2 ms·
I think you're misunderstanding. Traits like 'intelligence' are always 'expressed'. They aren't binary and ~100% determined by genetics like gender. You don't s
by gwern 10y ago
I think you're misunderstanding. Traits like 'intelligence' are always 'expressed'. They aren't binary and ~100% determined by genetics like gender. You don't select for 'would be expressed', you select for something like 'on average, has 0.1SD higher measured trait values'.
Even a trait which seems to be binary, like schizophrenia, is actually probably on a continuum (relatives of schizophrenics have lower IQs, many people have symptoms which wouldn't rise to the level of a full diagnosis) and can be seen as a liability threshold model where a large number of genes each have deleterious effects, yielding a normal distribution of risk and the higher your risk, the more likely random chance or an environmental insult will push you into a full-fledged schizophrenia. So if you were selecting on mental illness risk, it would be impossible to select for someone who 'does' or 'does not express' schizophrenia, but it would be perfectly meaningful to talk about halving or doubling the risk, which would translate to 0.5% vs 2% (in almost all of the people selected for less schizophrenia risk, they wouldn't have had schizophrenia either way; what it buys you is 1 less future out of 200 futures in which the person becomes schizophrenic).