the calculated probability density functions (PDFs) of the simulation model and the measurement data. The results of the validation and calibration procedures applied to the BEM model demonstrated that the proposed method provides a very effective method of verifying the model validity and enhancing the performance of the simulation model.