BACKGROUND: Tests sensitive to presymptomatic changes in Alzheimer's disease could be valuable for clinical trials. Accelerated long-term forgetting-during which memory impairment becomes apparent over longer periods than usually assessed, despite normal performance on standard cognitive testing-has been identified in other temporal lobe disorders. We assessed whether accelerated long-term forgetting is a feature of presymptomatic autosomal dominant (familial) Alzheimer's disease, and whether there is an association between accelerated long-term forgetting and early subjective memory changes. METHODS: This was a cross-sectional study at the Dementia Research Centre, University College London (London, UK). Participants were recruited from a cohort of autosomal dominant Alzheimer's disease families already involved in research at University College London, and had to have a parent known to be affected by an autosomal dominant Alzheimer's disease mutation, and not report any current symptoms of cognitive decline. Accelerated long-term forgetting of three tasks (list, story, and figure recall) was assessed by comparing 7-day recall with initial learning and 30-min recall. 7-day recognition was also assessed. Subjective memory was assessed using the Everyday Memory Questionnaire. The primary outcome measure for each task was the proportion of material retained at 30 min that was recalled 7 days later (ie, 7-day recall divided by 30-min recall). We used linear regression to compare accelerated long-term forgetting scores between mutation carriers and non-carriers (adjusting for age, IQ, and test set) and, for mutation carriers, to assess whether there was an association between accelerated long-term forgetting and estimated years to symptom onset (EYO). Spearman's correlation was used to examine the association between accelerated long-term forgetting and subjective memory scores. FINDINGS: Between Feb 17, 2015 and March 30, 2016, we recruited 35 people. 21 participants were mutation carriers (mean EYO 7·2 years, SD 4·5). Across the three tasks, we detected no differences between carriers and non-carriers for initial learning or 30-min recall. The proportion of material recalled at 7 days was lower in carriers than non-carriers for list (estimated difference in mean for list recall -30·94 percentage points, 95% CI -45·16 to -16·73; p=0·0002), story (-20·10, -33·28 to -6·91; p=0·0048), and figure (-15·41, -26·88 to -3·93; p=0·012) recall. Accelerated long-term forgetting was greater in carriers nearer to their estimated age at onset (p≤0·01 for all three tests). Mutation carriers' 7-day recognition memory was also lower across all tasks (list [mean difference -5·80, 95% CI -9·96 to -2·47; p<0·01], story [-6·84, -10·94 to -3·37; p<0·01], and figure [-17·61, -27·68 to -7·72; p<0·01] recognition). Subjective memory scores were poorer in asymptomatic carriers compared with non-carriers (adjusted difference in means 7·88, 95% CI 1·36 to 14·41; p=0·016), and we found a correlation between accelerated long-term forgetting and subjective memory in mutation carriers. INTERPRETATION: Accelerated long-term forgetting is an early presymptomatic feature of autosomal dominant Alzheimer's disease, which appears to pre-date other amnestic deficits and might underpin subjective memory complaints in Alzheimer's disease. Accelerated long-term forgetting testing might be useful in presymptomatic Alzheimer's disease trials. FUNDING: MRC, NIHR, Alzheimer's Research UK, Dementias Platform UK, Dunhill Medical Trust, ERUK, Great Western Research, Health Foundation, Patrick Berthoud Trust.