Objective: To describe the number, specifics and co-occurrence of shoulder pathologies detected by MRI in manual wheelchair users with spinal cord injury and to evaluate the association between shoulder pathologies and presence of shoulder pain. Design: Cross-sectional observation study. Setting: Community. Participants: Fifty-one wheelchair-dependent persons with spinal cord injury (44 males, 7 females, median age 50 years (IQR 14), median time since injury 24 years (IQR 16)) were allocated to pain or no-pain group based on the Wheelchair User Shoulder Pain Index. Interventions: Not applicable Outcome measures: All persons underwent shoulder MRI. Pathologies were scored blinded by two experienced radiologists. Participant characteristics, number and severity of shoulder pathologies were analyzed descriptively. Logistic regression was performed to evaluate the association between MRI findings and shoulder pain. Results: The median number of co-occurring MRI findings per person ranged from 0 to 19 (out of 31 possible findings). The cluster of MRI findings occurring most often together were tendon tears of supraspinatus (present in 84%), subscapularis (69%) and biceps (67%) and osteoarthritis of acromioclavicular joint (80%). When correcting for age and time since injury, the logistic regression showed no statistically significant correlation between the individual pathologies and shoulder pain. Conclusion: MRI findings of shoulder pathology are very frequent in persons with and without shoulder pain. Therefore, when diagnosing the cause of shoulder pain and planning interventions, health care professionals should keep this finding in mind and MRI should not be interpreted without careful consideration of clinical history and functional testing.