From the Whitehall II study, data involving 5431 older participants (45-69 at baseline) has revealed a significant effect of midlife sleep changes on later cognitive function. Sleep duration was assessed at one point between 1997 and 1999, and again between 2002 and 2004. A decrease in average night’s sleep from 6, 7, or 8 hours was significantly associated with poorer scores on tests of reasoning, vocabulary, and the MMSE. An increase from 7 or 8 hours (but not from 6 hours) was associated with lower scores on these, as well as on tests of phonemic and semantic fluency. Short-term verbal memory was not significantly affected. The magnitude of these effects was equivalent to a 4–7 year increase in age.

Around 8% of participants showed an increase from 7-8 hours of sleep over the five-year period (7.4% of women; 8.6% of men), while around a quarter of women and 18% of men decreased their sleep amount from 6-8 hours. About 58% of men and 50% of women reported no change in sleep duration during the study period. Some 27% of the participants were women.

The optimal amount of sleep (in terms of highest cognitive performance) was 7 hours for women, closely followed by 6 hours. For men, results were similar at 6, 7 and 8 hours.

Analysis took into account age, sex, education and occupational status. The Whitehall II study is a large, long-running study involving British civil servants. Sleep duration was assessed simply by responses to the question "How many hours of sleep do you have on an average week night?"

A very large Chinese study, involving 28,670 older adults (50-85), of whom some 72% were women, also supports an inverted U-shaped association between sleep duration and cognitive function, with 7-8 hours sleep associated with the highest scores on a delayed word recall test.

I would speculate that this finding of an effect of short-term verbal memory (in contrast to that of the Whitehall study) may reflect a group distinction in terms of education and occupation. The Whitehall study is the more homogenous (mostly white-collar), with participants probably averaging greater cognitive reserve than the community-based Chinese study. The findings suggest that memory is slower to be affected, rather than not affected.