Objective: Recollection and familiarity are independent processes that contribute to memory performance. Recollection is dependent on attentional control, which has been shown to be disrupted in early stage Alzheimer’s disease (AD), whereas familiarity is independent of attention. The present longitudinal study examines the sensitivity of recollection estimates based on Jacoby’s (1991) process dissociation procedure to AD-related biomarkers in a large sample of well-characterized cognitively normal middle-aged and older adults (N = 519) and the extent to which recollection discriminates these individuals from individuals with very mild symptomatic AD (N = 64). Method: Participants studied word pairs (e.g., knee bone), then completed a primed, explicit, cued fragment-completion memory task (e.g., knee b_n_). Primes were either congruent with the correct response (e.g., bone), incongruent (e.g., bend), or neutral (e.g., &&&). This design allowed for the estimation of independent contributions of recollection and familiarity processes, using the process dissociation procedure. Results: Recollection, but not familiarity, was impaired in healthy aging and in very mild AD. Recollection discriminated cognitively normal individuals from the earliest detectable stage of symptomatic AD above and beyond standard psychometric tests. In cognitively normal individuals, baseline CSF measures indicative of AD pathology were related to lower initial recollection and less practice-related improvement in recollection over time. Finally, presence of amyloid plaques, as imaged by PIB-PET, was also related to less improvement in recollection over time. Conclusions: These findings suggest that attention-demanding memory processes, such as recollection, may be particularly sensitive to both symptomatic and preclinical AD pathology.