Approximately 30% of healthy persons aged over 75 years show Abeta deposition at autopsy. It is postulated that this represents preclinical Alzheimer's disease (AD). We evaluated the relationship between Abeta burden as assessed by PiB PET and cognitive decline in a well-characterized, non-demented, elderly cohort. PiB PET studies and cognitive tests were performed on 34 elderly participants (age 73+/-6) from the longitudinal Melbourne Healthy Aging Study (MHAS). Subjects were classified as being cognitively 'stable' or 'declining' by an independent behavioural neurologist based on clinical assessment and serial word-list recall scores from the preceding 6-10 years. Decline was calculated from the slope of the word-list recall scores. Abeta burden was quantified using Standardized Uptake Value normalized to cerebellar cortex. Ten subjects were clinically classified as declining. At the time of the PET scans, three of the declining subjects had mild cognitive impairment, one had AD, and six were declining but remained within the normal range for age on cognitive tests. Declining subjects were much more likely to show cortical PiB binding than stable subjects (70% vs. 17%, respectively). Neocortical Abeta burden correlated with word-list recall slopes (r=-0.78) and memory function (r=-0.85) in the declining group. No correlations were observed in the stable group. Abeta burden correlated with incident memory impairment and the rate of memory decline in the non-demented ageing population. These observations suggest that neither memory decline nor Abeta deposition are part of normal ageing and likely represent preclinical AD. Further longitudinal observations are required to confirm this hypothesis.