Current criteria for the diagnosis of Alzheimer's disease require that the patient also fulfils the criteria for dementia. This means inevitably that the disease is well-established. As treatments become available there is an important need for early diagnosis. There are two components to diagnosis; the recognition that cognitive performance is below that which would be anticipated for the individual and secondly to relate this to a specific disease process. Research studies need to identify the very earliest changes i.e. before symptoms commence. Such studies have focused on at risk populations either by virtue of age or a positive family history. Both groups have demonstrated that memory is the salient early feature, particularly verbal memory, impairment of which may precede overt symptoms by many years. The other major approach has been that of neuroimaging. Group studies of at risk individuals, for example for those who carry an apoE4 allele, have shown relative biparietal, bitemporal hypometabolism in the at risk group. Similarly, structural neuroimaging may show hippocampal atrophy at an early stage. Serial studies to determine rate of change, may be more valuable than single cross-sectional studies. Recent techniques of positional matching and registration allow precise quantitation of rates of cerebral atrophy which may be useful in early diagnosis.