This study investigates whether use of subject-specific muscle synergies can improve optimization predictions of muscle excitation patterns and knee contact forces during walking. Muscle synergies describe how a small number of neural commands generated by the nervous system can be linearly combined to produce the broad range of muscle electromyographic (EMG) signals measured experimentally. By quantifying the interdependence of individual EMG signals, muscle synergies provide dimensionality reduction for the neural control redundancy problem. Our hypothesis was that use of subjectspecific muscle synergies to limit muscle excitation patterns would improve prediction of muscle EMG patterns at the hip, knee, and ankle and of contact forces at the knee using a subject-specific lower body musculoskeletal computer model. The predictions were evaluated against in vivo experimental data collected from a subject implanted with a force-measuring tibial prosthesis.