To estimate the validity of neurological gait evaluations in predicting falls in older adults. We studied 632 adults age 70 and over (mean age 80.6 years, 62% women) enrolled in the Einstein Aging Study whose walking patterns were evaluated by study clinicians using a clinical gait rating scale. Association of neurological gaits and six subtypes (hemiparetic, frontal, Parkinsonian, unsteady, neuropathic, and spastic) with incident falls was studied using generalized estimation equation procedures adjusted for potential confounders, and reported as risk ratio with 95% confidence intervals (CI). Over a mean follow-up of 21 months, 244 (39%) subjects fell. Mean fall rate was 0.47 falls per person year. At baseline, 120 subjects were diagnosed with neurological gaits. Subjects with neurological gaits were at increased risk of falls (risk ratio 1.49, 95% CI 1.11–2.00). Unsteady (risk ratio 1.52, 95% CI 1.04–2.22), and neuropathic gait (risk ratio 1.94, 95% CI 1.07–3.11) were the two gait subtypes that predicted risk of falls. The results remained significant after accounting for disability and cognitive status, and also with injurious falls as the outcome. Neurological gaits and subtypes are independent predictors of falls in older adults. Neurological gait assessments will help clinicians identify and institute preventive measures in older adults at high risk for falls.