Results from recent investigations of behavioral and genetic outcomes in older people with mild cognitive impairment (MCI) have been inconsistent. These conflicting results may be attributed to between-study differences in the diagnostic systems employed, as well as the use of unreliable neuropsychological measures. We investigated behavioral and genetic outcomes in older people classified as having MCI according to novel criterion that required evidence of cognitive impairment on three consecutive neurological/neuropsychological assessments. One hundred and seventy four healthy older people were evaluated semi-annually for 12 months. Of these, 23 subjects were rated as having MCI on three consecutive assessments and were compared to 23 matched control subjects. Subjects rated as impaired on one or two of the three semi-annual assessments were also identified. MCI and matched control groups were compared on a range of behavioral measures. The prevalence of the Apolipoprotein E4 (ApoE4) allele was determined in all groups, and estimates of anxiety and depressive symptomatology were obtained. Subjective cognitive complaints were also assessed. Many subjects were classified as impaired on one or two assessments, however relatively few (n = 23) recorded consistent cognitive deficits. The most severe impairment observed in MCI subjects was on a test of pattern-location associative learning, however MCI subjects did not have insight into this impairment. The prevalence of the ApoE4 allele was not different between matched control and MCI groups. These results indicate that individuals with MCI can be differentiated from healthy older people and older people with transient cognitive impairments, but that such differentiation requires serial assessment of cognitive function.