Purpose <p>Musculoskeletal models of the head and neck have employed kinematic relationships as constraints to effectively resolve the redundancy in specifying the cervical intervertebral joint movement. However, the veracity of the underlying skeletal kinematics is questionable as those relationships obtained from in vitro cadaveric studies remain untested against cervical spine kinematics in vivo. In this work, we introduce a new head–neck model that incorporates an updated set of relationships, characterizing cervical spine rhythms, derived from dynamic stereo-radiography (DSX) measurement in vivo.</p> Methods <p>Data of fourteen subjects performing neck full-range flexion–extension movements measured by a DSX and a surface-based motion capture system were used for model development and evaluation. For each subject, a new model was constructed using OpenSim, driven by surface marker motion data but with the rhythmic relationships based on DSX data; a model was also constructed using the latest approach in OpenSim (the HYOID model) as the benchmark.</p> Results <p>Cervical spine rhythms, quantified as relationships of the flexion–extension rotations between the C6 and C7 and each of the superior intervertebral joints (C1–C2 to C5–C6), were found to be segment- and position-variant. The new model compared favorably with the latest OpenSim head–neck model: the root-mean-square-error (RMSE) against DSX measurement improved from 5.1° to 2.2° overall when averaged across six joints, and most substantially at the C1–C2 joint (from 17.3° to 2.1°).</p> Conclusion <p>The current study provides a clearer and quantitative understanding of the rhythmic, coordinated motion of the cervical spine and an improved head–neck musculoskeletal model for OpenSim and other modeling applications.</p>