We have previously reported that rehabilitative reaching training initiated 4 days following an incomplete cervical spinal cord injury (SCI) in adult rats promotes plasticity and task-specific recovery. This training, however, also resulted in impairments in an untrained task. Here we examined whether delaying the rehabilitative training following cervical SCI is still effective in promoting task-specific recovery, but circumvents impairments in an untrained task, comparable to what has been reported in stroke models. Therefore, reaching training for a period of 6 weeks was initiated at Day 12 following a cervical dorso-lateral quadrant lesion. Thereupon the rats’ ability to reach and to walk on a horizontal ladder (i.e. the untrained task) was assessed, and 8 weeks post-injury cortical map changes were investigated through microstimulation. Further, we examined changes in phospho protein kinase A (pPKA) levels following an immediate and a delayed onset of reaching training in rats with cervical SCI. We found that delayed rehabilitative training was comparably effective as immediate training in promoting task-specific recovery and sprouting of injured axons. Importantly, delayed training did not impair the performance on horizontal ladder walking. Strikingly, only delayed reaching training restored cortical PKA levels that had dropped significantly over 2 weeks post-injury. Additionally, delayed training did not influence cortical map changes following injury, but decreased white matter damage. In conclusion, our results show that a short delay in the onset of training in a forelimb task significantly alters our outcome measures, which should be considered in future rehabilitative approaches.