The present study aims to examine the relationship between birth cohort, gender and educational level on the one hand and cognitive ability on the other. The data was taken from populations studies with roots in the longitudinal Swedish H70 study, originally begun in 1971. All participants were aged 85 at the time of examination. They were tested with a broad battery of cognitive tests to measure their level of cognitive ability. The results revealed that education was the unique predictor variable predicting most of the variance in cognitive ability. This points to the strong effect education has on measures of cognitive ability.