The optimal mechanical loading regimen for the healing of a tendon graft in a bone tunnel is unknown. We developed a rat model that directly tensions a healing tendon graft, without the use of confounding joint motion. Fifty cycles of either 0, 3, or 6 N of tension were applied to groups daily for 3 or 6 weeks. At 3 weeks the low load (3 N) group had the highest failure load (p = 0.009), but by 6 weeks there were no differences in failure load among groups. At 3 weeks the high load (6 N) group had greater osteoclast activity compared to the immobilized (0 N) group (p < 0.05), and by 6 weeks there were significantly more osteoclasts in the high load group compared to the low load group (p = 0.01). Bone volume fraction was higher in the immobilized group compared to the 3 N load group at 3 weeks (p = 0.014) and 6 weeks (p = 0.007). At 6 weeks, the immobilized group had greater trabecular number compared to both loading groups (p < 0.05). In conclusion, low magnitude loading had a beneficial early effect but continued loading led to poorer new bone formation over time and no beneficial effect at 6 weeks, perhaps due to delayed maturation from cumulative loads. © 2015 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res 34:852–859, 2016.