Objective: To describe the effects of upper body training on the physical capacity of people with a spinal cord injury. Data sources: The databases of PubMed, CINAHL, Sport Discus and Cochrane were searched from 1970 to May 2006. Review methods: The keywords 'spinal cord injury', 'paraplegia', 'tetraplegia' and 'quadriplegia' were used in combination with 'training'. The methodological quality of the included articles (both randomized controlled trials and controlled clinical trials) was assessed with the modified 'van Tulder et al.' checklist. Studies were described with respect to population, test design, training protocol and mode of training. The training effects on physical capacity, reflected by peak power output (PO peak ) and oxygen uptake (VO 2peak ), were summarized. Results: Twenty-five studies were included with a mean score of 8.8 out of 17 items on the quality checklist. The methodological quality was quite low, mostly because of the absence of randomized controlled trials. Therefore no metaanalysis was possible. In the 14 articles of acceptable quality the mean (SD) increase in VO 2peak and PO peak, following a period of training, was 17.6 (11.2)% and 26.1 (15.6)%, respectively. Conclusions: Due to the overall low quality of studies it is not possible to draw definitive conclusions on training effects for different lesion groups or training modes. The results of the relatively few studies with an acceptable quality seem to support the view that upper body exercise may increase the physical capacity of people with spinal cord injury. The magnitude of improvement in PO peak and VO 2peak , however, varies considerably among studies.