The rigidity of a pedicle screw implant is a critical biomechanical variable in lumbar spinal fusions. Sufficient rigidity is required for integration of bone grafts and to promote healing. Osteopenia, stress shielding, and compensatory hypermobility have been described as consequences of excessive rigidity. Little is known about the biomechanical characteristics of "semirigid" compared to "rigid" implants. A new implant, whose rigidity can be varied by selection of different implant components, was tested in vitro under well-defined loading conditions. The three-dimensional load-displacement behavior of all lumbar vertebrae involved in or adjacent to the two-level fusion was evaluated for two fusion modifications: bilateral rigid and bilateral semirigid. Cyclic fatigue loading was subsequently carried out under realistic conditions and motion testing repeated. The rigid device reduced the motion of the L3-4 transfixed segment in the primary movement planes by 87.3% with respect to the intact spine value in flexion/extension (FE), 86.3% in lateral bending (LB), and 76.8% in axial rotation (AR). The semirigid device achieved a reduction in motion of 79.6% (FE), 82.7% (LB), and 51.7% (AR). The semirigid implant was particularly easy to insert, because no bending of rods or plates was necessary. The implants showed no loosening or breakage after the fatigue testing. The results are compared to other available systems and the underlying biomechanics discussed.