The detection of mild cognitive impairment and dementia in high-functioning older adults can be difficult. It has also been observed that high-functioning persons show a lower prevalence of dementia than low-functioning persons. Three alternative explanations for this observation have been proposed in the literature: brain reserve capacity (BRC), cognitive reserve, and ascertainment bias. With data from a prospective, population-based study of incident dementia, the Canadian Study of Health and Aging (CSHA), we classified participants as being high- (HF) or low-functioning (LF) in three ways: educational and occupational attainment, and estimated premorbid IQ. We observed that fewer HF older adults were diagnosed with dementia after five years, which is in accordance with both the BRC and cognitive reserve models. Contrary to expectations, no difference on rate of memory deterioration was observed between those HF and LF persons who exhibited mild cognitive impairment at CSHA-1. However, HF persons who subsequently were diagnosed with dementia (CSHA-2) showed more rapid decline on five of the six memory measures over time than did LF persons diagnosed with dementia at CSHA-2. When performance on measures of memory functioning at CSHA-1 was examined for highly educated older adults, significantly more of those with dementia at CSHA-2 (n = 59) had scores falling within or below the average range in comparison to normative standards than those who continued to show no cognitive impairment (n = 145). Our findings suggest that the lower incidence of dementia for HF persons may be primarily the result of ascertainment bias, not underlying differences in brain or cognitive reserve.