Methods: A finite element model of the upper cervical spine was developed. A type II dens fracture was induced in the intact model to produce the injured model. The following three constructs were simulated on the intact and injured models: transarticular screw (C1– C2TA), lateral mass screw in C1 and pedicle screw in C2 (C1LM1–C2PD), and lateral mass screw in C1 and translaminar screw in C2 (C1LM1–C2TL). Results: In the intact model, flexion–extension range of motion (ROM) was reduced by up to 99% with C11–C2TA and 98% with C1LM1–C2PD and C1LM1–C2TL. The lateral bending ROM in the intact model was reduced by 100%, 95%, and 75% with C11–C2TA, C1LM1–C2PD, and C1LM1–C2TL, respectively. The axial rotation ROM in the intact model was reduced by 99%, 98%, and 99% with C11–C2TA, C1LM1–C2PD, and C1LM1–C2TL, respectively. The largest maximum von Mises stress was predicted for C1LM1–C2TL (332 MPa) followed by C1LM1–C2PD (307 MPa) and C11–C2TA (133 MPa). Maximum stress was predicted to be at the lateral mass screw head of the C1LM1–C2TL construct. Conclusions Our model indicates that the biomechanical stability of the atlantoaxial joint in lateral bending with translaminar screws is not as reliable as that with transarticular and pedicle screws. Translaminar screws experience large stresses that may lead to failure of the construct before the required bony fusion occurs.