Reports on cognitive decline with age have, over the years, come out with two general findings: older adults do significantly worse than younger adults; older adults are just as good as younger adults. Part of the problem is that there are two different approaches to studying this, each with their own specific bias. You can keep testing the same group of people as they get older — the problem with this is that they get more and more practiced, which mitigates the effects of age. Or you can test different groups of people, comparing older with younger — but cohort differences (e.g., educational background) may disadvantage the older generations. There is also argument about when it starts. Some studies suggest we start declining in our 20s, others in our 60s.
One of my favorite cognitive aging researchers has now tried to find the true story using data from the Virginia Cognitive Aging Project involving nearly 3800 adults aged 18 to 97 tested on reasoning, spatial visualization, episodic memory, perceptual speed and vocabulary, with 1616 tested at least twice. This gave a nice pool for both cross-sectional and longitudinal comparison (retesting ranged from 1 to 8 years and averaged 2.5 years).
From this data, Salthouse has estimated the size of practice effects and found them to be as large as or larger than the annual cross-sectional differences, although they varied depending on the task and the participant’s age. In general the practice effect was greater for younger adults, possibly because younger people learn better.
Once the practice-related "bonus points" were removed, age trends were flattened, with much less positive changes occurring at younger ages, and slightly less negative changes occurring at older ages. This suggests that change in cognitive ability over an adult lifetime (ignoring the effects of experience) is smaller than we thought.
Full text of the article available at http://www.apa.org/pubs/journals/releases/neu-24-5-563.pdf
 . Influence of age on practice effects in longitudinal neurocognitive change. Neuropsychology [Internet]. 2010 ;24(5):563 - 572. Available from: http://psycnet.apa.org/journals/neu/24/5/563/