We used a recently developed test of spatial memory--the Four Mountains Test--to investigate the core cognitive processes underpinning topographical disorientation in patients with amnestic mild cognitive impairment (a-MCI) and mild Alzheimer's disease (AD). Performance of these clinical groups was compared with age-matched controls, patients with frontotemporal lobar degeneration (FTLD), and patients with subjective memory impairments. We investigated the perception (concurrent match-to-sample) and short-term retention (2-s delayed match-to-sample) of the configuration of topographical features in computer-generated landscapes shown from different viewpoints. Thirty-one patients were tested (7 AD, 6 a-MCI, 7 temporal variant FTLD, 5 frontal variant FTLD, 6 subjective memory impairment) and 25 age- and gender-matched controls. Brain MRI was available for 27 patients; medial temporal lobe atrophy was assessed using a visual rating scale. Patients with a-MCI or mild AD were impaired on topographical short-term memory, but not perception. No other group differences were found on the topographical subtests. Notably, patients with temporal variants of FTLD performed normally, regardless of the laterality of damage. Subtests for the perception and retention of nonspatial aspects of the landscapes (weather conditions, seasonal and daily variations in lighting and color) were poor at differentiating the patient groups. These results indicate a core deficit in representing topographical layout, even for very short durations, within the context of more general long-term memory impairments found in AD, and suggest that this function is particularly sensitive to the earliest stages of the disease.