This is kind of a weird explanation of g. What g really means is that you can't design a meaningful intelligence test that doesn't correlate with all the other

This is kind of a weird explanation of g. What g really means is that you can't design a meaningful intelligence test that doesn't correlate with all the other intelligence tests - with slightly divergent axes for verbal and spatial/mathematical skills. Thus, for instance, backward digit span correlates with Raven's matrices, even though these tasks have nothing obvious in common. Someone who's good at one will be good at the other. It's also very hard to produce training/educational effects that show an effect on g, though dual N-back perhaps has some promise: http://en.wikipedia.org/wiki/N-back http://en.wikipedia.org/wiki/N-back However, these kinds of brain exercises have very little to do with education as we know it. The obvious null hypothesis is that we're looking at a physiological effect, such as the quality of myelin insulation in neurons. Obviously, we see the same correlation effect in CPU benchmarks - any benchmark at all will reveal that a Xeon is faster than a Celeron. The obvious null hypothesis is that the Xeon has smaller transistors and more of them. The causality behind neurological g is probably something just as crude and straightforward. One could argue, however, that when we compare Xeon motherboards to Celeron motherboards, we see faster DRAM and the like. Perhaps it's the fast CPU's environment, rather than its lithography, that makes it faster. But... this isn't an argument anyone would make without a strong prior conviction that all CPUs are created equal. It's unclear where such an idea comes from in the case of the human brain, but it doesn't seem evidentiary in nature.