On page 242 of his book Are We Getting Smarter?, James Flynn claims the original Wechsler Intelligence Scale (WISC) normed circa 1947.5 yields full-scale IQs 7.63 points higher than those of its revision (WISC-R) normed in 1972. Stats like that are known as the Flynn effect; the notion that IQ is going up at rate of 3 points per decade. Of course by definition the average U.S./U.K. IQ is always around 100, but tests must be constantly re-normed to keep it that way.

However I have long suspected (and largely proven) that the Flynn effect is exaggerated by a statistical artifact. For example, when comparing old and new versions of the Wechsler, half of the test subjects will take the old version first and the other half will take the new version first, so whatever practice effect from being tested twice affects both tests equally. However I had a hunch (as had A.S. Kaufman) that you get more of a practice effect from taking the WISC-R first than from taking the WISC first, thus WISC scores will be spuriously high when people take both tests.

Finally I found some obscure old paper that proves this right. White kids who took the WISC-R first scored 12 points higher on the WISC than WISC-R, but white kids who took the WISC first scored virtually the same on both tests! When you combine the two groups together, the WISC appears 6 points easier than the WISC-R, but that difference is entirely driven by the subset of the sample that took the WISC-R first

A similar (though less extreme pattern) is found for non-whites.

Source: A comparison of the WISC and WISC-R by order of administration
Michael Murphy