Regression to the mean deals with random error in the measurement of something. It tells us nothing about that something.
It only applies to measurement, not the particular cause and effect of a given thing you are studying. The cause and effect have to be established before we can measure and then say what any mean is.
http://en.wikipedia.org/wiki/Regression_toward_the_meanIn talking about IQ, we can say someone's IQ is Z and on any given day they will test within X points of it. As they take the test over time the scores will regress to the mean of Z.
As for kids. It has already been shown that intelligence is both heritable and is affected by environment. There is also the Flynn Effect. Thus one would expect that children of highly intelligent and committed parents would also be intelligent, probably more so.
Based on the above, "families" with parents of similar traits do not "slide back" to "average" which is the underlying assumption of misusing "regression to the mean" to describe a particular person's performance.
Another way it could be misused is the "all kids catch up" by third grade. Intelligence is not something that goes away as kids get older all things being equal. IQ remains as a fundamental part of their individuality.
