Clinical failures of internal fixation implants for the treatment of the thoracolumbar spine are generally attributed to fatigue. Few studies, however, have characterized changes in fixation rigidity with time or subjected spine-implant fixation constructs to fatigue loading until failure. Fatigue characteristics of five dorsally applied spinal fixation implants were determined using lumbosacral calf spines, with an L3 vertebrectomy, loaded cyclically in combined compression (maximum 605 N) and flexion (maximum 16 Nm) for up to 100,000 cycles. Displacement transducers monitored motion at the site of instability and at the segment above the implants. Flexibility and strain at these segments were then calculated. A one-way analysis of variance showed that there were no significant differences in flexibility of the five fixation constructs (P > .05). A multiple Bonferroni test revealed that the AO and Kluger fixateur interne and Steffee plates, with fixation at L2 and L4, allowed significantly more strain (P < .01) across the site of instability than did Harrington rods and Luque plates with fixation at L1, L2, L4, and L5. There were no significant differences between fixation constructs in initial strain above the implants. After 10,000 cycles, however, there were significant increases in strain across the segment above the Luque and Harrington implants (P < .05). Failure of the AO Schanz screw occurred in three of six constructs at a mean of 73,300 cycles. The Steffee screws falled in four of five constructs at a mean of 20,800 cycles. The rods of the Kluger fixateur interne broke in four of five constructs at a mean of 47,800 cycles, and one screw slipped at 11,000 cycles. There were no metal failures in the Harrington or Luque implants. In these tests, longer implants allowed less strain across the destabllized site, and this can enhance the possibility of fusion with these implants. There was increased strain, however, across the segment above, which can be associated clinically with early degeneration and destabilization at this site. Short segmental fixation might therefore be preferable because fewer motion segments are immobilized and there is less strain across the adjacent segments. In vitro fatigue testing of spinal implants helps demonstrate potential design weaknesses so that improvements can be made before their clinical use in humans.