The GEM scale is an objective assessment tool evaluating walking safety of elderly individuals. It includes 33 walking items divided into three subscales (A, B, and C). The purpose of this study was to estimate the internal consistency and factorial validation of the scale. Seventy-four subjects (> or = 65 years) recruited from geriatric units were assessed by a total of 11 physical therapists. The internal consistency data were analyzed by using the Cronbach alpha coefficient. An exploratory factor analysis was performed to analyze the factorial structure of the scale and determine the construct validation. The internal consistency for the three subscales revealed high Cronbach alpha (subscales A = 0.90; B = 0.76, and C = 0.85). The preliminary analyses of the factorial validation did not confirm the original structure of the scale. For subscales A and B, a three-factor solution was supported by the analyses and explained 61% of the total variance. For subscale C, a four-factor solution was extracted and explained 87% of the total variance. The three subscales showed excellent item homogeneity. The factorial validation results support a new structure for the GEM scale regrouping the items in two sub-scales under different factors. The reorganization of the walking items into representative factors will allow a better understanding and interpretation of the scale.