In amyotrophic lateral sclerosis (ALS), gait abnormalities contribute to poor mobility and represent a relevant risk for falls. To date, gait studies in ALS patients focused on the motor dimension of the disease, underestimating the cognitive aspects. Using a wearable gait analysis device, we compared gait patterns in ambulatory ALS patients with Mild Cognitive Impairment (ALS MCI+; n=18), and without MCI (ALS MCI-; n=24), and healthy individuals (HS; n=16) under two conditions: (1) normal gait (single task), (2) walking while counting backward (dual task). Finally, we examined if the occurrence and number of falls in the three months following the baseline test were related to cognition. In the single task condition, ALS patients, regardless of cognition, displayed higher gait variability than HS, especially for stance and swing time (p< 0.001). The dual task condition revealed additional differences in gait variability parameters between ALS MCI+ and ALS MCI- for cadence (p=0.005), stance time (p=0.04), swing time (p=0.04) and stability index (p=0.02). Moreover, ALS MCI+ showed a higher occurrence (p=0.001) and number of falls (p<0.001) at the follow-up. Regression analyses demonstrated that MCI condition predicted the occurrence of future falls (β= 3.649 p= 0.01) and, together with executive dysfunction, was associated with the number of falls (cognitive impairment: β= 0.63; p < 0.001; executive dysfunction: β= 0.39; p= 0.03), regardless of motor impairment at clinical examination. In ALS, MCI is associated with exaggerated gait variability and predicts the occurrence and number of short-term falls. This article is protected by copyright. All rights reserved.