Thinking positively: The genetics of high intelligence

I would be very surprised if high intelligence was anything other than the extreme edge of a normal distribution of the human population. For it to be anything other than this it would require people of high intelligence to be a sub-population that did not breed with the rest of humanity.

You could make this argument for any trait. However, some traits are the result of a single gene - e.g., sickle cell anemia and the accompanying malaria resistance. Yet some of these traits occur in large populations that are not strongly inbred.

Only single gene traits. Intelligence (however defined) is multi-genetic - there are thousands of genes that contribute to intelligence. Given this the only way that individuals with high intelligence could be anything other than edge of a normal distribution is if they were part of a human sub-population.

Edit. I should add that the humans are not completely one population because of genetic isolation and differential selection (especially over the last 10,000 years), but we are almost a single population. Like everything in genetics it gets fuzzy at the edges.

"there are thousands of genes that contribute to intelligence"

And the very definition of "intelligence" is incredibly complex and slippery, which is one of the reasons why I've always found trying to summarise such a complex property into a single numerical value such a silly exercise.

Summarizing a complex property into a single numerical value is very useful and not silly. For example, temperature is a numerical summary of huge number of molecular motions.

More on this here: http://slatestarcodex.com/2014/08/11/does-the-glasgow-coma-s...

What we mean in common language by intelligence may not be a single property. It might be more analogous to smell than temperature. An inherently multi-dimensional phenomena that can't be accurately described by a scalar measure.

It's actually not that difficult to turn a multidimensional measurement into a scalar quantiry; just represent the measurement as a vector composed of the deviation from median on each axis, then find the magnitude of the vector

But you're right that this doesn't tell you much about any of the individual dimensions. Maybe adding a variance across dimensions, so two numbers, would be more useful.