In order to investigate the mechanism of delayed progressive or secondary neuronal damage after the spinal cord injury, we developed a mild-compression injury model in the rat thoracic spinal cord. Our compression device consists of a soft silicone point of contact to the dura, in order to prevent violent injury that may cause axonal tears or hemorrhages in the spinal cord. Since rats often assume a 'standing' posture, i.e. raising head with lifting their fore-limbs, damage to the thoracic spinal cord was evaluated by measuring the frequency of 'standing', which effectively indicates hind limb function. Twenty-four hours after compression by a 20 g weight for 10 or 20 min, the standing frequency of the injured rat was almost the same as that of sham animals that underwent laminectomy without compression. However, the standing frequency decreased with time; the frequency of standing at 72 h was approximately 30-50% that of sham animals. In the compressed spinal cord tissue, microglial cells, detected by lectin staining, proliferated with time. An enormous amount of microglia was observed at 48 and 72 h after compression, although only a small amount of cells were positive to lectin staining at 24 h after the compression. These results suggest that our mild-compression spinal cord injury model showed late-onset or delayed neuronal damage that may be related to pathological microglia proliferation.