A better knowledge of long-term trajectories of cognitive decline is a central feature of the study of the process leading to Alzheimer's dementia. Several factors may mitigate such decline, among which is education, a major risk factor for Alzheimer's disease. The aim of our work was to compare the pattern and duration of clinical trajectories before Alzheimer's dementia in individuals with low and high education within the PAQUID cohort involving 20 years of follow-up. The sample comprises 442 participants with incident Alzheimer's disease (27.2% were male)--171 with low education (mean age=86.2 years; standard deviation=5.3 years) and 271 with higher education (mean age=86.5; standard deviation=5.4)--and 442 control subjects matched according to age, sex and education. At each visit and up to the 20-year follow-up visit, several cognitive and clinical measures were collected and incident cases of Alzheimer's disease clinically diagnosed. The evolution of clinical measures in pre-demented subjects and matched controls was analysed with a semi-parametric extension of the mixed effects linear model. The results show that the first signs of cognitive decline occurred 15 to 16 years before achieving dementia threshold in higher-educated subjects whereas signs occurred at 7 years before dementia in low-educated subjects. There seemed to be two successive periods of decline in higher-educated subjects. Decline started ∼15 to 16 years before dementia with subtle impairment restricted to some cognitive tests and with no impact during the first 7 to 8 years on global cognition, cognitive complaints, or activities of daily living scales. Then, ∼7 years before dementia, global cognitive abilities begin to deteriorate, along with difficulties dealing with complex activities of daily living, the increase in self-perceived difficulties and depressive symptoms. By contrast, lower-educated subjects presented a single period of decline lasting ∼7 years, characterized by decline concomitantly affecting specific and more global cognitive function along with alteration in functional abilities. This study demonstrates how early cognitive symptoms may emerge preceding Alzheimer's dementia particularly in higher-educated individuals, for whom decline occurred up to 16 years before dementia. It also demonstrates the protective role of education in the clinical trajectory preceding Alzheimer's dementia. We suggest that the initial decline in cognition occurs at the onset of comparable Alzheimer's disease pathology in both groups, and is associated with immediate decline to dementia in the lower education group. In contrast, higher education protects against further cognitive decline for ∼7 years until pathology becomes more severe.