In lumbar fusion, controversy remains regarding the effectiveness of cross-linking and the necessity of placing pedicle screws at the intermediate levels of the segment to be fused. The purpose of this study is to evaluate the stiffness of various rod/screw constructs used to instrument a three-level fusion with specific emphasis on the effect of cross-linking and the intermediate pedicle screws. Nine lumbar calf spines were mounted at L1 and L5. Pedicle screws (TSRH, Danek, Memphis, TN) were then placed bilaterally in the L2, L3, and L4 pedicles. Random sequence testing of the following constructs was then conducted: TSRH rods connected bilaterally to the L2 and L4 pedicles with and without a cross-link, and rods connected bilaterally at the L2, L3, and L4 levels with and without a cross-link. The tests were conducted on a modified MTS testing machine (MTS, Minneapolis, MN) and consisted of cyclic application of axial load, torsion, and flexion and extension. The tests yielded axial, sagittal, and torsional stiffness values. Statistical analysis was performed using log transformation and Fischer's test of least significant difference. In axial testing the use of additional screws in the intermediate pedicles increased stiffness an average of 160% (p = .007). The addition of a cross-link did not increase stiffness with axial loading. In flexion testing the six-screw construct was 84% stiffer when compared with the four-screw construct (p = 0.0001). There was no significant change in flexion stiffness with addition of cross-links. In torsional testing the six-screw construct was 38% stiffer than the four-screw construct (p = 0.042).(ABSTRACT TRUNCATED AT 250 WORDS)