The effect was found in the top 5% at a rate similar to the general distribution, providing evidence for the first time that the entire curve is likely increasing at a constant rate. The effect was also found for females as well as males, appears to still be continuing, is primarily concentrated on the mathematics subtests of the SAT, ACT, and EXPLORE, and operates similarly for both 5th and 6th as well as 7th graders in the right tail.
In other words, the Flynn effect doesn’t appear to be solely caused by rising scores among the lowest quartile. Rather, it seems to be just as prevalent among the top 5 percent. The smartest are getting smarter.
(Thanks to gwern.)