ObjectiveTo distinguish between patients with amyloid-positive (A+) and -negative (A−) amnestic mild cognitive impairment (aMCI) by simultaneously investigating navigation performance, visual exploration behavior, and brain activations during a real-space navigation paradigm.MethodsTwenty-one patients with aMCI were grouped into A+ (n = 11) and A− cases by amyloid-PET imaging and amyloid CSF levels and compared to 15 healthy controls. Neuropsychological deficits were quantified by use of the Consortium to Establish a Registry for Alzheimer's Disease–plus cognitive battery. All participants performed a navigation task in which they had to find items in a realistic spatial environment and had to apply egocentric and allocentric route planning strategies. 18F-fluorodeoxyglucose was injected at the start to detect navigation-induced brain activations. Subjects wore a gaze-controlled, head-fixed camera that recorded their visual exploration behavior.ResultsA+ patients performed worse during egocentric and allocentric navigation compared to A− patients and controls (p < 0.001). Both aMCI subgroups used fewer shortcuts, moved more slowly, and stayed longer at crossings. Word-list learning, figural learning, and Trail-Making tests did not differ in the A+ and A− subgroups. A+ patients showed a reduced activation of the right hippocampus, retrosplenial, and parietal cortex during navigation compared to A− patients (p < 0.005).ConclusionsA+ patients with aMCI perform worse than A− patients with aMCI in egocentric and allocentric route planning because of a more widespread impairment of their cerebral navigation network. Navigation testing in real space is a promising approach to identify patients with aMCI with underlying Alzheimer pathology.