This study evaluated the biofidelity of both the Hybrid III and the THOR-NT anthropomorphic test device (ATD) necks in quasistatic tension-bending and pure-bending by comparing the responses of both the ATDs with results from validated computational models of the living human neck. This model was developed using post-mortem human surrogate (PMHS) osteoligamentous response corridors with effective musculature added (Chancey, 2005). Each ATD was tested using a variety of end-conditions to create the tension-bending loads. The results were compared using absolute difference, RMS difference, and normalized difference metrics. The THOR-NT was tested both with and without muscle cables. The THOR-NT was also tested with and without the central safety cable to test the effect of the cable on the behavior of the ATD. The Hybrid III was stiffer than the model for all tension-bending end conditions. Quantitative measurement of the differences in response showed more close agreement between the THOR-NT and the model than the Hybrid III and the model. By contrast, no systematic differences were observed in the head kinematics. The muscle cables significantly stiffened the THOR-NT by effectively reducing the laxity from the occipital condyle (OC) joint. The cables also shielded the OC upper neck load cell from a significant portion of the applied loads. The center safety significantly stiffened the response and decreased the fidelity, particularly in modes of loading in which tensile forces were large and bending moments small. This study compares ATD responses to computational models in which the models include PMHS response corridors while correcting for problems associated with cadaveric muscle. While controversial and requiring considerable diligence, these kinds of approaches show promise in assessing ATD biofidelity.