The development and testing of an observational gait assessment instrument, the Spinal Cord Injury Functional Ambulation Inventory (SCI-FAI) is described. To assess validity and reliability, 22 subjects with incomplete spinal cord injury were scored by four raters. Raters scored each subject three times, once live (LS) and twice from videotaped records (VS1, VS2). A moderate-good negative correlation (r = -0.742 and -0.700, for VS1 and VS2, respectively) was found between the gait score and time required to walk a demarcated path. Inter-rater reliability was moderate-good for the live score and the videotaped records (ICC = 0.703, 0.800, and 0.840, respectively). Intra-rater reliability was good (ICC = 0.903, 0.960, 0.942, and 0.850 for Raters 1-4, respectively). To assess sensitivity, another group of 19 subjects with SCI were assessed prior to and following participation in an intensive walking program. A moderate correlation was found between change in gait score and change in lower extremity strength (Pearson r = 0.58). These results indicate that the SCI-FAI is a reliable, valid and sensitive measure of walking ability in individuals with spinal cord injury. In addition, the results suggest that gait analysis using this instrument is equally reliable whether the observation is performed live or from videotaped records.