Study design: Experiments in a mouse model of complete paraplegia. Objectives: To evaluate the effect of non-assisted treadmill training on motor recovery and body composition in completely spinal cord-transected mice. Settings: Laval University Medical Center, Neuroscience Unit, Quebec City, Quebec, Canada. Methods: Following a complete low-thoracic (Th9/10) spinal transection (Tx), mice were divided into two groups that were either untrained or trained with no assistance. Training consisted of placing the mice during 15 min with no further intervention (that is no tail pinching or body weight support) on a motorized treadmill (8-10 cm s À1 ) five times per week for 5 weeks. Locomotor performances were assessed weekly in both groups using two complementary locomotor rating scales. After 5 weeks, all mice were killed and adipose tissue, soleus, and extensor digitorum longus muscles were dissected for analyses. Results: No significant difference in locomotor performances or in muscle fibre type conversion was found between trained and untrained mice. In contrast, body weight, adipose tissue, whole muscle, and individual fibre cross-sectional area (CSA) values were significantly lower in trained compared with untrained animals. Conclusions: Non-assisted treadmill training in these conditions did not improve motor performances and contributed to further accentuate body composition changes post-Tx, suggesting that assistance provided manually, robotically, or pharmacologically may be key to spinal learning and recovery of locomotor function and body composition.