Most theories of human mental evolution assume that selection favored higher intelligence and larger brains, which should have reduced genetic variance in both. However, adult human intelligence remains highly heritable, and is genetically correlated with brain size. This conflict might be resolved by estimating the coefficient of additive genetic variance (CVA) in human brain size, since CVAs are widely used in evolutionary genetics as indexes of recent selection. Here we calculate for the first time that this CVA is about 7.8, based on data from 19 recent MRI studies of adult human brain size "in vivo": 11 studies on brain size means and standard deviations, and 8 studies on brain size heritabilities. This CVA appears lower than that for any other human organ volume or life-history trait, suggesting that the brain has been under strong stabilizing (average-is-better) selection. This result is hard to reconcile with most current theories of human mental evolution, which emphasize directional (more-is-better) selection for higher intelligence and larger brains. Either these theories are all wrong, or CVAs are not as evolutionarily informative as most evolutionary geneticists believe, or, as we suggest, brain size is not a very good index for understanding the evolutionary genetics of human intelligence.