Cervical and upper thoracic spinal cord injury causes impairments in respiratory muscle performance, leading to variable degrees of pulmonary dysfunction and rendering deep breathing difficult for affected individuals. In this retrospective study, we investigated the effects of self-directed respiratory muscle training in this context by assessing pulmonary function relative to spinal cord injury characteristics. A total of 104 spinal cord injury patients (tetraplegia/paraplegia; 65/39, acute/subacute/chronic; 14/42/48) were admitted for short-term (4–8 weeks) in-patient clinical rehabilitation. Initial evaluation revealed a compromised pulmonary function with a percentage of predicted value of 62.0 and 57.5 in forced vital capacity in supine and forced vital capacity in sitting positions, respectively. Tetraplegic patients had more compromised pulmonary function compared with paraplegic patients. At follow-up evaluation, the percentage of predicted value of forced vital capacity in supine and sitting position improved overall on average by 11.7% and 12.7%, respectively. The peak cough flow improved by 22.7%. All assessed pulmonary function parameters improved significantly in all subgroups, with the greatest improvements found in patients with tetraplegia and subacute spinal cord injury. Therefore, short-term self-directed respiratory muscle training should be incorporated into all spinal cord injury rehabilitation regimens, especially for patients with tetraplegia and subacute spinal cord injury, as well as those with chronic spinal cord injury.