The general intelligence factor, g for short, is a latent variable that explains about half of the variance in intelligence.
Its construction stems from the positive manifold—the observation that cognitive subtest results are positively correlated.
It is constructed through Factor analysis, often as the first unrotated factor. It is a population-level latent variable and is not directly measurable for individuals. It is however possible to construct IQ tests that are heavily g-loaded, with correlations over 0.9.
Some items are more g-loaded than others. Strongly g-loaded subtests include Raven’s progressive matrices, matrix reasoning, backward digit span memorization—while weakly g-loaded subtests include simple reaction time, spelling, forward digit span memorization.
g is positively correlated with brain volume, even when controlling for height1.